For audit, review & validation firms

Your constraint isn't demand. It's reviewer hours.

Every IRR review starts the same way: chase the document request, read a set of documents you've read a hundred variations of, tie the same forty figures into the same workpaper template, and work through the same program. VetALM puts AI on that pass, with every figure tied to its page and every finding checked against its sources. Your senior people start where the judgment begins.

Request a demo See the economics

Each client is its own institution, with its own history and program
Your reviewers approve every finding
Workpapers export to Word, PDF, and Excel

The Workpapers view of a VetALM engagement: tests performed, each with a one-line result and a Pass or Exception badge.

The economics

The same team, more engagements.

Review work is priced by scope and delivered by hours. Most of those hours go to evidence handling rather than analysis: reading, transcribing, cross-referencing, formatting. That's the portion VetALM's AI absorbs.

40–80h → 10–25hTarget effort per engagement

44Document types on the request list, twelve of them core

73Review program questions, answered with citations

100%Findings traceable to source evidence

Effort figures are targets for a typical community-bank IRR review, not a guarantee. Scope, corpus quality, and your own review depth all move the number.

Where the hours go back

Four places a review loses time.

Chasing the document request

The PBC list lives in someone's spreadsheet and the follow-up lives in email. In VetALM the request is part of the engagement: forty-four document types across seven categories, twelve of them core. Request a type with a due date, and the client answers through its own portal, uploading a folder at once while VetALM matches each file to its request line. Your reviewer accepts each file or sends it back with a reason the client sees word for word. Minutes printed inside an ALCO packet count as collected, with a link to the page. For credit unions, two years of NCUA Call Reports arrive without asking. What the list gives you is an accurate picture of what you have to work with, early enough to scope around it.

Tying out the same forty figures

Betas, decay rates, average lives, EVE and NII by scenario, policy limits, approval dates. The AI reads each one, and code then checks it against the page it cites: the quote and the figure must both be there, or the reading is set aside. The limits schedule is built from the board-approved policy, never from a limit printed on a vendor report. The tie-out becomes a review rather than a re-typing.

Running the same tests

Results against the board limits, limit bands checked for gaps, the limits on the vendor's reports checked against the policy, board figures tied to the model, balances tied to the Call Report, NII back-tested against actuals, assumptions checked against typical ranges, review cadence, prior findings. VetALM runs them and lays each one out as a workpaper: the criterion, the figures compared, where each came from, and Pass or Exception.

Working through the program

Seventy-three questions, from governance to back-testing, each answered from the documents with verbatim citations. An answer that can't be cited becomes "Unable to determine" rather than a guess. A No or Partial raises a finding. Your firm keeps its own question bank and tailors it per client.

The difference

Findings you'd put your name to.

An AI that writes findings fast is easy to build. One whose findings survive a senior reviewer, a peer review, and an examiner is the hard part. That is where VetALM spends its effort.

Written from settled facts, then checked against the pages.

Findings are written the way your team writes them: the condition, the criterion it was measured against, why it matters, and a recommendation, classed as a Finding, Technical Exception, or Best Practice. The writer works only from facts the workpapers already settled, and a gate in code refuses any figure, limit, or document outside them. Then a reasoning model reads each finding against the pages it cites. A finding that holds stands, one with a wrong figure is corrected from the source, and one whose conclusion doesn't hold is withdrawn with its reasons.

Seventeen deterministic rules, plus an AI read kept only where a majority of reads agree
Limits measured against the board-approved policy in force at each date, never the vendor's benchmark
Every correction shows the source quote it came from
A final check on the report itself before it can be generated

A finding withdrawn by the source check, followed by each claim that failed and the page quote that contradicts it. — A finding the source check withdrew before a reviewer had to: the two reports it called inconsistent were different rate scenarios.

On a full validation

VetALM covers the program. Your analysts cover the model.

An IRR review or ALM audit is mostly documents and governance, and that is what VetALM does. A full model validation adds a quantitative half: testing the model against the institution's own data. VetALM doesn't do that today, and we won't tell you it does. What it does is take the rest off your analysts' desks.

VetALM does

The document request, the reading, and every figure tied to its page
Governance, review cadence, and prior-finding follow-up
The limits schedule and every result measured against it
Board reporting tied to the model, balances tied to the Call Report, and an NII back-test
Assumption support and reasonableness, the 73-question program, and drafted findings

Your analysts do

Reconcile and sample the loan and deposit data behind the model's inputs
Measure betas, decay, and prepayment speeds from the institution's history
Benchmark the model against an independent, calibrated alternative
Stress the key assumptions to show which ones move EVE and NII
Assess the vendor's method and what its certification doesn't cover

Fitting your practice

It slots into how your firm already works.

Your firm, its clients, and every engagement in one place.

Each client is an institution with its own profile, engagement history, and tailored program, and each engagement holds its documents, workpapers, answers, findings, and audit trail. Every record carries your firm and every query is scoped in the service layer. Clients join through their own portal, where they see only their requests and the findings you release.

Roles map to how review actually runs: firm administrator, engagement lead, senior reviewer, reviewer, read-only
Who may approve a finding or sign off is a server-side capability check
Sign-off seals the engagement's audit trail

The audit trail: timestamped actions with acting entity and before and after content hashes. — Attributable, time-stamped, append-only: the record if a conclusion is ever questioned.

Deliverables in the formats you already send.

An executive summary, the workpaper, and an issues register, in Word, PDF, and Excel. The workpaper carries the findings, tests performed, limits schedule, and analysis exceptions, with the full program and document register as appendices. The AI drafts the prose, but generation runs only once every finding has a reviewer's decision, so nothing AI-drafted reaches a client unread.

Release approved findings to the client, who responds with support; you accept the response or send it back
Rejections require a reason, which becomes part of the record
Regenerate at any time, and the trail records that too

The Outputs view listing a generated issues register, workpaper, and executive summary in Word, PDF, and Excel.

Questions firms ask

Independence, defensibility, and what you sign.

Can we call a VetALM engagement a model validation?

Not on VetALM alone. It doesn't re-perform the model, test the input data at the account level, or measure assumptions from the institution's history, and a full validation needs those. For an IRR program review or ALM audit, VetALM covers the core of the work. For a validation, pair it with your own quantitative testing; VetALM does the documents and governance so your analysts' hours go there.

Does using AI compromise our independence?

Independence is about who forms the conclusion, and that stays entirely with your reviewers. VetALM is a tool your firm operates, much like the workbooks you'd build to test a client, except you don't rebuild them each engagement. The AI reads, tests, answers, and drafts; your reviewers override any answer and accept or dismiss every finding. No deliverable is produced while a finding is still waiting on a reviewer.

What do we tell an examiner or a peer reviewer about the tooling?

That every finding cites the page it came from, that every reading was checked against its page before anything used it, that the seventeen detection rules are deterministic comparison logic with AI supplying only the cited figures they compare, that every finding was checked against its sources by a second read, that the whole engine is scored against a golden dataset before each release, and that the engagement's audit trail records who approved what and when. Those are the questions that get asked, and the platform is designed to answer them with records rather than assurances.

Our clients' documents are confidential. How is that handled?

Documents are encrypted in transit and at rest, scoped to your firm, and served only through permission-checked routes. Document content is processed by Anthropic's Claude through its commercial API, under terms that exclude it from model training. Client users see only their own requests and the findings you release.

Can we keep our own workpaper template and methodology?

The generated workpaper follows a structure built around the evidence and findings, and the findings themselves are yours to edit before generation. If your template differs materially, that's worth a conversation during a pilot. The output layer is the most adaptable part of the system.

What does a pilot look like?

A demo on a sample dataset first, about thirty minutes. If the fit looks right, a pilot on one live engagement of your choosing, run alongside your normal process so you can compare the output to what your team would have produced.

Run one engagement through it.

We'll start with a walkthrough on a sample dataset, then talk about a pilot on a live engagement of yours, run in parallel with your normal process so the comparison is real.

30-minute guided walkthrough, no slideware
Tailored to the documents your clients actually send
Works from the documents your clients already send