Wirewalk Instituteof AI
fin-210 Intermediate AvailableFlagship

Building and Deploying AI in Banking and Lending

For US bank, credit union and lending-fintech teams who build, validate or approve AI: make the calls that decide what ships, what it costs and what it is really worth.

Level
Intermediate
Study time
16 hours
Lessons
18 in 6 modules
Exam
20 questions

About this course

Keelridge Bank, a fictional composite $46 billion regional bank, has promised investors $42 million in AI savings and $150 million in new personal-loan originations. Release decisions are due in March and OCC examinations of model risk management and BSA/AML start in May. You work through the program as the people who own it: the credit modeler whose challenger model looks good until a proxy explains its lift, the model risk head with 31 overdue validations, the BSA Officer with 7,900 aged alerts, the fraud strategist with a 900-hold-a-day desk and a Friday scam wave, the credit administrator whose drafter put a wrong ratio in front of committee, and the digital banking head whose assistant has to recognize an error notice in a customer's own words.

Each of the six modules solves one business problem with the approaches professionals actually choose between (refresh, build or buy; in-house testing or a platform; rules, ML scores or agents; vendor document AI or GSE tools; keyword routers or gated generative assistants), and every lesson works a full example on realistic synthetic data with the real code, prompt or policy text, its output and an expert critique. You see each approach fail the way it fails in banks (train/serve skew, a reason table keyed by position, labels that lie, an invoice at 2.6 times forecast, a doom loop) and practice the fix in a faded exercise with a model answer. Python notebooks can be run or worked from the printed results, so non-coders are never blocked.

The case does not flatter AI. By the end, the evidence supports less than a third of the promised $42 million as savings the bank can book, net of what the tools cost to run, and about $120 million of the $150 million in originations, and a draft board slide showing $23.7 million has to be taken apart line by line. Learning to say that plainly, with an evidence-gated path to close part of the gap, is a core outcome. Every module ends in a Decision Lab: a branching scenario with business metrics, a cliffhanger event and several endings with an expert debrief. The course closes with a strategic capstone simulation and a release decision package for one use case, written for a model risk committee with a one-page board summary; Ellery gives rubric feedback on it as practice. Real-world examples are cited sidebars attributed to their sources; regulatory and product status is stated as of October 2026, with no endorsements and no vendor performance claims. The course is educational, not legal advice, and the certificate attests to completion and assessment only; it is not a regulatory or professional credential.

What you will be able to do

  • Evaluate a challenger credit model against the champion at the business's operating point on out-of-time data, and derive per-applicant adverse-action reasons that meet the Regulation B specificity requirement and stay stable under small input changes
  • Analyze a credit model for outcome disparities and proxy features, compare less discriminatory alternatives on approvals and disparity, and recommend one in a committee memo that names the federal, state and private-law exposures that remain after the July 21, 2026 Reg B amendments
  • Classify each component of a bank AI system as an SR 26-2 model, a deterministic process or a generative or agentic tool outside SR 26-2, tier it by use and materiality, and specify proportionate controls, including an internal standard for out-of-scope tools
  • Validate a vendor ML score on the bank's own data, with effective-challenge questions, outcomes by period and segment, and drift detection, and write approval conditions with monitoring triggers, owners and actions
  • Set and defend an operating threshold for an AML, fraud or onboarding decision on out-of-time data within a staff-capacity and customer-friction budget, so that recall of the cases that matter holds and the SAR and Regulation E clocks are met
  • Test a generative AI tool in lending operations or customer conversations against a gold set with pass thresholds set before testing, covering numbers tied to source, rights-triggering intents and contact and consent rules enforced outside the model, and decide pass, conditional pass or fail
  • Create a release decision package for one bank AI use case that combines scope classification, evidence, consumer-law clocks, third-party terms, fairness and customer-harm review, monitoring and a tested rollback, with a recommendation a model risk committee can act on and a one-page board summary
  • Evaluate the business case for a bank AI deployment by separating booked savings from capacity freed and customer losses avoided, netting run and review costs, and choosing build, buy or extend on total cost, traceability, data boundary and change control

Syllabus

6 modules · 18 lessons · about 16 hours

Module 1: Growing Digital Personal Loans Without Breaking Reg B

Decide whether a gradient-boosted model earns its place in personal-loan underwriting, turn its attributions into specific adverse-action reasons, and find the proxy behind most of its lift.
3 lessons
  1. Champion or Challenger: Testing Whether a Gradient-Boosted Model Earns Its Place
  2. From Attributions to Adverse-Action Reasons That Survive Regulation B
  3. Fairness Testing and the Less Discriminatory Alternative After the 2026 Regulation B Amendments
  • Decision Lab · 12 minThe Approval-Rate PromiseTake a gradient-boosted credit model toward a CEO's public growth promise while reasons, a news story and a geography proxy test every shortcut.

Module 2: Thirty-One Overdue Validations: Model Risk Under SR 26-2

Triage a bank's AI inventory under SR 26-2, challenge a vendor fraud score on the bank's own data, and govern the generative and agentic tools the guidance leaves out.
3 lessons
  1. Reading SR 26-2 and Triaging a Bank's AI Inventory
  2. Effective Challenge of a Vendor Fraud Score on the Bank's Own Data
  3. Governing What SR 26-2 Leaves Out: A Credit-Memo Agent and a Policy Bot
  • Decision Lab · 12 minThe Validation QueueRun model risk management through a backlog, a vendor's two-week ask, a silent model update and an examiner's request letter.

Module 3: The AML Alert Backlog

Clear an aged alert backlog with ML triage and agents without losing SAR-producing alerts, missing the SAR clock or leaking SAR information.
3 lessons
  1. Auto-Closing Alerts Without Losing SAR-Producing Recall
  2. Agentic Alert Review: Approval Gates, Sampling and a File an Examiner Can Follow
  3. Reviewing an AI-Drafted SAR Narrative: Evidence, Clocks and Confidentiality
  • Decision Lab · 14 minSeventy-Nine Hundred AlertsClear an aged AML alert backlog with contractors, models or agents before the exam, and live with the threshold you chose when a subpoena arrives.

Module 4: Stopping Scam Money in Both Directions

Choose warn and hold points for a real-time scam score within a friction and staffing budget, keep Reg E claims handling intact, and keep deepfake-opened mule accounts from receiving the money.
3 lessons
  1. Warn, Hold or Release: Operating Points for a Real-Time Scam Score
  2. When the Customer Says "I Was Tricked": Regulation E Around a Fraud Model
  3. Mule Accounts at the Front Door: Deepfakes, CIP and a Review Queue You Can Staff
  • Decision Lab · 12 minThe Friday Scam WavePick a scam score and operating point within a friction budget and a hold desk's capacity, then handle a spoofed-bank scam wave and its Reg E claims.

Module 5: Faster Loan Files and Savings the CFO Can Book

Speed up commercial memos and self-employed mortgage income with AI while every number ties to source, then rebuild the program's savings bridge so booked savings, capacity and customer outcomes are reported honestly.
3 lessons
  1. Spreading and Credit-Memo Drafting: Tying Every Number Back to the Spread
  2. Document AI for Self-Employed Mortgage Income, With the Underwriter Still Deciding
  3. The Business Case: Turning Hours Freed Into Savings the CFO Can Book
  • Decision Lab · 12 minThe Committee Memo That Was WrongPick and control a credit-memo drafter, respond when committee approves a loan on a wrong ratio, and decide what savings the CFO can book.

Module 6: AI That Talks to Customers, and the Release Decision

Ship a servicing assistant that recognizes error notices and a collections voice agent that obeys contact and consent rules, then assemble the release decision package for the model risk committee.
3 lessons
  1. A Servicing Assistant That Recognizes an Error Notice
  2. Collections Voice Agents: Contact Limits, Artificial-Voice Consent and Disclosure
  3. Capstone: The Keystone Release Decision Package
  • Decision Lab · 12 minLaunch WeekendDesign and gate a servicing assistant against a competitor's launch date, then live through a duplicate-debit Saturday that tests whether it hears disputes.
  • Capstone simulation · 35 minKeystone: The Board, the Exam and the Long GameRun Keelridge's AI program for eight months: choose the bets, handle a vendor model change and a CFO's number, face the board and the OCC, and decide what to scale.

Exam and certificate

The exam unlocks once you have completed every lesson and played each Decision Lab to an ending. It draws 20 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 40 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.

Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.

The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by Zest AI, FairPlay AI, Wolters Kluwer, ValidMind, Microsoft, Anthropic, OpenAI, Nasdaq Verafin, NICE Actimize, BioCatch, Featurespace, Visa, Socure, Early Warning Services (Zelle), nCino, Moody's, Amazon Web Services, Ocrolus, ICE Mortgage Technology, Fannie Mae, Freddie Mac, or Salient. Product and company names are trademarks of their respective owners.

More in AI in Financial Services

All 15 courses in this track
fin-001 Free intro Available

AI in Financial Services: The One-Hour Briefing

A free one-hour briefing built on eight real financial-services jobs, not an overview: spreading financials for a credit memo, writing adverse-action reasons from an ML score, drafting a SAR narrative, setting fraud-score thresholds, checking an earnings-call summary, reviewing an advisor's AI meeting note, summarizing a commercial loss run and reviewing a photo-based auto estimate. Each lesson shows where AI helps, where it must not decide, and the rule that governs that task.

1 h · 9 lessons Free intro
fin-110 Basic Available

AI for Banking Professionals

Practice banking work on a self-contained fictional bank packet: repair service drafts, reconcile credit facts and reason records, summarize investigation leads, compare review burdens and prepare a scoped recommendation. Supplied outputs let you complete the work without a model account, code installation or real customer records.

12 h · 18 lessons See price at checkout