Financial Services & Insurance

Financial data that ties out, and AI decisions you can defend.

We connect fragmented financial data, put AI to work on the decisions that matter, and build in the validation, controls and accountability a regulated business needs. From fraud and lending to claims and portfolios, every engagement starts from an outcome you already measure.

Talk through your financial data See what we have built in financial services

Regional banks · Credit unions · Insurers · Asset and wealth managers

1FraudCard payment held: three countries in 40 minutes
2LendingDecline reasons ready for the adverse action notice
3ClaimsClaim fast-tracked: photos match the policy
4InvestmentsPortfolio fit shown with the checks it passed

Illustrative: every payment, loan, claim and portfolio decision, recorded with its reason.

Four exceptions we find at almost every firm

None comes from a shortage of data or models, and none is cleared by buying another platform. The right-hand column is what each one costs.

Exception	What we find	What it costs
One customer, five records	The core, the card platform, the loan or policy system, the CRM and the warehouse each hold a version of the customer. Seven figures to reconcile	Every household view, fraud pattern or cross-sell list starts with reconciliation, and the signals in the gaps are missed.
The inventory nobody can produce	Models live in spreadsheets, vendor tools and analysts’ notebooks, each documented a different way, if at all.	Nobody can say quickly which models are material, which are drifting or who answers for their outputs, and the inventory takes weeks to assemble for an examiner or auditor.
Generative AI parked in legal review	An assistant that drafts customer letters or answers product questions waits for a sign-off nobody knows how to give. Read a use case as a validator would	Capability already paid for sits unused, because nobody can show what it will and will not say.
Quarter end by hand	Regulatory filings, board packs and investor reports rebuilt in spreadsheets every quarter.	Expensive analyst time, later decisions, and figures that do not tie out the first time.

Engagements, with the figures the clients reported

Two engagements in financial services. The clients are described, not named.

A global asset manager

A decision layer inside its investment and risk platform

Portfolio, risk and mandate data the platform already held, turned into suitability-aware portfolio choices with their criteria shown, for investment, sales and risk teams alike.

Consolidated attributes · suitability scoring · role-aligned views · transparent criteria

Read the case study

Reported by the client

fewer manual comparison steps25%
faster portfolio-fit identification15%
better investment–distribution alignment32%

A global financial information and ratings provider

Contract compare-and-comply across MSAs, SOWs, POs and invoices

Terms extracted and checked against the approved clause library, exception-first review for legal and procurement, and purchase orders and invoices matched to the governing terms, with attorney review intact.

Read the case study

15–25% of documents nonstandard: the only ones that now need a reviewer

Seven figures that tie out only once they are defined

When finance, risk and the business report different figures, the dashboard is rarely at fault. The basis is: the definition, the source, the cut-off or the owner. Pick a figure to see each basis and the policy decision that ties it out.

Net interest marginBanking

Basis	How it is calculated
ATax-equivalent or not	Tax-exempt income grossed up to a taxable equivalent, or reported as booked.
BAverage or period-end	Earning assets averaged daily, monthly or taken at period end.
CWhich earning assets	Nonaccrual loans and certain securities included in the base, or left out.

The policy decision that ties it outTax-equivalent or not, and over which average of which earning assets?

Source systems: the general ledger, the core, treasury, regulatory reports.

Delinquency rateLending

Basis	How it is calculated
ACount or balance	The share of loans past due, or the share of balances, which large loans can swing.
BPast due from when	Counted from the contractual due date, or after a grace period.
CWhich loans in the base	Charged-off, modified and deferred loans counted or left out.
DHow late counts	30, 60 or 90 days and over, reported as one figure or as buckets.

Worked example, illustrative. 1,000 loans; the 40 that are 30 or more days past due are larger than average and hold 6% of the balance. The delinquency rate is 4.0% by count and 6.0% by balance.

The policy decision that ties it outCount or balance, from which due date, and which loans are in the base?

Source systems: the loan servicing system, collections, risk, regulatory reports.

Customer countRetail banking

Basis	How it is calculated
APeople, households or accounts	One person with three accounts counted once, three times, or as part of a household.
BActive means what	Any transaction in 90 days, a balance above zero, or a digital log-in.
CJoint and business accounts	Joint owners counted once or twice; a sole proprietor counted as a person or a business.

The policy decision that ties it outWho counts as one customer, and what makes them active?

Source systems: the core, the CRM, digital banking, marketing.

Loss ratioInsurance

Basis	How it is calculated
AIncurred or paid	Losses incurred, including reserves, or only what has been paid.
BWith or without adjustment expense	Loss adjustment expenses included in the ratio or reported apart.
CEarned or written premium	Losses against premium earned in the period, or premium written.
DCalendar or accident year	Losses booked this year, or losses from accidents that happened this year.

Worked example, illustrative. For every 100 of premium earned, 60 is paid in claims and 72 is incurred once reserves are added: a paid loss ratio of 60% and an incurred loss ratio of 72%.

The policy decision that ties it outIncurred or paid, with which expenses, against which premium, and by which year?

Source systems: claims, policy administration, actuarial, finance.

Combined ratioInsurance

Basis	How it is calculated
AStatutory or GAAP	Expenses measured against written premium in statutory reporting, or earned premium under GAAP.
BPolicyholder dividends	Dividends to policyholders counted in the ratio or left out.
CCatastrophes	Catastrophe losses included, or shown separately to reveal the underlying ratio.

The policy decision that ties it outStatutory or GAAP, and are catastrophes and dividends in?

Source systems: finance, actuarial, statutory filings, investor reporting.

Fraud loss ratePayments

Basis	How it is calculated
AGross or net	Losses before recoveries and chargebacks, or after.
BAgainst volume or count	Basis points of dollar volume, or a share of transactions.
CWhich losses count	Scams the customer authorized, and first-party fraud, counted or left out.
DWhich date	Counted by transaction date or by the date the fraud was reported.

The policy decision that ties it outGross or net, against volume or count, and which kinds of loss count?

Source systems: card and payment platforms, the fraud system, disputes, finance.

Assets under managementAsset management

Basis	How it is calculated
ARegulatory or marketing	The regulatory figure filed with the SEC, or the broader figure in marketing material.
BDiscretionary or advisory	Only assets you manage with discretion, or advisory and model-delivery assets too.
CCounted once	Assets in a fund that invests in your other funds counted once or twice.
DCommitted or invested	For private funds, capital committed or capital actually invested.

The policy decision that ties it outWhich assets count, counted once, and valued as of when?

Source systems: the portfolio accounting system, the CRM, regulatory filings, marketing.

Which of these gets a footnote in every board pack? Bring it to a 30-minute call, and we will tell you what it would take to define it once, at the source.

Book a call on one of these figures

The work, by line of business

Banks and credit unions, insurers, and asset and wealth managers: two pieces of work each, then the controls every firm needs.

Banks and credit unions

Fewer false alarms, faster cases

Lower false-positive rates and investigation cost, with detection coverage kept.

Fraud models stop losses as they happen; anti-money-laundering monitoring finds patterns for investigators to report. We join transaction, device and customer signals for both, rank alerts by real risk, and show investigators the reasons behind every score.

Start with AI & Agent Pilot

Faster credit decisions, with the reasons attached

Quicker approvals, sharper risk segmentation and fewer manual reviews.

Credit and pricing models built with the specific reasons for each decision, outcomes tested across groups before go-live, and the documentation your validators ask for.

Start with AI & Agent Pilot

Insurers

Faster claims triage, clearer exceptions

Shorter cycle times on simple claims, and adjusters’ time spent where judgment matters.

First notice of loss read and routed, simple claims fast-tracked, and complex or suspicious ones sent to the right adjuster with the documents already summarized. Adjusters decide denials.

Start with AI & Agent Pilot

Underwriting that shows its working

Lower underwriting expense and consistent risk selection, with the AI program regulators expect.

Submissions and external data read and summarized for underwriters, pricing and risk models documented and tested for unfair discrimination, and every model owned in a written AI program.

Start with AI Governance Quickstart

Asset and wealth managers

Portfolio choices with the criteria shown

Faster portfolio fit and closer alignment between investment and distribution.

Portfolio, risk and mandate data the platform already holds, turned into suitability-aware choices that show the criteria each one met.

Case study: A decision layer for a global asset manager

Advisor time back

More time with clients, and suggestions that pass suitability before they are made.

Meeting preparation, client summaries and suggested next steps drafted from the client’s own holdings and history, with mandate and suitability checks applied and the advisor reviewing everything before it reaches a client.

Start with AI & Agent Pilot

Across every firm

Model risk and AI governance

One inventory for statistical models, vendor scores and generative AI: owners, risk tiers, validation, performance and third-party AI, kept current.

Start with AI Governance Quickstart

Quarter-end reporting that ties out

Reconciliation across the general ledger, operational systems and regulatory reports automated, with lineage from every filed figure back to its source.

Start with Data Platform Accelerator

Contracts and documents

Agreements and filings checked against your clause library and policies, with only the exceptions sent to counsel.

Case study: Contract compare-and-comply

A line of business we have not listed, or a priority that cuts across all three? Thirty minutes is usually enough to agree the first step.

Talk through your line of business

The rules, dated, and the data they cover

The rules moved twice in 2026, and Colorado’s law arrives on January 1. Each date links to its note, as in the notes to the accounts; the systems the rules cover follow.

What changed, and when

Dec 3, 2025Regulation S-P amendments, larger firms
Apr 17, 2026SR 26-2 replaces SR 11-7
Jun 3, 2026Regulation S-P amendments, smaller firms
Today
Jan 1, 2027Colorado SB 26-189 takes effect

Note 1Issued Apr 17, 2026 · replaces SR 11-7
Interagency model risk management guidance (SR 26-2) (opens in a new tab)
Supervisory guidance, not a rule, and scaled to risk: the agencies expect it to matter most for banks with over 30 billion dollars in assets, and generally not for smaller banks’ models unless their model risk is significant. Credit unions answer to the NCUA. Generative and agentic AI are outside its scope for now.
Note 2In force
ECOA and Regulation B (opens in a new tab)
Applicants denied credit get the specific principal reasons, however complex the model that made the decision.
Note 3In force
Gramm-Leach-Bliley Act (opens in a new tab)
Customer financial information protected by a written security program, with privacy notices and limits on sharing.
Note 4Adopted state by state
The NAIC model bulletin on insurers’ use of AI (opens in a new tab)
Not a national rule: about half the states have adopted it, some with changes, and other state insurance laws can add to it. Where adopted, insurers keep a written AI program: governance, documented data and models, validation, monitoring and oversight of third-party AI.
Note 5Larger firms Dec 3, 2025 · smaller Jun 3, 2026
SEC Regulation S-P amendments (opens in a new tab)
Advisers, broker-dealers, funds and transfer agents keep an incident response program. When sensitive customer information has been, or is reasonably likely to have been, accessed without authorization, affected customers are generally told within 30 days, subject to the rule’s exceptions; service providers must tell the firm within 72 hours.
Note 6Colorado only · from Jan 1, 2027
Colorado’s law on automated decisions (SB 26-189) (opens in a new tab)
Where covered automated tools materially influence lending or insurance decisions about Colorado consumers: notice, an explanation after an adverse outcome, a way to correct data, and human review. The attorney general’s rules are due by January 2027.

A summary for orientation, not legal advice. Each rule links to its source. Last checked October 2026.

Source systems we reconcile

System	Formats and feeds	What it holds
Core banking and deposits	Accounts · transactions · general ledger	Accounts, balances, transactions and the ledger they post to.
Cards and payments	ISO 8583 · ISO 20022 · NACHA	Authorizations, settlements, ACH and wire activity, disputes and chargebacks.
Lending	Origination · servicing · bureau	Applications, decisions, bureau data, payments and collections.
Policy and claims	Policy administration · claims · ACORD	Policies, endorsements, premiums, first notice of loss, reserves and payments.
Investments	Holdings · trades · market data	Portfolios, positions, trades, mandates and the benchmarks they are measured against.
Customer and relationship	CRM · KYC · household	Know-your-customer records, relationships, households and service history.

Your AI use case, as a validator would read it

Pick a use case to see the decision it drives, its likely risk tier, the rules in play and the evidence a validator will ask for before go-live.

New in 2026: the banking agencies’ revised model risk guidance leaves generative and agentic AI outside its scope. Assistants, letter drafting and agents still need governing, through your own program for now.

Credit decision models

Decision it drives: Whether an applicant is approved, and on what terms.
Line of business: Banks and credit unions
Likely risk tier: High
Risk attributes: Customer financial information Yes
Speaks to customers Possibly
Decides credit, price, coverage or a claim Yes
Independent validation likely Yes
Could treat groups of customers differently Yes
Rules in play: ECOA and Regulation B: the specific principal reasons for every adverse action.
Fair lending laws: no applicant treated differently on a prohibited basis, which means testing outcomes, not only inputs.
Model risk guidance (SR 26-2) at larger banks, or your own model governance: validation and monitoring scaled to the model’s risk.
Colorado, from January 2027: notice, an explanation after a denial, data correction and human review.
Evidence a validator will ask for: Specific reasons generated with every decision and matched to the notice.
Outcomes tested across groups before go-live and on live decisions.
Validation by someone who did not build the model.
First step: Start with AI Governance Quickstart

Customer financial information	Yes
Speaks to customers	Possibly
Decides credit, price, coverage or a claim	Yes
Independent validation likely	Yes
Could treat groups of customers differently	Yes

Fraud and anti-money-laundering alerts

Decision it drives: Which payments are held and which alerts investigators see first.
Line of business: Banks and credit unions
Likely risk tier: High
Risk attributes: Customer financial information Yes
Speaks to customers No
Decides credit, price, coverage or a claim Possibly
Independent validation likely Yes
Could treat groups of customers differently Possibly
Rules in play: Bank Secrecy Act: suspicious activity reported within set timeframes, so a model is judged by what it misses as well as what it catches.
Model risk guidance (SR 26-2) at larger banks, or your own model governance: alert models validated and monitored in proportion to their risk.
Evidence a validator will ask for: Alert levels tuned on your own history, with samples below the line reviewed too.
Investigators see the reasons behind every score.
Model changes logged and approved like any other model change.
First step: Start with AI & Agent Pilot

Customer financial information	Yes
Speaks to customers	No
Decides credit, price, coverage or a claim	Possibly
Independent validation likely	Yes
Could treat groups of customers differently	Possibly

Customer service assistant

Decision it drives: What customers are told about products, fees and their accounts.
Line of business: Banks, credit unions and insurers
Likely risk tier: Moderate
Risk attributes: Customer financial information Yes
Speaks to customers Yes
Decides credit, price, coverage or a claim No
Independent validation likely Possibly
Could treat groups of customers differently Possibly
Rules in play: Outside the 2026 model risk guidance: generative AI is governed by your own program.
Gramm-Leach-Bliley Act: identity verified before any account information is shared.
UDAAP: what it says about products, fees and rights must be accurate and not misleading.
Evidence a validator will ask for: Answers only from approved product terms and fee schedules.
Disputes, complaints and hardship requests handed to a person.
Conversations logged and sampled for quality.
First step: Start with AI & Agent Pilot

Customer financial information	Yes
Speaks to customers	Yes
Decides credit, price, coverage or a claim	No
Independent validation likely	Possibly
Could treat groups of customers differently	Possibly

Drafting customer letters and notices

Decision it drives: What customers receive in writing, including notices with required content.
Line of business: Banks, credit unions and insurers
Likely risk tier: Moderate
Risk attributes: Customer financial information Yes
Speaks to customers Yes
Decides credit, price, coverage or a claim No
Independent validation likely Possibly
Could treat groups of customers differently Possibly
Rules in play: Outside the 2026 model risk guidance: generative AI is governed by your own program.
Required notices, such as adverse action, privacy and claim decisions, keep their required content.
UDAAP: accurate and not misleading.
Evidence a validator will ask for: Required wording locked; the model drafts only what it is allowed to change.
One common model: compliance approves the templates, and letters with regulated content are reviewed one by one.
Every letter kept as sent.
First step: Start with AI & Agent Pilot

Customer financial information	Yes
Speaks to customers	Yes
Decides credit, price, coverage or a claim	No
Independent validation likely	Possibly
Could treat groups of customers differently	Possibly

Underwriting and pricing models

Decision it drives: Who is offered cover, and at what price.
Line of business: Insurers
Likely risk tier: High
Risk attributes: Customer financial information Yes
Speaks to customers No
Decides credit, price, coverage or a claim Yes
Independent validation likely Yes
Could treat groups of customers differently Yes
Rules in play: NAIC model bulletin, adopted in about half the states: a written AI program covering governance, testing and third-party models.
Rate and rule filings: rating factors filed with, and justified to, each state.
Colorado, from January 2027: notice, an explanation after an adverse outcome, data correction and human review.
Evidence a validator will ask for: Every external data source documented, with the reason it is used.
Outcomes tested for unfair discrimination before filing and after go-live.
A named owner for each model in the written AI program.
First step: Start with AI Governance Quickstart

Customer financial information	Yes
Speaks to customers	No
Decides credit, price, coverage or a claim	Yes
Independent validation likely	Yes
Could treat groups of customers differently	Yes

Claims triage and fraud flags

Decision it drives: How fast a claim moves, and who handles it.
Line of business: Insurers
Likely risk tier: High
Risk attributes: Customer financial information Yes
Speaks to customers Possibly
Decides credit, price, coverage or a claim Yes
Independent validation likely Possibly
Could treat groups of customers differently Possibly
Rules in play: Unfair claims settlement practices laws: claims handled promptly and fairly, with decisions an adjuster can explain.
NAIC model bulletin, where adopted: AI used in claims sits inside the written AI program.
Evidence a validator will ask for: The model fast-tracks or routes; adjusters decide denials.
Reasons recorded for every routing decision.
Fast-tracked claims audited by sample.
First step: Start with AI & Agent Pilot

Customer financial information	Yes
Speaks to customers	Possibly
Decides credit, price, coverage or a claim	Yes
Independent validation likely	Possibly
Could treat groups of customers differently	Possibly

Advisor and relationship manager assistant

Decision it drives: What advisors suggest to clients.
Line of business: Asset and wealth managers
Likely risk tier: High
Risk attributes: Customer financial information Yes
Speaks to customers Possibly
Decides credit, price, coverage or a claim Possibly
Independent validation likely Possibly
Could treat groups of customers differently No
Rules in play: Regulation Best Interest and fiduciary duty: recommendations suit the client, with the basis documented.
Communications rules: anything that reaches clients is reviewed as required and retained.
Regulation S-P: client information protected, with notice within 30 days after a breach.
Evidence a validator will ask for: Suitability and mandate checks applied before anything is suggested.
Every suggestion shows the factors it met.
Drafts reviewed by the advisor before they reach a client, and kept.
First step: Start with AI & Agent Pilot. Case study: A decision layer for an asset manager

Customer financial information	Yes
Speaks to customers	Possibly
Decides credit, price, coverage or a claim	Possibly
Independent validation likely	Possibly
Could treat groups of customers differently	No

Contract and document review

Decision it drives: Which contract terms reach counsel for review.
Line of business: Every firm
Likely risk tier: Low
Risk attributes: Customer financial information Possibly
Speaks to customers No
Decides credit, price, coverage or a claim No
Independent validation likely No
Could treat groups of customers differently No
Rules in play: Confidentiality: client and counterparty documents stay inside your controls.
Records: findings kept with the documents they came from, for as long as your retention rules require.
Evidence a validator will ask for: Checks run against your own clause library and policies.
Exceptions go to counsel; nothing is approved automatically.
An audit trail on every finding.
First step: Start with AI & Agent Pilot. Case study: Contract compare-and-comply

Customer financial information	Possibly
Speaks to customers	No
Decides credit, price, coverage or a claim	No
Independent validation likely	No
Could treat groups of customers differently	No

A first read, not a determination. What applies depends on your institution, who supervises it, its size and how the tool is used; it is not a legal or regulatory determination. The rules in full: The rules, dated, and the data they cover.

High: data protection, model validation, fair-treatment testing, compliance and governance sign-off.
Moderate: data protection, testing on your own data, and a named owner.
Low: content ownership, access and a periodic check.

A model not on the list, or one already in production without this evidence? Thirty minutes is usually enough to tier it and name the first step.

Talk through one model or assistant

Three ways in, smallest commitment first

A six-minute self-check, a two-week sprint, then one decision through governance or one bounded build. Each ends with a decision, and none commits you to the next.

1Free · about six minutes
Data Maturity Assessment
Fifteen questions on lineage, quality, ownership, access and platform: what an auditor asks of any figure. An instant result, and no email required.
Score your data foundation
2Two weeks
AI & Data Assessment Sprint
Readiness scored against your own systems, the gaps on one register with a view of the cost, and the next steps in order. The right start when the question is where to begin.
Ask about a two-week sprint
32 to 4 weeks, once access is in place
A first decision or a first build
One AI use case through governance, one use case running, or one data foundation, built in your environment once access, security and vendor approvals are in place, against a number you already track, then a decision to scale, change or stop.
AI Governance Quickstart · AI & Agent Pilot · Data Platform Accelerator
Talk through a first step

Questions from risk, compliance and the business

Does customer data leave our controls?

No. We work in your cloud, on your accounts, under your entitlements and your third-party risk review, and what we build is yours to run. The first diagnostic usually needs definitions, metadata and a sample, not full customer records. Where the work needs customer information, your vendor and security reviews come first.

Can you work inside our model risk management program?

Yes, and we build for it from the start: the inventory entry, the development documentation, the testing evidence and the monitoring plan your validators ask for, so independent validation reviews a finished package rather than a reconstruction.

Does the new model risk guidance apply to us?

It depends on who supervises you and your size. The banking agencies expect SR 26-2 to matter most for banks with over 30 billion dollars in assets, and generally not for smaller banks’ models unless their model risk is significant; credit unions answer to the NCUA. Whatever applies, a proportionate inventory, testing and monitoring are what auditors, boards and examiners ask about.

Our generative AI is outside the new model risk guidance. Does it still need governing?

Yes. The revised guidance issued in April 2026 leaves generative and agentic AI out of scope for now, and the agencies plan to ask the industry about AI model risk. Until they say more, the questions are the same ones your program already asks: what it is used for, how it was tested, what it must never say or do, and who watches it. The AI Governance Quickstart takes one real use case through that path.

Do you work with insurers and asset managers, or only with banks?

With all of them: banks, credit unions, insurers, and asset and wealth managers. The work differs: for banks and credit unions it is usually fraud, credit and the customer view; for insurers, claims, underwriting and the written AI program; for asset and wealth managers, portfolio data and advisor tools.

Our core and CRM vendors now ship AI. Where does building still pay?

Where a question crosses systems no single vendor sees, such as fraud across cards and deposits, or a household view across products. Elsewhere, use what your vendors ship; it is often the fastest route. We help you decide which features to switch on and bring them under your model risk and third-party risk programs.

What is the smallest first commitment?

Six minutes and no email: the free Data Maturity Assessment. After that, the Assessment Sprint takes two weeks. A first decision through governance, or a first build, takes two to four weeks once access and approvals are in place, and ends with a decision to scale, change or stop.

Bring the question nobody can answer with one number.

The model inventory, the alert backlog, the ratio nobody agrees on, or the assistant still waiting for legal. Thirty minutes with the person who would lead the work, on what it would take to close it out.

Book 30 minutes on that question

A figure that will not tie out? Thirty minutes on the figure, the model or the backlog.

Book time on your financial data

Financial data that ties out, and AI decisions you can defend.

Four exceptions we find at almost every firm

Engagements, with the figures the clients reported

A decision layer inside its investment and risk platform

Contract compare-and-comply across MSAs, SOWs, POs and invoices

Seven figures that tie out only once they are defined

Net interest marginBanking

Delinquency rateLending

Customer countRetail banking

Loss ratioInsurance

Combined ratioInsurance

Fraud loss ratePayments

Assets under managementAsset management

The work, by line of business

Banks and credit unions

Fewer false alarms, faster cases

Faster credit decisions, with the reasons attached

Insurers

Faster claims triage, clearer exceptions

Underwriting that shows its working

Asset and wealth managers

Portfolio choices with the criteria shown

Advisor time back

Across every firm

Model risk and AI governance

Quarter-end reporting that ties out

Contracts and documents

The rules, dated, and the data they cover

Interagency model risk management guidance (SR 26-2)↗ (opens in a new tab)

ECOA and Regulation B↗ (opens in a new tab)

Gramm-Leach-Bliley Act↗ (opens in a new tab)

The NAIC model bulletin on insurers’ use of AI↗ (opens in a new tab)

SEC Regulation S-P amendments↗ (opens in a new tab)

Colorado’s law on automated decisions (SB 26-189)↗ (opens in a new tab)

Source systems we reconcile

Your AI use case, as a validator would read it

Credit decision models

Fraud and anti-money-laundering alerts

Customer service assistant

Drafting customer letters and notices

Underwriting and pricing models

Claims triage and fraud flags

Advisor and relationship manager assistant

Contract and document review

Three ways in, smallest commitment first

Data Maturity Assessment

AI & Data Assessment Sprint

A first decision or a first build

Questions from risk, compliance and the business

Bring the question nobody can answer with one number.

Interagency model risk management guidance (SR 26-2) (opens in a new tab)

ECOA and Regulation B (opens in a new tab)

Gramm-Leach-Bliley Act (opens in a new tab)

The NAIC model bulletin on insurers’ use of AI (opens in a new tab)

SEC Regulation S-P amendments (opens in a new tab)

Colorado’s law on automated decisions (SB 26-189) (opens in a new tab)