EY Forward Deployed
↑↓ spine   ←→ detail   space read everything
Interview command center · prepared 30 Oct 2026 · for Johnny Chung

EY Forward Deployed

Forward Deployed Engineer, Applied AI, Senior Manager, Financial Services Consulting. Ten spine slides tell the story top to bottom; each opens rightward into its detail and examples.

9
people researched from the hiring post, chain of command mapped
3
FDE Applied AI levels open in FSO: Senior, Manager, Senior Manager
$172.8K–374K
posted NYC base range for Senior Manager (national $144K–329.1K)
Oct 18
requisition posted; refreshed on LinkedIn Oct 30 with 25 applicants
↓ next spine slide · → detail for the current slide · space walks every slide in order · Home back hereFacts tagged verified came from ey.com or primary sources; unverified items came from search snippets or inference.
The read · what this is

This is a governance-first AI practice hiring an architect who has sat on the client's side of the model-risk table

What they are building

  • A forward-deployed Applied AI team inside EY's Financial Services Organization (FSO), hiring at three levels at once (Senior, Manager, Senior Manager), reposted Oct 18–28 in Charlotte, Dallas, New York, Tampa.
  • It sits in FS Technology Consulting under Mudit Gupta's FS Consulting AI practice. His earlier hiring post called it the "Financial Services Technology Consulting Office."
  • Context: EY's AI services sales grew 49% in FY26; the May 2026 Microsoft deal ($1B+ over five years) names FDEs as the delivery mechanism.

What the hiring leaders believe, in their words

"In practice, governance is what makes these systems deployable at scale."Mudit Gupta, PYMNTS, 22 Jun 2026
"Differentiation shifting to domain-specific data, workflow design, and the control layer."Scott Keipper, 13 May 2026
"Without a platform-based approach, banks risk creating 15 versions of the same AI function."Sameer Gupta, ey.com, 6 Apr 2026

Why you fit

  • Eight years inside two banks' risk functions as the person who built the analytical platform, wrote the documentation, and got it through independent validation and in front of the board.
  • Eight years at Accenture running onshore and offshore delivery for Fortune 20 clients, then data leadership at GroupM and Havas.
  • 2026 retooling into hands-on agentic engineering: multi-agent systems with eval harnesses, guardrails, audit logs, and tested cost invariants. You know what the CRO signs and what validation asks, and you can build the thing.
Sources careers.ey.com req 1393575733; ey.com FY26 results (8 Oct 2026); EY–Microsoft release (21 May 2026); PYMNTS; aimmediahouse citing Business Insider; ey.com insights. Right → the role decoded, then where it sits.
The read · the role decoded

Senior Manager here means architect, govern, advise, and mentor, with building as the proof rather than the job

LevelYearsNYC rangeCenter of gravity
Senior3+$128.4–200.6KBuilds LLM, RAG, agentic components; "supports delivery"
Manager7+$150.7–261.6KLeads multidisciplinary teams; MLOps and CI/CD governance
Senior Manager10+$172.8–374KReference architectures, standards, client platform strategy

One thing the posting gets wrong

It lists SR 11-7 as a preferred framework. The Fed, OCC and FDIC superseded it with SR 26-2 on 17 Apr 2026 and scoped GenAI and agentic AI out of it. Mention it once, lightly, as "SR 26-2, formerly 11-7." With a hiring leader who came up through audit and risk, that lands.

What the Senior Manager posting repeats

  • Define and govern reference architectures and engineering patterns for AI/ML, GenAI, RAG and agentic systems; own end-to-end architecture incl. service boundaries, orchestration layers, data flows, security controls.
  • Evaluation and LLM Ops standards: scalable approaches to model evaluation and benchmarking, lifecycle governance, monitoring and observability tooling.
  • Advise clients on multi-year AI architecture strategy: platform investment, build vs operate, sequencing; reduce duplication and create reusable capabilities.
  • Lead large program delivery and mentor senior engineers and technical leads.
  • Use "Claude Code, Codex, or equivalent agentic coding platforms" in day-to-day engineering. Named in all three levels.
  • Soft bar: "sound judgment in regulated environments," comfort with ambiguity, explain trade-offs to non-technical audiences.
  • Preferred: consulting background; SR 11-7, FINRA, Reg BI; Databricks, Kafka, lakehouse patterns; GPU-accelerated workloads; small-language-model design patterns; multimodal familiarity.
  • Hybrid: 40–60% in person over an engagement. New York is one of the four cities.
Sources careers.ey.com reqs 1393575733 (SM), 1393540633 (Senior), Manager via freehire mirror; Fed SR 26-2 (17 Apr 2026); Sullivan & Cromwell memo Apr 2026.
The read · where it sits

The nine names are a chain of command with AI leaders beside it and sector sponsors around it, not a panel of peers

Sid Khosla
EY Americas Financial Services Vice Chair since 9 May 2026. About $9B practice, about 14,000 people. Ex Americas BCM Leader, ex FSO Strategy & Transactions MP.
Ron Giammarco
Americas FS Deputy Vice Chair and Americas Consulting FS Leader (Banking & Capital Markets, Wealth & Asset Mgmt, Insurance, Real Estate). FinCrime technology roots.
Scott Keipper
Americas FS Technology Consulting Leader; EY US Governing Board. Tampa. Owns the FS Databricks relationship publicly.
Mudit Gupta posted the job
Partner, Americas FS Consulting AI Practice Leader; leads engineers, architects, data scientists; leads the FS Databricks business. Most likely hiring manager.

AI leaders beside Mudit

Phil Andriyevsky "Phil A." ~90%
Partner, Americas FS AI & Data Leader. Go-to-market, alliances (Snowflake voice), AI/data product. Incubated EY's WAM AI & data business.
Sameer Gupta
Principal, Americas FS AI Leader. Data scientists, LLM and agentic frameworks, knowledge graphs. Author of the EY banking and WAM GenAI surveys.

Also on the FS GenAI page

  • Rani Bhuva, Americas FS Responsible AI Leader
  • Abhinav Goel, Americas WAM AI Leader
  • Harish Arora, MD, FS Technology Consulting

Adjacent sponsors and consumers

Paul Sussex
Principal, Americas Digital & FS Cloud Leader; now "Integrated Solutions Leader, Productivity & Technology, FS." ServiceNow disputes and fraud ops offerings. Navy veteran, CISSP.
Gurdeep Batra
Managing Partner, Americas FS Strategy, Innovation & Client Service Leader. Formerly FS Consulting Deputy Leader and Global/Americas WAM Consulting Leader. Irvine.
Adam Girling
Principal; Americas Banking & Capital Markets Sector Leader (2026). Regulatory capital, risk and finance transformation. Cambridge mathematics. Not an AI practitioner.
Note Titles verified on ey.com bios. Reporting lines are inferred from titles and from Mudit's "FS Technology Consulting Office" hiring post. Confirm who owns the FDE pod in the recruiter call.
Your positioning · the thesis and the intro

Lead with what you built inside the bank, then the engineering, then the consulting years

The 90-second intro, in your voice

I've spent the last eight years inside bank risk functions, at MUFG and then US Bank, as the person who builds the analytical platform and then gets it through model validation and in front of the board. At US Bank I replaced a $1M vendor commitment with an in-house loss-simulation platform, written in Python with Claude Code agents on Snowflake pipelines. It's now one codebase shared by the risk, credit and modeling teams, and I wrote the documentation that got it through independent validation.

Before banking I spent eight years at Accenture running onshore and offshore analytics teams for Fortune 20 telecom and media clients, then led data teams at GroupM and Havas. So the client-facing half of this role is familiar ground.

This year I went deep on agentic engineering through Gauntlet's program. I shipped a multi-agent red-teaming platform with an eval harness and a cost invariant enforced by a test, and a clinical copilot with a deterministic verifier and an 80-case regression suite. At home I run Claude Code across a five-machine fleet with local models. The combination I bring is the one your posting describes: I know what a CRO will sign and what validation will ask, and I can build the system that produces that evidence.

Three rules for the room

  • Never say "not a SWE." At Riverside you said it four times and it became the story. Say what you built and governed, and name the gaps as gaps.
  • Workflow where precision matters, agent where judgment matters. Say it in that order. At Riverside you said "we'd need LangGraph to scale," which inverted it.
  • Translate climate every time: scenario, loss, PD and LGD, CECL reserve, capital. The hazard is an input; the pipeline is credit risk.

One-sentence thesis

You are the rare FDE who has been the client: the risk VP who built, validated and defended the platform. That is what "forward deployed into a bank" needs most, and the engineering is now current.

Sources Resume v4 EY FDE (Downloads, regenerated 30 Oct 2026); Riverside interview notes (Jun 2026). Right → proof mapped to the posting, then the gaps and bridges.
Your positioning · proof

Every must-have in the posting has a named artifact behind it

Posting requirementYour evidenceSay it as
Reference architectures, engineering standards, design reviewsWritten project guardrails (CONSTITUTION articles) every change is checked against before merge; planner, architect, coder, reviewer, tester, deployer pipeline with human approval gates and an independent reviewer agent; standing gap log that scopes the next iteration."I govern builds with written guardrails and gates, and the reviewer is independent of the implementers."
Evaluation standards, LLM Ops, observability80-case clinical eval suite with golden-hash regression, full tracing, per-run cost and token metering; AgentForge all-local inference invariant enforced by a test; model-drift and cost monitoring on a five-machine fleet."Evals are the acceptance test the validator reads. Cost is a tested invariant, not a dashboard."
Enterprise knowledge AI, RAGAI risk and regulatory news monitor at US Bank: 9 risk categories, relevance scoring, Claude briefs, auditable monthly archives tracing every headline to source. RepIntel: 1.15M of 13.65M CFPB complaints matched to 23 banks."Grounded, auditable retrieval built for a risk team that gets examined."
Agentic frameworks, multi-agent orchestration, memoryAgentForge: orchestrator, red team, dual judge, documentation agent. Cortex: five persistent assistants on private infrastructure with MCP memory tools and health monitoring. LangGraph clinical copilot with deterministic verifier."Bounded agents with explicit tool permissions, budgets and a verifier in front of the output."
Judgment in regulated environments, model riskUS Bank platform passed independent model validation; MUFG Board Risk Committee exposure report on a $250B+ portfolio; GARP SCR; SR 26-2 and FINRA 2026 fluency."I've written the documentation validation signs and presented the result to a board committee."
Client advisory, consulting, build vs operate8 years Accenture (Fortune 20 telecom, media); vendor RFPs and methodology assessment at MUFG (MSCI, Oliver Wyman, S&P); build-vs-buy case won at US Bank; pitched and ran P&L at Havas."I've sat on both sides of the build-vs-buy decision and run the vendor lifecycle."
Python, data platforms, cloud, GPU workloadsPython, SQL, SAS; Snowflake pipelines; AWS, Azure; FastAPI, Docker; NVIDIA DGX Spark (GB10), vLLM, Ollama, MLX; model routing across Claude, OpenAI, Qwen, Nemotron, DeepSeek with fallback chains."Local and frontier models routed by task, with the cost math to show for it."
Claude Code and agentic coding at enterprise scaleDaily driver since early 2026; open-sourced build workflow adopted by much of the Gauntlet cohort; Codex cross-verifies each change."Happy to screen-share the pipeline rather than describe it."
Sources Resume v4 EY FDE; Gauntlet project notes; cortex and projects directories.
Your positioning · gaps

Four gaps to manage, each with a one-line bridge and a boundary you will not cross

1 · Enterprise-scale platform operations

Gap: Kubernetes, service mesh, API governance at bank scale are not your depth.

Bridge: "I set the standards and the gates; I've run onshore and offshore engineering teams that operate at scale, and I ship FastAPI and Docker services myself." Name Kubernetes and Kafka as known, not mastered. The Senior and Manager FDEs under you carry that depth.

2 · Databricks (Mudit leads the FS Databricks business)

Gap: Your pipelines are Snowflake.

Bridge: Speak lakehouse concepts (Delta, Unity Catalog, MLflow) and your MUFG enterprise data catalog implementation. Phil Andriyevsky is EY's Snowflake voice, so ask how FS clients split between the two rather than pretending.

3 · Microsoft stack: AutoGen, Semantic Kernel, Azure AI Foundry, Fabric

Gap: Your orchestration is LangGraph, MCP and custom.

Bridge: "Frameworks are interchangeable; the control layer isn't." Know by name what EY runs on: Azure AI Foundry and Fabric (orchestration), NVIDIA NIM and NeMo Guardrails (EY.ai Agentic Platform), Copilot Studio.

4 · Banking domains beyond credit and climate

Gap: Fraud, AML, disputes and wealth are not delivery experience.

Bridge: Lead with credit loss, CECL, board reporting, vendor lifecycle and RepIntel. For FinCrime (Giammarco), disputes (Sussex) and wealth (Batra, Phil) carry one architecture sketch each and say "here is how I'd structure it," never "I've delivered it."

Also from the Riverside loop (Jun 2026) you didn't volunteer the data-quality story, didn't lead with having built on their business, and fumbled names. Fixes live in the positioning, architecture and don'ts rows.
The people · who you are talking to

Nine people, three roles in the hire: the chain that owns it, the AI leaders beside it, and the sponsors around it

owns the hire
Mudit Gupta
Partner, FS Consulting AI Practice Leader · posted the job
Will probe: runtime control of agents, getting past proof-of-concept paralysis, Databricks, leading engineers.
beside
Phil Andriyevsky
Partner, FS AI & Data Leader · "Phil A."
Will probe: pilot to production, operating model and workforce, platform over point solutions, can you help sell.
beside
Sameer Gupta
Principal, FS AI Leader
Will probe: reusable platform versus bespoke, ROI attribution, agentic versus RPA, model routing and cost.
owns the org
Scott Keipper
FS Technology Consulting Leader · US Governing Board
Will probe: production inside systems of record, the control layer as architecture, cross-sector reuse, delivery discipline.
sponsor
Paul Sussex
Digital & FS Cloud Leader · Integrated Solutions
Will probe: security-first LLM design, embedding AI into workflow platforms, cloud economics, real Claude Code fluency.
sponsor
Gurdeep Batra
FS Strategy, Innovation & Client Service Leader
Will probe: the revenue outcome before the stack, wealth fluency, enterprise operating model, client posture.
sponsor
Adam Girling
Banking & Capital Markets Sector Leader
Will probe: regulatory constraints on AI, data lineage and quality, what a CRO or CFO would sign.
owns consulting
Ron Giammarco
FS Deputy Vice Chair · Consulting FS Leader
Will probe: delivery economics of the FDE model, FinCrime and KYC, auditability for examiners, talent.
owns FS
Sid Khosla
Americas Financial Services Vice Chair
Will probe: total cost of agents, which 20% of use cases, when not to use GenAI, trust as the scaler.
Right → one slide per person, in this order: who they are, what they say publicly with dates, what they will probe, your angle, and one question to ask them.
The people · the poster

Mudit Gupta came up through audit analytics and risk, so he hears "governance" as the enabler, not the brake

Mudit Gupta, CPA
Partner, EY Americas Financial Services Consulting AI Practice Leader; leads the FS Consulting Databricks business
Where
New York. CPA licensed NY and NJ; CGMA u
School
Rutgers BS Accounting + Computer Science (honors); Harvard Business School alumnus (program type unstated)
Team
"Engineers, architects and data scientists" implementing AI and data platforms in regulated industries, across Banking, Insurance, Wealth
Path: continuous-auditing and audit-analytics work 2007–2010 (Rutgers CarLab symposium, MetLife) u; FS AI & Data Risk Leader by Mar 2024; FS Consulting AI Practice Leader 2025–26. LinkedIn articles: AI in FS Internal Audit, continuous auditing. Likely the hiring manager.

What he says

Leaders think agentic AI "could advance faster without governance constraints... But in practice, governance is what makes these systems deployable at scale."PYMNTS, 22 Jun 2026
Synthetic data is "often wrongly assumed to be inherently safe" and can leak sensitive signals or replicate historical bias.PYMNTS, 22 Jun 2026
"Effective governance and runtime control of agentic AI will become a competitive advantage, not merely a compliance requirement."LinkedIn, undated u
Banks "start with productivity because it's low risk... establish proof points... then move on to transformation."Banking Dive, 2026

Also: Mar 2024 EY webcast on GenAI in risk and compliance with Adam Girling; EY was headline sponsor of the Evident AI Symposium NY, 22 Oct 2026 u.

What he will probe

  • Runtime control of agents: permissions, approval gates, kill switches, audit trail.
  • Getting a client past "proof of concept paralysis" into production.
  • Databricks fluency; data platform patterns in regulated settings.
  • Can you lead engineers and architects, not just prototype.

Your angle

  • You are an auditor's candidate: documentation that passed validation, archives that trace every headline to source, invariants enforced by tests.
  • Tell the US Bank validation story and the AgentForge audit-log story in his vocabulary: "runtime control."
  • Ask: "How is the FDE pod structured relative to your practice and Phil's AI & Data org, and what do the first 90 days at a client look like?"
Sources ey.com/en_us/people/mudit-gupta; EY FS GenAI offerings page; EY webcast Mar 2024; PYMNTS 22 Jun 2026; Banking Dive; LinkedIn snippets (login-walled).
The people · "Phil A."

Phil Andriyevsky owns go-to-market for FS AI and data, so he will listen for whether you can help sell what you build

Phil Andriyevsky
Partner, EY Americas Financial Services AI and Data Leader identity ~90%
Where
New York metro (inferred)
School
BS Finance and Information Systems Management, University at Albany
Remit
GTM strategy, revenue, team scaling, partnerships, AI/data product; the CDO and CTO agenda end to end
Before
Principal, Asset & Wealth Management Data & Analytics Leader; incubated EY's WAM AI & data business; 18+ years
Why "Phil A.": the LinkedIn slug phil-a-6988154 indexes as "Partner, Financial Services AI and Data Leader," his exact ey.com title, and he is co-listed with Mudit and Sameer on EY's FS GenAI page. No other EY FS "Phil A." surfaced. Leads EY FS veteran integration; Pat's Run sponsor 10+ years.

What he says

About 95% of AI pilots never reach production; 84% of desk workers embrace agentic AI while 56% worry about job security. AI in FS is "ultimately a workforce transformation."ey.com, "How AI in financial services is transforming operating models," 5 May 2026
"Digital workers" compress the asset-management fund-accounting cycle from weeks to hours.ey.com, 8 Jun 2026, with Kaimowitz and Dubey
Governed AI personalization in wealth using small language models: about 10x lower latency and advice-detection guardrails.ey.com, 24 Sep 2026 (Goel, Mangla, O'Reilly), the FS case study of record

EY's Snowflake voice: 2026 FS AI & Data Predictions webinar (14 Jan 2026), Snowflake Summit panel (Jun 2025). AI Summit NY Dec 2025 panel "The AI value gap": platforms over isolated use cases.

What he will probe

  • Pilot to production: what changed the last time you crossed that line.
  • Operating model and workforce change, not just the system.
  • Platform and data products versus point solutions; Snowflake.
  • Commercial instinct: can you scope and help close the next phase.

Your angle

  • Your US Bank story is adoption, not just a build: you embedded Claude into the team's research workflow and coached them to extend it; they now build their own Power Automate agents.
  • Snowflake pipelines at US Bank are a direct match.
  • Ask: "The September piece chose small models for latency and guardrails. How often do FDEs pick small over frontier in production, and who owns that call?"
Sources ey.com/en_us/people/phillip-andriyevsky; ey.com insights 5 May, 8 Jun, 24 Sep 2026; snowflake.com webinars; ey.com AI Summit NY page; InvestOps speaker bio; LinkedIn search snippets.
The people · FS AI leader

Sameer Gupta wants one platform instead of fifteen versions, and he spoke about model gateways and cost four days ago

Sameer Gupta
Principal, EY Americas Financial Services AI Leader (HBR bylines him "Global FS AI" u)
Where
New York
School
MBA Finance & Strategy, IIM Bangalore; BTech Electrical, IIT Delhi
Team
Data scientists, AI practitioners, domain pros: deep learning, knowledge graphs, LLMs, agentic frameworks; customer risk, fraud, compliance automation
Before
EY since 2014 u; partner and head of consumer finance, risk and fraud at Opera Solutions (2003–2014) u
Do not confuse with Sameer Gupta, EY India Tax Leader, or the Apollo CDAO of the same name.

What he says

"Without a platform-based approach, banks risk creating 15 versions of the same AI function, just like they once used 15,000 spreadsheets." "The biggest barrier to scaling AI isn't the algorithm, it's the change management."ey.com, "Five hallmarks of effective AI strategies in banking," 6 Apr 2026
"The biggest challenge is not measuring usage, it is proving attribution."CNBC, 5 May 2026
Agentic AI makes "large-scale process automation that was not possible with rules-based approaches like robotic process automation" possible.MIT Technology Review Insights with EY, 4 Sep 2025
Banks use "gateways to route workloads to models best suited" to each task; cost control is technology plus behavior.CIO Dive, Evident AI Symposium panel, 26 Oct 2026

What he will probe

  • Reusable capability (gateway, eval harness, agent framework) versus bespoke builds.
  • Governance and privacy as engineering, closed models on trusted platforms.
  • ROI attribution: what the CFO counts as realized value.
  • Agentic process automation versus RPA; model routing and cost.

Your angle

  • Your fleet routes Claude, OpenAI, Qwen, Nemotron and DeepSeek with fallback chains; AgentForge held platform inference at $0 with a tested invariant. That is his gateway and cost theme with receipts.
  • Attribution: the US Bank platform's outputs feed CECL reserves and credit decisions. It is used, not admired.
  • Ask: "What evidence does an FDE engagement have to produce for the client's CFO to count it as realized value?"
Sources ey.com/en_us/people/sameer-gupta; ey.com 6 Apr 2026; EY-Parthenon GenAI in Banking survey 19 Sep 2025; GenAI in WAM survey 16 Sep 2025; MIT TR Insights; HBR sponsored Jan 2026; CNBC; Global Finance; CIO Dive.
The people · FS technology consulting leader

Scott Keipper runs the technology org the team sits in and judges AI by whether it lives inside systems of record

Scott Keipper
Principal; EY Americas Financial Services Technology Consulting Leader; EY US Governing Board; Global Governance Council
Where
Tampa
School
MS Accounting, Florida State
Scope
End-to-end technology delivery for FS: platforms, AI, data engineering, cybersecurity, large transformation programs
Honors
Consulting Magazine 2024 Leaders in Technology Consulting
Path: pre-EY career and start year are not public. Fronts EY's FS Advanced Analytics page (with Sameer) and was the EY voice when Databricks named EY its 2024 FS Partner of the Year. An Eric Keipper is an EY MD in Government & Public Sector; different person.

What he says

"For banks and the startup ecosystem, we expect consolidation around a smaller set of core model providers, with differentiation shifting to domain-specific data, workflow design, and the control layer."On Anthropic's finance agents, 13 May 2026 (Business Insider via aimmediahouse)
Products that integrate into existing systems of record and work inside existing governance and risk architecture will win.Same interview, paraphrased
"Builds high-performing teams dedicated to solving complex technology challenges in financial services... As a leader, I try to set an example for my teams by challenging myself to go beyond what I think I'm capable of."ey.com bio

No bylined articles or podcasts under his name. A LinkedIn snippet says he hosted the FS panel at AI Summit NY, Dec 2025 u.

What he will probe

  • Production inside the client's systems of record and existing MRM and governance.
  • The control layer as architecture, not slides: evals, guardrails, observability, lifecycle.
  • Reuse across banking, insurance and wealth; reference architectures and standards.
  • Delivery discipline on big programs, team building, and whether you can run a client room.

Your angle

  • "One codebase shared by risk, credit and modeling" is integration into the client's operating fabric. Say it that way.
  • Your control layer: eval gates, guardrails, audit log, HITL, tested cost invariant. Map it to his phrase.
  • Ask: "What does EY's reusable control layer look like across FS clients today, and where do FDEs extend it rather than rebuild it?"
Sources ey.com/en_us/people/scott-keipper; consultingmag.com 16 May 2024; aimmediahouse.com 13 May 2026; EY Databricks FS Partner of the Year release (Aug 2024); EY FS advanced-analytics page.
The people · digital and cloud

Paul Sussex is a security-credentialed cloud leader who took Claude Code 101 in July, so show the workflow instead of describing it

Paul J. Sussex
Principal; EY Americas Digital and FS Cloud Leader; LinkedIn: "Productivity and Technology, Integrated Solutions Leader for Financial Services" u
Where
New York
School
Master of Information Technology, Virginia Tech
Creds
CISSP, CISA, CRISC, CIPP/IT, ITIL, PMP; Kellogg exec ed "AI Applications for Growth" (Feb 2024); Anthropic "Claude Code 101" (Jul 2026) u
Also
US Navy veteran; inventor on US patent 12,639,705 (digital-wallet key-shard backup, granted 26 May 2026); Wreaths Across America, NPower, Girls Who Code
Path: 25+ years; management roles at an unnamed "leading multinational technology corporation" before EY (not Accenture, no evidence). EY contact for the ServiceNow FS offerings: Fraud Operations, Disputes, Vulnerability Management.

What he says

"When a customer calls the bank about a charge they don't recognize, they're not just filing a complaint. They're testing whether their bank has their back." "It is about redesigning workflows so automation handles repeatable decisions."ey.com, "Why financial dispute management is a smart AI starting point for banks," with Jennifer Lucas, 26 May 2026
"Dispute operations rarely fail because of a single task. They fail at the handoffs between tasks."Companion piece by Surabhi Prasad, 27 Jul 2026

Numbers he uses: $9–10 processing cost per dispute; chargebacks to 324M by 2028; 30–50% manual-processing reduction from modernization. ServiceNow disputes offering claims 30–40% claims-volume reduction.

What he will probe

  • Security-first LLM design: data handling, access control, auditability, prompt injection.
  • Embedding AI into operational workflow platforms (ServiceNow, case management, core) rather than bolting on a chatbot.
  • Cloud and cost economics of deployed AI.
  • Coaching and growing engineers; real agentic-coding fluency.

Your angle

  • AgentForge is AI security tooling by construction: prompt injection, exfiltration and tool-misuse probes with a tamper-evident log. Lead with it.
  • Offer to screen-share your Claude Code pipeline with its approval gates. He has taken the course; he will recognize real use.
  • Ask: "Integrated Solutions launched in August. How does an FDE build hand off into a managed service like the ServiceNow disputes offering?"
Sources ey.com/en_us/people/paul-j-sussex; ey.com alliances pages (ServiceNow, disputes, fraud ops); Google Patents US12639705B2; EY Integrated Solutions release 27 Aug 2026; LinkedIn snippets.
The people · strategy, innovation, client service

Gurdeep Batra sells "business-led, technology-enabled" transformation and will want the revenue outcome before the stack

Gurdeep Batra
Managing Partner; EY Americas Financial Services Strategy, Innovation and Client Service Leader
Where
Irvine, CA
School
Master of Management, Devi Ahilya University; BE, College of Technology and Engineering, Udaipur u
Before
Americas FS Consulting Deputy Leader; Global and Americas WAM Consulting Leader. Pre-EY: Publicis Sapient EVP (2006–2018), Infosys (2000–2006) u
Builds
Key team member, EY Nexus for Wealth & Asset Management (interoperability layer, advisor desktop, CRM data flow)
Correction: he is no longer the WAM Consulting Leader. He sits above the WAM practice with a growth and client-service mandate. Three title variants circulate in 2026; the ey.com one is above.

What he says

An integrated, enterprise-wide approach to GenAI (strategy, infrastructure, risk governance, data, center of excellence, workforce) beats point use cases.ey.com, "The transformation imperative: GenAI in wealth and asset management," with Abhinav Goel, 31 Oct 2023
88% of financial institutions use more than five providers for wealth services; 54% cite change-management and implementation cost as the main risk.ey.com, "We are entering the Wealth as a Service era," 16 Jan 2025
"AI is the tip of the spear for differentiation in all aspects of asset management."EY Future of Asset Management Study, Jan 2026 (contact)

2025–26 LinkedIn: Executive Leadership AI Immersion Series in Silicon Valley; "EY's first AI Immersion Session for Wealth Management."

What he will probe

  • Can you connect a build to a C-suite growth agenda and a number.
  • Wealth fluency: advisor desktop, next-best-action, onboarding, investment ops, WaaS economics.
  • Enterprise operating model for AI versus pilot sprawl.
  • Client posture with a skeptical bank or wealth CTO; Sapient-style product delivery vocabulary.

Your angle

  • Open with the US Bank outcome (credit decisions, reserves, a $1M vendor displaced), then the stack.
  • Use product language he recognizes: PRD, workflow map, gap loop, agile delivery with gates.
  • Ask: "Do clients leave the AI Immersion sessions with a scoped first build, and is that where an FDE enters?"
Sources ey.com/en_us/people/gurdeep-batra (and en_ca, en_lt); linkedin.com/in/gurdeepbatra (public); ey.com insights 2021, 2023, 2025; EY Future of Asset Management Study PDF Jan 2026; International Banker 7 Oct 2024; directory snippets for dates.
The people · banking and capital markets

Adam Girling is a Cambridge mathematician who runs the banking sector, and he frames everything as safety and soundness

Adam Girling
Principal, FS Consulting; EY Americas Banking and Capital Markets Sector Leader (2026; some pages still say Deputy)
Where
New York
School
MA Mathematics, University of Cambridge
Tenure
Nearly 30 years in banking and capital markets; 20+ at EY
Focus
Regulatory capital (Basel), risk and finance transformation, data management, scenario planning
Correction: not wealth, not technology. He led Financial Performance and Risk solutions for US BCM, was Khosla's deputy, and appears to have succeeded him as BCM leader when Khosla became Vice Chair in May 2026 (inferred). He is a sector sponsor of the team, not an AI practitioner.

What he says

"The safety and soundness of the banking system is critical for the economy... our work to modernize and transform the industry both supports economic growth and furthers the resiliency of the banking system."ey.com bio
Regulators rescinded the 2023 Basel III Endgame on 19 Mar 2026 and replaced it with a new proposal (expanded risk-based approach, new standardized approach).ey.com, "US Basel III Proposal: what the changes mean," with Richard Tuosto, 16 Apr 2026
Move data from a risk and compliance burden to a value-driven asset; AI-ready data products; GenAI and agentic AI make data quality existential.ey.com, "How data management in FS creates value," with Khosla, Lele, Andriyevsky, 9 Apr 2026

Co-panelist with Mudit on the Mar 2024 GenAI in risk and compliance webcast.

What he will probe

  • Regulatory constraints on AI: model risk (SR 26-2), explainability, lineage and controls.
  • Whether a CRO or CFO would sign what you built; data quality in regulated pipelines.
  • Live capital and reporting topics: Basel proposal, RWA data remediation, regulatory reporting automation.

Your angle

  • Home turf. Scenario analysis, PD and LGD, CECL, Board Risk Committee reporting, model validation. Use the words "effective challenge" and "inventory."
  • Translate: scenario loss to reserve to capital. Then show the same event-exposure-decision-control pattern for fraud and ops.
  • Ask: "With SR 26-2 scoping out GenAI and agentic models, what are your banks using as the interim control standard, and what do you expect the RFI to change?"
Sources ey.com/en_gl/people/adam-girling (and en_us, en_au); ey.com insights 20 Sep 2023, 9 Apr 2026, 16 Apr 2026; EY risk and regulatory transformation page; EY webcast Mar 2024; LinkedIn snippets.
The people · consulting FS leader

Ron Giammarco runs all of FS Consulting from a financial-crime technology background and talks about how the work gets delivered

Ron Giammarco
EY Americas Financial Services Deputy Vice Chair; EY Americas Consulting Financial Services Leader
Where
New York (lives New Canaan, CT)
School
MBA, Columbia Business School; BBA Accounting, Baruch College
Tenure
30+ years, essentially career EY u
Before
Americas FS IT Consulting Partner; Americas Financial Crime Operations, Technology & Managed Services Leader (Cognitive Investigator platform)
Scope: technology, data and AI-led transformation across Banking & Capital Markets, Wealth & Asset Management, Insurance and Real Estate consulting, plus talent and culture for the Consulting division. Keipper, Sussex, Mudit and (formerly) Batra sit in his line; Tom Campanile is his current deputy.

What he says

"AI is reshaping financial services, challenging organizations to pair a growth mindset with the regulatory confidence required for lasting transformation."ey.com bio
"Innovative financial institutions are moving to a dynamic KYC process." The Americas are characterized by "synthetic identity, identity theft, and account takeover."The Paypers interview, 19 Nov 2020
Value comes from "rethinking how the work gets delivered," not just doing it faster.Yahoo Finance segment shared on LinkedIn, autumn 2026 u

Sep–Oct 2026 posts: the EY–Aaru behavioral-simulation alliance (replicated a 3,600-investor study in a day); frontier AI changing the cyber threat. Moderated a Sibos 2020 panel on real-time compliance.

What he will probe

  • Delivery economics: FDE versus staff augmentation; reuse, managed-services leverage, margin.
  • FinCrime is native ground: AML alert triage, false positives, dynamic KYC, model validation under BSA/AML.
  • Auditability and defensibility to OCC, Fed and FINRA examiners.
  • Talent and culture: upskilling client teams and EY juniors while shipping.

Your angle

  • Your ticket-to-deployed-app-in-a-week pipeline with gates is a "how the work gets delivered" story. Tell it as delivery model, not as tooling.
  • For FinCrime, sketch bounded agent triage of AML alerts with HITL and an audit trail. Structure, not claimed delivery.
  • Ask: "What does a healthy FDE engagement look like at month six: embedded team, managed service, or client-run?"
Sources ey.com/en_us/people/ron-giammarco (and en_id, en_lv); The Paypers 19 Nov 2020; Finextra Sibos 2020; EY Cognitive Investigator page; EY–Aaru release 30 Sep 2026; LinkedIn public page.
The people · FS vice chair

Sid Khosla runs the $9B FS practice and asks what agents actually cost to run, not what tokens cost

Sid Khosla
EY Americas Financial Services Vice Chair (effective 9 May 2026, succeeded Shawn Smith)
Where
New York
School
BS Computer Science, Punjabi University, Patiala; Stanford graduate degree (EY says MBA; one unofficial page says otherwise) u
Before
EY since 2015: Americas BCM Leader; Americas FSO Strategy & Transactions MP. Founding partner, Wipro Consulting Services, led its FS advisory. Earlier BCG, Lehman u
Honors
M&A Advisor 40 Under 40; 2026 Outstanding 50 Asian Americans in Business
Personal: LEGO, rebuilds postwar motorcycles, "entrepreneur at heart." Unlikely to interview a Senior Manager; if he does, stay on economics and trust.

What he says

"20% of generative AI cases will drive 80% of the value." Efficiency use cases get "commoditized"; real value is "customer engagement and growth." ML and rules "better suited for tasks needing precision," GenAI for "language understanding, content creation, complex data synthesis." AML stays rules-based.American Banker podcast, as BCM Leader
"The real economics of running agentic workflows at scale include infrastructure, operations, people, risk and the hidden cost of engineering around AI's limitations."Yahoo Finance, 2026, via LinkedIn u
"Trust isn't a constraint on innovation, trust is what makes innovation scalable." "Know Your AI."LinkedIn, 2026

What he will probe

  • Total cost of agents versus token cost; realized ROI versus pilots.
  • Which 20% of use cases, and why those.
  • When not to use GenAI; how rules, ML and LLMs combine.
  • Competitive advantage first, technology second.

Your angle

  • AgentForge: platform inference $0, about $0.03 per run billable, invariant enforced by test. Fleet routing by task. That is his "total cost" argument in practice.
  • Your workflow-versus-agent discipline is his rules-versus-GenAI split. Agree with him in your own words.
  • Ask: "Which 20% is FSO betting the FDE team on this year?"

Misattribution trap

  • "An agent in banking is not just summarizing a document" is Deepak Khosla (Impetus). "America can't slow AI" is Vinod Khosla. Quote neither.
Sources ey.com/en_us/people/sid-khosla; americanbazaaronline.com 11 May 2026; americanbanker.com podcast; EY "Total Cost of Agents" insights 2026; linkedin.com/in/sidkhosla (public); sidkhosla.wixsite.com (unofficial).
Regulation · the April 2026 reset

Know the April 2026 reset cold: SR 26-2 replaced SR 11-7, scoped out GenAI and agentic AI, and an RFI is coming

Model risk: Fed SR 26-2 · OCC Bulletin 2026-13 · FDIC (17 Apr 2026)

  • Supersedes SR 11-7 (2011) and SR 21-8 (2021 BSA/AML statement). Same lifecycle: development, implementation, use, independent validation, governance, inventory, effective challenge.
  • Now explicitly risk-based and tailored to size and complexity; mid-size banks no longer need to imitate GSIB practice.
  • GenAI and agentic AI are out of scope as "novel and rapidly evolving"; banks are pointed to broader risk-management and governance practices.
  • The agencies said an RFI on model risk management including GenAI and agentic AI is coming "in the near future."

Your line

"SR 26-2 took GenAI out of scope, which moved the burden to governance. So the engineering has to produce the evidence an examiner would have asked MRM for anyway: an inventory entry, lineage, evals, human-in-the-loop points, and an audit log."

What it changes for an FDE

  • No safe harbor from "it's not a model." The control burden sits with risk management and the engineering team, so build the evidence trail from day one.
  • Tailoring helps mid-size banks: the same platform can carry lighter documentation where the risk profile allows, which is a selling point.
  • The RFI means the standard will move within the engagement; design for change logs and re-validation, not a one-time sign-off.
Sources federalreserve.gov SR2602; Sullivan & Cromwell memo Apr 2026; OCC Bulletin 2026-13. Right → FINRA, the Fed's posture, and capital.
Regulation · broker-dealers, the Fed's posture, capital

FINRA already expects hallucination controls, agent permissions and logged agent actions, and the Fed says it wants to support adoption

Broker-dealers and wealth: FINRA

  • Regulatory Notice 24-09 (27 Jun 2024): existing rules apply to GenAI whether built in-house or embedded in third-party tools.
  • 2026 Annual Regulatory Oversight Report (9 Dec 2025), new GenAI section: procedures to catch hallucinations, where to keep a human in the loop, agent permissions and data access, logging agent actions and decisions, guardrails on agent behavior.
  • Regulatory Notice 26-14 (9 Jul 2026): proposal to modernize Rule 2210 (communications with the public) citing GenAI. Reg BI still governs recommendations.
  • Quote to carry: "Those rules apply whether member firms are directly developing GenAI tools for their proprietary use or when leveraging the technology of a third party."

Fed posture: Vice Chair for Supervision Bowman, 1 May 2026

  • FSOC AI roundtable: "Our approach should support banks in implementing AI tools safely, effectively, and efficiently."
  • Revised guidance "now applies narrowly to traditional models and basic AI applications." Third-party risk guidance being simplified; flexibility for smaller banks.
  • FSB sound-practices consultation expected Q3 2026; cited Anthropic's Mythos model as the double-edged cyber example.

Capital (Girling's topic)

  • 2023 Basel III Endgame rescinded 19 Mar 2026; new US proposal with an expanded risk-based approach and new standardized approach. AI use cases here: RWA data remediation, capital calculation lineage, regulatory reporting automation.

Example answer

"For a wealth client I would map the agent's permissions and logging to the FINRA 2026 report before writing a prompt: what it can read, what it can do, what gets logged, where the human signs. Then Rule 2210 review for anything client-facing."

Sources finra.org AI key-topics page; FINRA RN 24-09, 26-14; FINRA 2026 Annual Regulatory Oversight Report; federalreserve.gov Bowman speech 1 May 2026; ey.com Basel proposal piece 16 Apr 2026.
The firm · EY in 2026

EY's stack is Microsoft plus NVIDIA, its audit relationship with Anthropic rules out a Claude alliance, and AI sales grew 49% last year

The stack to name

  • EY.ai Agentic Platform with NVIDIA (Mar 2025): NIM model catalog, NeMo Guardrails, AI Foundry; 150 agents for 80,000 staff at launch, 50,000+ agents built in nine months.
  • Microsoft ($1B+ over five years, May 2026): Azure AI Foundry and Fabric for orchestration, Copilot to 400,000 staff, FDEs named as the delivery mechanism.
  • Databricks (Mudit, Keipper) and Snowflake (Phil) for data; ServiceNow (Sussex) for workflow.

The moment

  • FY26 (8 Oct 2026): $57B revenue, Consulting $17.4B, AI services sales +49%, NVIDIA GSI Partner of the Year.
  • "Forward Deployed Engineer" entered EY's vocabulary with the UK&I launch on 28 Apr 2026: senior engineers embedded in client delivery teams to get past pilots.
  • The FSO team is hiring at three levels in the same month. That is a practice being built, not a seat being backfilled.

The nuance to hold, not raise

  • Anthropic's S-1 (27 Oct 2026) names EY as auditor. Independence rules bar a commercial Claude alliance; PwC, KPMG and Deloitte all have one.
  • The posting still names Claude Code as expected tooling. Treat Claude as a tool and a model option inside the gateway, never as a partnership pitch.
  • Say FSO for the FS business unit; EY-Parthenon is strategy. Purpose line: "Building a better working world."
Sources ey.com newsroom (Mar 2025, May 2026, Oct 2026); ey.com/en_uk FDE release 28 Apr 2026; Bloomberg Law 27 Oct 2026; big4news 12 Aug 2026. Right → the dated timeline, then alliances by person and vocabulary.
The firm · structure

EY in one picture: four service lines, ten Regions in three Areas, and a financial-services unit that cuts across all of it

EY Global · Ernst & Young Global Limited · Chair and CEO Janet Truncale (Jul 2024, ex-head of Americas FSO)

FY26: $57B revenue (+4.7%), 415,182 people, AI services sales +49%. Global Consulting Vice Chair Errol Gardner.

Assurance

$18.9B · 130,279

Consulting

$17.4B · 131,678, now the largest by headcount

EY-Parthenon

all of Strategy & Transactions since Mar 2025 · $6.8B

Tax

$13.8B

Americas Area = USLI Region (US, Latin America, Israel) + Canada Region

Dante D'Egidio, EY Americas and US Managing Partner since 1 Jul 2026. Americas revenue $25.9B. Whitt Butler, Americas Vice Chair for Consulting (25,000+ consultants). Ten Regions since 1 Jul 2025, down from 18.

Financial Services Organization (FSO), Americas · created 2000 · about $9B, about 14,000 people

Sid Khosla, Vice Chair (May 2026); Ron Giammarco, Deputy Vice Chair and Americas Consulting FS Leader. Banking & Capital Markets · Wealth & Asset Management · Insurance. EY's postings: "the only professional services firm with a separate business unit dedicated to the financial services marketplace."

FS Consulting → Technology Consulting (Scott Keipper) → FS Consulting AI practice (Mudit Gupta) → FDE Applied AI seats

Titles verified on ey.com; the line from the AI practice to the FDE seats is inferred from the postings. ey.com also lists Matthias Loh (Charlotte) as "FSO Technology Consulting Leader"; one page is stale.

Say it in one breath

"EY has four service lines; this seat is in Consulting, which is $17.4B and now the largest by headcount. Financial services is the one industry EY runs as its own business unit, the FSO, under Sid Khosla since May; inside it, Ron Giammarco runs consulting, Scott Keipper runs technology consulting, and Mudit Gupta leads the AI practice that posted the role."

Vocabulary

  • Partner vs Principal: both equity owners; by consensus Partner is the CPA, Principal is not. Managing Director is the senior non-equity rank. EY calls the ranks PPMDD.
  • Sub-service lines in Consulting: Business, Technology, Risk, People.
  • GDS: about 74,000-person offshore delivery network. Talent Shared Services is its HR back office; requisitions post in its name.
  • ey.ai ecosystem as EY lists it: Adobe, Databricks, Dell, IBM, Microsoft, NVIDIA, OpenAI, SAP, ServiceNow, Snowflake.

Watch item

The Senior Manager requisition page on careers.ey.com read "The Job is no longer available" on 31 Oct while LinkedIn still listed it with 26 applicants. Ask the recruiter whether it was filled, reposted or consolidated.

Sources EY FY26 results (8 Oct 2026); EY US Transparency Report 2025; Global Strategy Update (25 Jun 2025); EY-Parthenon release (Mar 2025); ey.com executive pages and bios; careers.ey.com; Going Concern (headcount).
The firm · timeline

Three years from EY.ai to a forward-deployed practice, dated

Sep 2023
EY.ai launched after a $1.4B investment; EYQ internal LLM; early Azure OpenAI access via Microsoft.
18 Mar 2025
EY.ai Agentic Platform with NVIDIA: NIM model catalog, NeMo Guardrails, AI Foundry; 150 agents for 80,000 staff; FS, financial crime and regulatory compliance among first domains.
12 Jun 2025
EY Risk Agents on the platform (third-party-risk document review; Uber first client).
7 Apr 2026
Agentic AI across the global audit platform (EY Canvas) on Microsoft Foundry, Fabric and Azure.
28 Apr 2026
UK&I launches "Forward Deployed Engineer" roles: senior AI engineers embedded in client delivery teams. The US FSO roles are the Americas analogue.
11 May 2026
Sid Khosla named Americas FS Vice Chair.
21 May 2026
EY–Microsoft $1B+ over five years; FDEs integrated with EY practitioners; Copilot to 400,000 staff.
1 Jul 2026
Dante D'Egidio becomes EY US and Americas Managing Partner.
27 Aug 2026
Integrated Solutions portfolio launched; FS solution with SymphonyAI. Sussex's new title maps here.
30 Sep 2026
Aaru alliance (behavioral simulation; FS use cases: message testing, pricing, M&A).
8 Oct 2026
FY26 results: $57B (+4.7%); Consulting $17.4B; AI services sales +49%; 381,000 Copilot licenses; 415,000 people.
18–30 Oct 2026
FSO FDE Applied AI postings at Senior Manager, Manager and Senior levels; Anthropic's S-1 names EY as auditor (27 Oct).

Older context they may assume you know

  • Project Everest (the audit and consulting split) collapsed in Apr 2023 at over $1B cost, followed by about 3,000 US cuts, mostly consulting.
  • CEO Janet Truncale consolidated 18 regions into 10 super-regions from 1 Jul 2025; the US sits in a US, Latin America and Israel super-region.
  • No consulting-wide layoffs reported in 2026; 2026 reports covered audit staff-level trims and offshoring of US executive assistants.
  • EY was 2024 Databricks FS Partner of the Year; 119 alliances; alliance revenue growing at a 34% CAGR (FY26 release).
Sources ey.com newsroom releases cited by date; cfodive Sep 2023; accountingtoday Apr 2023; peoplematters 2025; goingconcern 2026; Bloomberg Law 27 Oct 2026.
The firm · alliances and vocabulary

Each leader has a platform they front; match the vocabulary to the person in the room

Alliances by person

PlatformEY voiceWhat to say
DatabricksMudit Gupta (leads FS Databricks business); Scott Keipper (quoted in the 2024 Partner of the Year release)Lakehouse, Delta, Unity Catalog, MLflow; your MUFG data catalog work
SnowflakePhil Andriyevsky (webinars 2023, Jan 2026; Summit 2025)Your US Bank pipelines; ask how FS clients split the two
ServiceNowPaul Sussex (disputes, fraud ops, vulnerability management)AI embedded in workflow, redesign the handoffs
MicrosoftFirm level ($1B+ May 2026)Azure AI Foundry, Fabric, Copilot Studio by name
NVIDIAFirm level (EY.ai Agentic Platform; GSI Partner of the Year)NIM, NeMo Guardrails; your DGX Spark and Earth-2 work
OpenAINo confirmed EY alliance (Frontier Alliance lists BCG, McKinsey, Accenture, Capgemini)Do not assert one
AnthropicEY is the auditor (S-1, 27 Oct 2026)Claude as tool and model option; never a partnership

Phrases that signal you did the reading

  • "The control layer" (Keipper) and "runtime control" (Mudit) for governance as architecture.
  • "Control before autonomy" (Nikhil Lele, Jun 2026): agents earn autonomy in phases.
  • "Business-led, technology-enabled" (Batra) and "platform-based approach" (Sameer).
  • "Regulatory confidence" (Giammarco) and "Know Your AI" (Khosla).
  • "Total cost of agents" (EY 2026 series): $0.04 chat becomes $1.20 orchestration.

Example: answering "what do you know about EY's AI platform?"

"The Agentic Platform you built with NVIDIA gives you a model catalog and guardrails layer, Microsoft gives you orchestration and the data fabric, and the Integrated Solutions launch in August packages them by outcome. The gap I'd expect to fill as an FDE is the client-specific control layer and the integration into their systems of record."

Sources ey.com alliance pages (Databricks, Snowflake, ServiceNow); EY newsroom; openai.com Frontier Alliance page; Bloomberg Law; ey.com insights (Lele 18 Jun 2026; Total Cost of Agents 2026).
Architecture · the whiteboard

Whiteboard this when they ask how you would architect a governed agentic system inside a bank

Layer 1Systems of record and workflow
Core banking, case management (ServiceNow), advisor desktop, data warehouse. Integrate into the flow and the existing governance; never a side chatbot. (Keipper, Sussex)
Receipt: one codebase shared by risk, credit and modeling at US Bank; Snowflake pipelines for S&P data.
Layer 2Orchestration
Deterministic workflow for precision steps (checks, calculations, rules, ML scores). Bounded agents for judgment steps (triage, summaries, drafting) with explicit tool permissions and budgets. Autonomy is earned by eval evidence. (Lele "control before autonomy"; Khosla rules vs GenAI)
Receipt: planner to deployer pipeline with human gates; LangGraph copilot with a deterministic verifier in front of output.
Layer 3Model gateway
Route by task: frontier, small language model, or local. Fallback chains, token and cost budgets, prompt and version registry. (Sameer gateways; Khosla total cost)
Receipt: routing across Claude, OpenAI, Qwen, Nemotron, DeepSeek; AgentForge platform inference held at $0 by a tested invariant.
Layer 4Knowledge and data
Lakehouse (Databricks or Snowflake), lineage, data products, retrieval with citations and freshness. Data quality is the existential risk. (Girling, Andriyevsky)
Receipt: MUFG enterprise data catalog; risk news monitor tracing every headline to source; RepIntel entity matching at scale.
Layer 5Control layer
Eval suites as acceptance gates (golden sets, regression on every change), guardrails (NeMo, Guardrails AI), human-in-the-loop approval points, tamper-evident audit log, tracing and drift monitoring (LangSmith, Arize, Datadog). (Mudit "runtime control"; Keipper "control layer")
Receipt: 80-case eval with golden-hash regression; AgentForge red team, dual judge, audit log; per-run cost metering.
Layer 6Evidence for MRM and examiners
Inventory entry, documentation, validation artifacts, change log mapped to the SR 26-2 lifecycle and FINRA's 2026 expectations (HITL, agent permissions, logged actions).
Receipt: wrote the documentation that passed independent validation at US Bank; Board Risk Committee report at MUFG.
Use Draw layers 1 to 6 top to bottom, then say which two the client already has and which one you would build first. Name the EY components where they fit: NIM, NeMo Guardrails, Azure AI Foundry, Fabric, Databricks. Right → the first 90 days as a worked example.
Architecture · example: the first 90 days

If asked how you would run the first ninety days at a bank client, pick one decision, get to shadow production fast, and leave capability behind

Weeks 1–2 · Pick the decision

  • Choose one decision the output changes: an alert disposition, a dispute outcome, a reserve estimate, an advisor action. Not "a copilot."
  • Baseline it before any build: cycle time, cost per case, error rate, rework.
  • Map the control owners on day one: model risk, compliance, information security, data owners. Get the inventory entry opened.
  • Confirm the data path and the system of record you will integrate with.

Weeks 3–6 · Thin slice in shadow mode

  • Deterministic workflow for the precision steps; one bounded agent for the judgment step; verifier in front of the output.
  • Eval suite first: golden set drawn from real cases, regression on every change, red-team cases for injection and leakage.
  • Run in shadow against live cases; audit log and tracing on from the first run.
  • Weekly review with the control owners using the evidence, not slides.

Weeks 7–12 · Production with a human in the loop

  • Promote to HITL production; measure realized change against the baseline; report total cost including people and ops.
  • Hand the client team the runbook, the eval suite and the approval process; coach them to extend it.
  • Return a reusable component to the EY platform (a gateway policy, an eval pattern, a connector) so the next client starts further along.
  • Scope phase two with the business owner: the next decision, priced.
Why this lands It answers Phil (pilot to production, workforce), Sameer (attribution, platform reuse), Mudit (runtime control), Keipper (systems of record), Giammarco (how the work gets delivered) in one story.
Your answers · the five stories

Five stories, each aimed at the person who will care most

1 · Displacing the vendor at US Bank Mudit · Keipper · Batra

Situation: a $1M vendor commitment for loss modeling with a black-box methodology. Task: a transparent, defensible alternative the credit team would actually use. Action: built a Python Monte Carlo loss platform on an open-source stack (NVIDIA Earth-2, CLIMADA, HAZUS), Claude Code agents writing and testing code, Snowflake pipelines for S&P data, 36+ building types. Result: vendor displaced; one codebase shared by risk, credit and modeling; outputs feed PD and LGD adjustments and CECL reserves.

2 · Getting through independent validation Girling · Mudit

Wrote the model documentation the platform needed; partnered with validation and governance through effective challenge; designed the evidence trail (assumptions, lineage, tests) before the code was finished. Lesson you state: design for the validator from day one and production stops being a separate project.

3 · Board Risk Committee at MUFG Giammarco · Khosla

Founding member of the Americas risk analytics program on a $250B+ wholesale and project-finance portfolio; wrote and presented the exposure report to the Board Risk Committee; ran the vendor lifecycle (MSCI, Oliver Wyman, S&P); argued for a bottom-up client-level approach and built the data environment that made it possible.

4 · AgentForge's cost and security invariant Sameer · Sussex · Khosla

Multi-agent red team for clinical LLM workflows (orchestrator, red team, dual judge, documentation agent). Routed every agent role to local open models on a DGX Spark: platform inference cost $0, only the customer's target model billable at about $0.03 per run, all-local invariant enforced by a test, tamper-evident audit log. Led a team of three; wrote the PRD and the five-module architecture.

5 · Leaving capability behind Phil · Giammarco

Embedded Claude into the US Bank team's research workflow, then coached them to extend it; built Power Automate agents for recurring data work and taught the team to build their own; open-sourced the build workflow from Gauntlet and a good share of the cohort now runs on it. Your framing: forward deployed means the client runs it after you leave.

Delivery Keep each under 90 seconds. Lead with the result for senior listeners (Khosla, Giammarco, Batra); lead with the mechanism for technical ones (Mudit, Sameer, Sussex). Right → technical questions, judgment questions, and the numbers to carry.
Your answers · technical questions

The technical questions, answered in the order they will probably come

"Walk me through an AI system you took to production in a regulated environment."

US Bank platform: the decision it informs (credit, reserves), the validation it passed, the shared codebase, the adoption. Then the news monitor as the GenAI example: grounded, auditable, in weekly use. End on what you would do differently with SR 26-2 in force.

"How do you decide between a workflow and an agent?"

Precision and auditability steps are deterministic: rules, ML scores, calculations, KYC checks. Judgment and language steps get a bounded agent with explicit tools, budgets and a verifier. Agents earn autonomy through eval evidence, never by default. Example: alert scoring deterministic, alert narrative agentic with human sign-off.

"How do you evaluate an LLM system?"

A golden set tied to business acceptance; regression on every change (golden-hash); a red-team suite for injection, leakage and tool misuse; production monitoring for drift; cost as a tested invariant. Evals are the acceptance test the validator reads, so write them in the validator's language.

"What does governance look like at runtime?"

Tool permissions and scopes, budgets, approval gates, a kill switch, a tamper-evident log, an inventory entry. Map it to the SR 26-2 lifecycle and FINRA's 2026 report: human in the loop, agent permissions, logged actions, guardrails. Mudit's phrase is "runtime control"; use it.

"How do you prove ROI and attribution?"

Baseline before building; name the decision the output changes; measure cycle time, cost per case, error and rework; report total cost including ops and people; call it realized only after shadow and HITL production. Sameer's line is "proving attribution is the hard part"; agree and show the instrumentation.

"How do you use Claude Code day to day?"

Planner, architect, coder, reviewer, tester, deployer pipeline; human approval gates; the reviewer agent independent of the implementers; written project guardrails every change is checked against; Codex cross-verifies; a standing gap log scopes the next iteration. Offer the screen-share.

Watch On the workflow-versus-agent question, this is the exact place you inverted the answer at Riverside. Say "workflow" first.
Your answers · judgment questions

The judgment questions, where the consulting years and the bank years do the work

"You've been inside banks for eight years. Can you sell?"

Eight years at Accenture scoping and delivering for Fortune 20 clients; pitched and ran P&L at Havas; led 10 analysts at GroupM. Inside the bank you sold the build-versus-buy case against a funded vendor and got three teams onto one codebase. Then ask what the Senior Manager sales objective looks like here.

"Why EY, and why now?"

The posting describes the work you have been doing from the client's chair. EY has the client access, a dedicated FS unit, and a platform (Microsoft, NVIDIA, Databricks) to do it at scale, and it is hiring at three levels on the back of 49% AI growth. You want to be the FDE who sits across the table from your former peers.

"Your climate work is niche. How does it transfer?"

It is scenario to loss to PD and LGD to CECL to capital; the hazard is an input and the pipeline is credit risk. The same event, exposure, decision, control pattern runs fraud, disputes and AML. Then name one of those and sketch it.

"Tell me about leading engineers."

Onshore and offshore statisticians and data scientists at Accenture; 10 analysts at GroupM; 5 developers at Havas; a team of three on AgentForge where you wrote the PRD and architecture; a fleet of agents governed by written guardrails and an independent reviewer. Mentoring senior engineers is in the JD; have one specific story ready.

"Where are you not deep?"

Enterprise platform operations at bank scale: Kubernetes, service mesh, API governance across hundreds of services. Say how you cover it: standards and gates you set, the Manager and Senior FDEs who carry it, and a named learning plan. Do not pad it with modesty about "not being a SWE."

"How would you handle a client who wants Claude in production?"

Treat it as model selection inside the gateway: fit, cost, data residency, the client's own approved-vendor list. Do not raise EY's auditor relationship with Anthropic yourself; if they raise it, show you understand independence and move on.

Tone Short answers, then stop. Let them pull the next layer. Senior Managers who fill silence sound like Managers.
Your answers · their numbers

Carry their numbers, not yours: the EY statistics the room has already memorized

StatisticSourceUse it when
About 95% of AI pilots never reach productionAndriyevsky, ey.com, 5 May 2026Explaining why an FDE model exists
77% of banks launched GenAI; 31% started agentic; 40% of use cases miss expectations; 79% rank governance the top improvement priorityEY-Parthenon GenAI in Banking, 19 Sep 2025 (Sameer)Governance as the gating factor
Wealth and asset managers: 95% scaled multiple use cases, 78% exploring agentic, only 27% report substantial impact; 86% surprised by regulatory complexity; 87% use closed-source models on trusted platformsEY GenAI in WAM survey, 16 Sep 2025 (Sameer)Wealth conversations with Batra or Phil
70% of banks use agentic AI (16% deployed, 52% piloting); 56% say it excels at fraud detectionMIT Technology Review Insights with EY, 4 Sep 2025Agentic reality check
Large banks 50% current AI use vs 20% for mid-sizeEY Cash Management Survey, 27 Oct 2026 (32 banks)Mid-market and regional bank clients
A $0.04 chat assistant becomes a $1.20 orchestration once tools, planning and subagents are addedEY "Total Cost of Agents," 2026Cost conversation with Khosla or Sameer
$9–10 processing cost per dispute; chargebacks to 324M by 2028; 30–50% manual reduction from modernizationSussex and Lucas, ey.com, 26 May 2026Disputes with Sussex
EY FY26: $57B revenue, Consulting $17.4B, AI services sales +49%EY FY26 results, 8 Oct 2026"Why now" and why they are hiring at three levels
FSO: about $9B, about 14,000 peopleKhosla appointment coverage, 11 May 2026Scale of the practice you would join
Rule Quote a number with its source and date, once. The room wrote these reports; precision reads as respect.
Your questions · what to ask

Eight questions to ask, grouped by what each answer tells you

Team and structure

  • "Who owns the FDE pod day to day, and how does it relate to the FS AI practice, the AI & Data org, and FS Technology Consulting?" Mudit · Keipper
  • "How do FDEs split time between client sites and building reusable EY platform capability?" Mudit · Sameer
  • "What did the first FDE engagements in FSO teach you that changed how you staff the next ones?" any

Engagement model and stack

  • "Do FDEs default to EY's stack, Azure AI Foundry plus NVIDIA NIM, or the client's, and who decides?" Keipper · Sussex
  • "How does an engagement end: managed service, client-run, or the next build?" Giammarco · Sussex

Success and economics

  • "How is an FDE Senior Manager measured at 6 and 12 months, and what share is sales?" Mudit · Phil
  • "What evidence counts as realized value for a client CFO?" Sameer · Khosla

Career

  • "What is the path from Senior Manager to Managing Director or Partner on this track, and what has to be true?" Mudit · Giammarco

Each person slide carries one more question tuned to that person's own published work.

Rule Ask two per conversation, not eight. Pick the two whose answers you cannot get from the posting. Right → the don'ts and the names.
Your questions · don'ts and names

Nine things not to say, and the names to get right

Do not

  • Say "not a SWE" or "not production." Describe what you built and governed; name gaps as gaps.
  • Cite SR 11-7 as current. Say "SR 26-2, formerly 11-7."
  • Call Girling or Keipper wealth leaders, or Batra the WAM leader. Girling is BCM sector; Keipper is FS technology; Batra is strategy, innovation and client service.
  • Attribute Deepak Khosla's or Vinod Khosla's lines to Sid Khosla.
  • Pitch an Anthropic partnership. EY is Anthropic's auditor. Claude Code as tooling is fine; the posting names it.
  • Lead with climate. Lead with credit loss, validation and board reporting.
  • Say EY-Parthenon when you mean FSO Consulting.
  • Claim Databricks depth you don't have. Speak lakehouse concepts; ask how FS clients split between Databricks and Snowflake.
  • Mention other processes or offers. If asked about timing, say you are in conversations and can move quickly.

Names, said right

NameSay
Mudit GuptaMOO-dit GUP-ta
Phil Andriyevskyan-dree-YEV-skee; "Phil" in conversation
Sameer Guptasa-MEER
Scott Keipperconfirm with recruiter (KY-per or KIP-er)
Paul SussexSUSS-ex
Gurdeep Batragur-DEEP BUT-ra
Adam GirlingGUR-ling
Ron Giammarcojah-MAR-koh
Sid KhoslaKHOS-la (Sid, not Siddharth, in public)

From Riverside

  • You fumbled Contruent, Kaseware and Joe Manning. Write the nine names on the pad before the call starts.
Pronunciations are best effort from name origin; the recruiter will know. Ask once, early, and it reads as care.
Recruiter screen · the first conversation

The recruiter call is a thirty-minute fit check with six beats; know what is being scored at each one

MinuteBeatWhat they are scoringYour move
0–4Their pitch: FSO, the FDE team, three levels open, hybrid, four citiesWhether you already understand the roleLet them talk. Nod to specifics you know (FS Technology Consulting, Mudit's practice) without showing off.
4–11"Walk me through your background"Coherent story, seniority, FS fit, hands-on AI, consulting historyThe recruiter version of the intro: 75 seconds, result first, no stack names beyond Python and Claude Code.
11–15Motivation: why EY, why consulting again, why now, what is happening at US BankFlight risk, realism about consulting lifeOne clean 2026 narrative. Decide it before the call (see "before the first conversation").
15–20Must-have screen: 10+ years, Python, agentic systems, regulated FS, consulting, mentoringTicks against the posting; can they advance you to the hiring managerShort yes-with-proof answers. One artifact per must-have, then stop.
20–26Logistics: level, location, hybrid, travel, authorization, notice, compensation, other processesWill an offer land; are expectations inside the bandAsk for the band's typical landing point before giving a number. Then anchor once.
26–30Your questions and next stepsJudgment, preparation, seriousnessTwo or three questions from the recruiter list. Confirm the loop and the timeline. Ask what to prepare.

Who it probably is

Mudit Gupta's earlier hiring post pointed candidates to FS technology recruiters Pete Eels-Miller and Elle Gray (EY Talent Shared Services). Recruiters at EY screen for fit and comp, then hand to the hiring manager. They do not judge architecture; they judge clarity, seniority and whether an offer would close.

Three outcomes to secure

  • Advance to the hiring-manager round at Senior Manager, not Manager.
  • A comp expectation on record in the upper half of the NYC band, framed as scope and market, not need.
  • The loop and timeline in writing: who, how many rounds, any exercise, decision date.

Before you dial

  • Resume v4 is the one they have. The 2026 sentence is decided. LinkedIn matches.
  • Nine names on the pad. Your number on the pad. Your notice period on the pad.
Right → their questions with draft answers to edit (background and motivation; level, logistics and compensation), then your questions and what to listen for. Drafts marked edit are yours to rewrite; brackets mark facts only you can fill in.
Recruiter screen · their questions, part 1

Background and motivation: six questions they will ask, with draft answers in your voice to review and change

"Walk me through your background." edit

"Twenty years in data and AI, split three ways. Twelve client-facing at Accenture and two agencies, leading onshore and offshore analytics teams for Fortune 20 clients. Eight inside bank risk at MUFG and US Bank, building the platforms, getting them through model validation, presenting to a board risk committee; at US Bank I replaced a $1M vendor with a platform three teams now share. This year, hands-on agentic engineering through Gauntlet, shipping with Claude Code daily. The FDE role is those three combined."

"Why EY, and why this role?" edit

"The posting describes work I've done from the client's side: taking AI into production inside a bank's governance. EY has the client access, a dedicated FS unit, the Microsoft and NVIDIA platform, and is building this team at three levels now. I'd rather do it across many institutions than one."

"What's your current situation, and what's prompting the move?" edit · decide the 2026 sentence

Option A: "I lead scenario analysis at US Bank, and in 2026 I invested deliberately in applied AI engineering: the Gauntlet fellowship and an AI engineering lead role at GT School [confirm framing]. I want the seat where that combination is the job."

Option B (if no longer full-time at US Bank): "I stepped back from US Bank in [month] to retool in applied AI; since then I've shipped agentic systems weekly and led AI engineering at GT School. This is the role I retooled for." Three sentences; LinkedIn must agree.

"Why consulting again, and do you know what a Senior Manager does here?" edit

"Twelve years client-side before banking, so the rhythm is familiar: scoping, delivery, the client relationship, developing people. I know Senior Managers carry a sales objective; I've run a P&L at Havas and sold a build-versus-buy decision inside a bank against a funded vendor. An FDE sees ten banks' problems in the time an employee sees one."

"How hands-on are you today?" edit

"Daily. Python, Claude Code and Codex, LangGraph, local and frontier models. This year I shipped a multi-agent red-teaming platform with an eval harness and a tested cost invariant, and a clinical copilot with an 80-case regression suite. I can screen-share the pipeline in the technical round."

"What do you know about EY's AI work in financial services?" edit

"The Agentic Platform built with NVIDIA, the Microsoft commitment in May that names forward-deployed engineers, your FS leaders' published view that governance is what makes agents deployable, and the three FDE levels posted this month. From the banking survey, the number that stuck is 40% of use cases missing expectations."

Rule Every answer under 60 seconds on a recruiter call. They are filling a form, not learning your architecture. Save depth for the hiring manager.
Recruiter screen · their questions, part 2

Level, logistics and compensation: where the call is actually decided

"We have this at Senior Manager and Manager. Which level are you targeting?" edit

"Senior Manager. Twenty years, VP titles at two global banks with a board-level reporting line, teams of up to ten, and a consulting background. The Senior Manager posting is architecture governance and client advisory, which is the work I've been doing. I'd want to understand how you calibrate between the two levels, but Senior Manager is the fit."

"What are your compensation expectations?" edit · your number

First, turn it: "The posting shows $172,800 to $374,000 for New York. Where does a Senior Manager with my profile typically land in that band, and what's the bonus target?" Then anchor once: "Given the scope, a base around $300,000 with total cash in the mid-$300s is where I'd expect to land. I'm flexible on structure." Do not give salary history; New York City employers may not ask for it, and the recruiter will not expect it.

"Where are you based, and how do you feel about 40 to 60 percent in person and client travel?" edit

"New York City. Hybrid at that level is how I've worked for years, and client-site time is the point of a forward-deployed role. Travel to Charlotte, Dallas or Tampa for engagements is fine [confirm any constraints: family calendar, recovery, school year]."

"Work authorization, notice period, and when could you start?" fill in

"[Authorized to work in the US without sponsorship / confirm]. Notice is [two to four weeks / confirm]. I could start [date or 'within a month of an offer']." Have exact answers; vagueness here reads as a flight risk.

"Are you interviewing elsewhere? What's your timeline?" edit

"I'm in a couple of conversations, and this one is a priority because it matches the work most closely. I can move at your pace; if your loop is two or three weeks, that works." Do not name firms or stages.

"Anything in your background we should know before the hiring manager sees your resume?" edit

"Two things worth saying plainly: my delivery experience is credit and market risk rather than fraud or wealth, and my platform work has been Snowflake rather than Databricks. Both transfer, and I'd rather the hiring manager hear it from me than discover it."

Numbers to have on the pad NYC band $172.8K–$374K · typical EY Consulting SM cash about $280K · your anchor $300K base, $345K+ total, $280K floor · SM bonus target 15–20% · no equity at EY.
Recruiter screen · your questions and what to listen for

Ask the recruiter about the loop, the level and the seat; listen for whether the practice is real

Ask about the process

  • "Who is the hiring manager, and who else is on the loop?" (You expect Mudit Gupta; listen for an MD under him.)
  • "How many rounds, and is there a technical exercise, case or presentation? What should I prepare?"
  • "What's the timeline to decision, and when do you need someone to start?"

Ask about the seat

  • "Is this Senior Manager seat tied to a specific client engagement or building the bench? How many SM seats are open?"
  • "Where does the team sit: FS Technology Consulting under the FS Consulting AI practice?"
  • "What distinguishes a Senior Manager from a Manager hire on this team?"
  • "How much of the role is Databricks-specific?"

Ask about the offer

  • "What does the comp structure look like: base, bonus target, signing, and when is the next promotion cycle?"
  • "Is the upper half of the posted band realistic for a profile like mine?"
  • "What's the New York office expectation versus client sites, and how much travel to the other three cities?"

Signals that the practice is real

  • A named hiring manager who is a Partner or MD in the AI practice; a defined loop with a technical conversation; a start date tied to client demand.
  • "Three levels because we're standing up pods" rather than "we'll see what level you come in at."
  • Comp discussion that references the band's upper half without flinching.

Yellow flags

  • "We may bring you in at Manager and promote." Ask what the promotion cycle and criteria are before accepting that framing.
  • "The role is mostly at one client" with no platform time: staff augmentation with a better title.
  • No technical round at all for an architect-level seat.
Close "Thanks. I'll send a short note tonight confirming my interest at Senior Manager and my availability. Who should I expect to hear from next, and when?" Then send that note the same day.
The call · comp and process

Anchor in the upper half of the NYC band and expect a practical, project-based loop that ends with partners

The band and the market

ReferenceFigure
Posted NYC base range (SM)$172,800–$374,000
Posted national base range (SM)$144,000–$329,100
NYC band midpointabout $273,000
Typical EY Consulting SM cash (StrategyCase, RoadToOffer, Jul 2026)~$240K base + ~$40K bonus
Levels.fyi SM total (Jun 2026) · Glassdoor 90th pct (Aug 2026)~$261K · ~$396K
Equitynone at EY

The same SM band appears on another FSO posting (CTO, Semantic AI, Oct 24), so it is the standard band. An "FDE premium" above typical SM cash is plausible but unverified.

Your position

  • Anchor $300K base, bonus at the SM target (15–20%), total cash at or above $345K. Floor $280K base.
  • Rationale you state: architect-level scope with 10+ years, FS regulatory depth, New York, and a posted ceiling of $374K.
  • If they hold the line on base: ask for a signing bonus, a defined MD-track review at 18–24 months, and the senior end of the SM title.

Process, as EY and candidates describe it

  • Recruiter screen. Mudit's earlier hiring post named FS tech recruiters Pete Eels-Miller and Elle Gray. Confirm hiring manager, loop, and whether there is a technical exercise.
  • One or two rounds with Senior Managers or MDs. Cases are practical and drawn from your own projects, not MBB style. Expect an architecture conversation and possibly a coding or writing sample (EY's "How we hire" page allows for both).
  • Partner round: Mudit, possibly Keipper or Phil. Behavioral themes: leadership, problem solving, client management. EY calls its approach strengths-based.
  • Senior Managers carry sales objectives. Have the "can you sell" answer ready.
  • Decision typically 5–10 business days after the final round.

Who to call for an inside read

  • Brian Rauch (Kick Drum owner-partner, your June conversation) spent about ten years at EY earlier in his career. Fifteen minutes on culture, promotion rhythm and how FSO Consulting treats hands-on builders is worth asking for.
Sources careers.ey.com SM posting; strategycase.com, roadtooffer.com (Jul 2026); Glassdoor, Levels.fyi snippets; ey.com How we hire; PrepLounge; resumeadapter. Right → before the first call, then sources.
The call · the loop and how to practice

EY Consulting interviews laterals on their résumé and behavior, with partners last; rehearse four formats anyway

#RoundWhat it scoresFormat
1Recruiter screen, 30 minMotivation, level, comp, location, travel, authorizationPhone or video
2Hiring leader or panel, 45–60 minRésumé deep-dive; EY's three scoring questions: relevant experience, the action you took, what it led toOften a Manager, SM and Partner panel
3Technical or architecture conversation, 60–90 minA client scenario from your résumé, or a reference architecture for RAG and agents in a bank with evals, guardrails and model-risk governanceConversation; written case or coding uncommon at SM
4Partner or Principal, 45–60 min (sometimes a second partner for SM calibration)Fit, judgment, commercial instinct, "why EY," owning engagements from day one, developing people, a view on market trendsOne to one
5DecisionAbout 5–10 business days; references and background check after the offer; independence onboardingVerbal, then written

What candidates report

  • "EY doesn't typically do case interviews for experienced hires at the senior manager level; instead, you typically get a panel... asking questions about your experience and resume." (Fishbowl)
  • Where a case appears it is conversational and implementation-focused; MBB-style cases live at EY-Parthenon. One lateral reported a 90-minute case plus a partner presentation, so ask which shape yours is.

Rehearse four formats

  • 10–15 CAR stories, each checked against the three scoring questions.
  • Two résumé projects told as a client case and as a technical design.
  • A 45-minute decomposition of an under-specified bank problem: clarify, define success, map inputs, split into workstreams, go deep on one, surface trade-offs.
  • A 10-minute demo narrative with its eval story.

Prompts to run cold

  • "Walk me through the US Bank platform as if I were the CRO."
  • "Scope the first 90 days of an agent that triages SR 26-2 validation findings across three legacy inventories."
  • "A client architect says LLMs are non-deterministic, so no production. Respond."
  • "Why EY, why FSO, why now, and why not a bank or an AI lab?"

Independence flag that is specific to you

EY audits U.S. Bancorp (since 2003) and Anthropic. Your US Bank employment and any U.S. Bancorp or Anthropic holdings are disclosure items; EY's Global Monitoring System requires Managers and above to report and sell restricted securities. Have the household holdings list ready.

Sources ey.com interview tips and How we hire; Fishbowl and Glassdoor excerpts; Blind posts 2022–2026; strategycase.com; Harvard FAS FDE guide (Aug 2026); aced.io; ey.com independence practices; USB 8-K via stocktitan; Bloomberg Law 27 Oct 2026. Round order and durations are synthesis.
The call · qualities EY names

Thirteen qualities EY names, and your one proof for each

Quality (EY's words)Your proofSay it as
Innovation mindsetReplaced a $1M vendor with an open-source platform that passed validation"A black box could not pass validation; the saving followed."
Analytics mindsetMonte Carlo losses into PD, LGD, CECL; 80-case golden-hash eval suite"Evals are the acceptance test, written in the validator's language."
Global mindsetEight years onshore and offshore delivery; MUFG Americas program with global teamsOne concrete cross-time-zone episode, not a summary.
Strong communicatorWrote and presented the Board Risk Committee exposure report"I turn a model into a decision a board can act on."
Team player (teaming)One codebase shared by risk, credit and modeling; matrix team across four functions"It mattered once credit and modeling trusted it enough to share it."
Courage to leadBottom-up approach against the top-down default; build over a funded vendor"I took the position the default would not survive validation, and built the alternative."
IntegrityValidation-grade documentation; naming your own gaps first"I would rather you hear my gaps from me."
Inclusiveness, developing peopleCoached the team to extend the Claude workflow; open-sourced the cohort's workflow"I measure myself on what the team can do without me."
Lifelong learner2026 retooling: Gauntlet, GARP SCR, a self-run model fleet"When agentic engineering became the job, I went and learned it."
Commercial acumen (SM)Accenture scoping and delivery; Havas P&L; vendor RFPs; build-vs-buy won"I scope the next phase at the end of every story."
Ownership in ambiguity (FDE)Founding member who set MUFG's program scope and methodology from scratch"There was no playbook; we wrote the one the team still uses."
Judgment in regulated environments (FDE)SR 26-2, FINRA's agent checklist, inventory entry in week one"Governance is where production starts."
Energy and enthusiasmBuilds constantly outside work; an EY alumnus called it "exactly what we're looking for"Only if asked: "I build things; my kids are my first users."
Sources ey.com What we look for; EY values (en_gl About us); Transformative Leadership story page; EY FS Cyber M/SM and FDE Senior Manager postings; Kick Drum notes (Jun 2026); your resume. Right → how to show a quality without naming it.
The call · showing the qualities

Show a quality by telling the story; let the interviewer name it

How to show a quality without naming it

  • Never say "I have an innovation mindset." Tell the vendor story in three sentences and let them label it.
  • One quality per story, two at most. Stories that try to prove five things prove none.
  • Match EY's scoring: relevance (why this example), action in the first person, a result with a number or a decision. Close the loop every time.
  • Partners read "Better Us" signals: credit the team by role, say what you learned, name what you would do differently.
  • Candor is a quality here: "I have not done that," then the adjacent thing you have done.

Where each cluster lands

  • Innovation, analytics, learning: the US Bank build, the eval suite, the 2026 retooling. Technical rounds.
  • Communication, teaming, developing people: the Board Risk Committee report, three teams on one codebase, coaching the team to extend the workflow. Panel and partner rounds.
  • Courage, integrity, ownership: bottom-up over top-down at MUFG, build over a funded vendor, naming your gaps first. Partner round.
  • Commercial acumen, regulated judgment: Accenture and Havas years, vendor RFPs, SR 26-2 and FINRA fluency, the 90-day plan. Hiring manager and partner rounds.

Stories you still need

  • Inclusiveness: a quieter or more junior voice you brought into a decision, and what changed.
  • Being wrong: a time you challenged a senior and the senior was right; what you learned.
  • A recovery: a project that went sideways and how you brought it back. The vendor story is a win; have a recovery too.
  • Global mindset: one concrete cross-time-zone episode from the offshore years.

Ask the recruiter

"Does the panel use EY's standard behavioral questions, and should I expect the 'what business objectives were you expected to meet this year' line?"

Sources ey.com interview tips (three scoring questions); Transformative Leadership story page; PrepLounge and hackingthecaseinterview partner-round guidance; your resume and notes.
The call · before the first conversation

Settle three things before the recruiter call, and keep the unverified list in your pocket

1 · The 2026 narrative

Your EY resume (v4, regenerated today) shows US Bank through present plus Gauntlet 2026. Your current work also includes leading AI engineering at GT School. Decide the one sentence that covers US Bank, Gauntlet and GT School without a gap or a contradiction, and make LinkedIn match it before anyone at EY looks.

2 · Which resume

Two EY versions exist in Downloads: the v4 "EY FDE" PDF (executive framing, FS risk heavy, Claude Code and DGX Spark named) and the June "AI engineering" markdown with a utilities slant. Send v4. Check that the 90-second intro in the positioning row matches its bullets word for word on the numbers ($1M vendor, 36+ building types, $250B+, 80 cases).

3 · Confirm with the recruiter

  • Hiring manager: Mudit directly, or a Managing Director under him.
  • Who is on the loop; whether there is a technical exercise or presentation.
  • How much Databricks weight the role carries.
  • Office expectation for New York; travel to Charlotte, Dallas, Tampa.

Unverified: know it, don't assert it

  • "Phil A." is Phil Andriyevsky: about 90% on an indexed LinkedIn headline matching his exact title. Confirm on the call by his title, not his name.
  • Reporting lines (Khosla, Giammarco, Keipper, Mudit) are inferred from titles and the "FS Technology Consulting Office" post.
  • Khosla's Stanford degree: EY says MBA; an unofficial page differs. Never say "Stanford MBA" as fact.
  • Mudit's EY start year, partner year and HBS program; whether his 2007–2010 continuous-auditing work was at MetLife or for MetLife.
  • Batra's Infosys and Publicis Sapient dates; Sameer's 2014 start; Sussex's prior employer; Giammarco as career EY: all directory or snippet level.
  • Sussex's "Claude Code 101" and Integrated Solutions title: LinkedIn snippets.
  • Keipper hosting the AI Summit panel and Mudit at the Evident Symposium: snippets.
  • An FDE pay premium above the standard SM band: no public evidence.
Why this slide exists Interview prep fails on confident wrong facts more often than on missing ones. Everything above is good enough to prepare on and not good enough to say out loud.
The call · sources

Sources

Postings
careers.ey.com req 1393575733 (SM), 1393540633 (Senior), 1445825633 (FS FDE Tech Consulting), 1431253633 (CTO Semantic AI); Manager via freehire.me mirror; linkedin.com/jobs/view/4466611730; ey.com/en_uk newsroom 28 Apr 2026 (FDE launch).

EY bios (ey.com/people)
mudit-gupta; phillip-andriyevsky; sameer-gupta; scott-keipper; paul-j-sussex; gurdeep-batra (en_us, en_ca, en_lt); adam-girling (en_us, en_gl, en_au); ron-giammarco (en_us, en_id, en_lv); sid-khosla; traci-gusher; phil-hennessey.

EY insights and pages
FS GenAI offerings page; AI in financial services (5 May 2026); digital workers in asset management (8 Jun 2026); governed AI personalization in wealth (24 Sep 2026); five hallmarks of AI strategy in banking (6 Apr 2026); EY-Parthenon GenAI in Banking (19 Sep 2025); GenAI in WAM survey (16 Sep 2025); agentic AI in banking (18 Jun 2026); data management creates value (9 Apr 2026); US Basel III proposal (16 Apr 2026); dispute management (26 May 2026); WaaS (16 Jan 2025); Future of Asset Management Study (Jan 2026); Nexus for WAM; Total Cost of Agents; Cognitive Investigator; ServiceNow FS offerings; risk and regulatory transformation page.
EY newsroom
EY.ai Agentic Platform with NVIDIA (18 Mar 2025); Risk Agents (12 Jun 2025); agentic AI in audit (7 Apr 2026); EY–Microsoft (21 May 2026); Integrated Solutions (27 Aug 2026); Aaru alliance (30 Sep 2026); CrowdStrike trust layer (Sep 2026); FY26 results (8 Oct 2026); Cash Management Survey (27 Oct 2026); D'Egidio appointment (Dec 2025); Databricks FS Partner of the Year (Aug 2024).

Press and third party
PYMNTS 22 Jun 2026; Banking Dive 2026; aimmediahouse.com 13 May 2026 (Keipper, citing Business Insider); americanbazaaronline.com and newindiaabroad.com 11 May 2026 (Khosla); americanbanker.com podcast (Khosla); The Paypers 19 Nov 2020 and Finextra Sibos 2020 (Giammarco); MIT Technology Review Insights 4 Sep 2025; HBR sponsored Jan 2026; CNBC 5 May 2026; Global Finance 6 May 2026; CIO Dive 26 Oct 2026; International Banker 7 Oct 2024; consultingmag.com 16 May 2024; Google Patents US12639705B2; snowflake.com webinars (2023, 14 Jan 2026); InvestOps speaker page; evidentinsights.com; big4news 12 Aug 2026; Bloomberg Law 27 Oct 2026; openai.com Frontier Alliance page; cfodive Sep 2023.
Regulation
federalreserve.gov SR 26-2 (17 Apr 2026); Sullivan & Cromwell memo (Apr 2026); federalreserve.gov Bowman speech (1 May 2026); finra.org AI key topics; FINRA RN 24-09 (27 Jun 2024), 2026 Annual Regulatory Oversight Report GenAI section (9 Dec 2025), RN 26-14 (9 Jul 2026); OCC Bulletin 2026-13.

Comp and process
strategycase.com and roadtooffer.com EY consulting salary (Jul 2026); Glassdoor and Levels.fyi snippets (2026); ey.com/en_us/careers/how-we-hire; PrepLounge (May 2025); resumeadapter (Jun 2026); Credly EY Transformative Leadership badge.

Your side
Johnny_Chung_Resume_AI_v4_EY_FDE.pdf and generate_resume_v4_ey_fde.py (Downloads); resume-ey-ai-engineering-pu.md (Jun 2026); Riverside and Kick Drum interview notes (Jun 2026); Gauntlet project notes; cortex, projects and SR 26-2 memory files.

Blocked
LinkedIn profile pages (HTTP 999), ZoomInfo, Consulting Magazine profile, rmany.org, hrtoday.in. Facts from those appear only as snippets and are tagged unverified.
Research run 30 Oct 2026 across five parallel threads; every quotation checked against a fetched page unless tagged otherwise.