Dawid Kotur, CEO and co-founder of Curvestone AI, argues that lenders should point AI at bounded, repetitive, document-heavy compliance checks first. They should keep a person responsible for judgment, and they should require a record of the evidence behind every result. That is his case, made in interviews with Modern Lender (6 July 2026) and The Intermediary (3 August 2026). It is not independent proof that AI can safely review every lender’s cases. This piece sets out the argument, the build-versus-buy reasoning, and the limits of the evidence.
What “built narrow” means in practice
The phrase is our shorthand for the approach Kotur describes, not a quotation from either interview. It means starting with one specific workflow and the lender’s own criteria. Expansion comes later, and only when the evidence, controls and operational readiness support it. The interviews point to high-volume routine checks as the starting point, rather than an attempt to transform a whole business at once. Experienced staff are then left to focus on judgment calls and edge cases.
As an Amazon Associate I earn from qualifying purchases.
The reason is the problem Kotur describes. In The Intermediary he says manual compliance sampling typically covers “around 10%” of cases. That is his description, not an independently checked industry statistic. He ties the push for wider oversight to the regulatory context of Consumer Duty and the weakness of relying on small samples. The interviews do not establish that Consumer Duty requires full AI review of every case, and no regulator is quoted.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteHow the described workflow operates
Curvestone says its software reviews mortgage and commercial-finance cases. The material it works from is messy: scans, photos, emails, call transcripts, fact-finds, bank statements and IDs. The steps, as described:
- The system reads the varied case materials.
- It checks them against criteria the lender defines.
- It produces findings with reasoning and source evidence.
- A human reviewer approves or overrides each finding.
- Exceptions go to a specialist, and an audit trail is kept.
The company also claims to process “thousands of checks a quarter”. These are company and interview descriptions. The sources do not include independent testing of the product, an accuracy rate, a controlled comparison against manual review, or verified return on investment.
The defensibility test
The sharpest line comes from The Intermediary, which attributes this to Kotur: “if you cannot explain and defend an automated decision after the fact, you shouldn’t be using it.” This is the practical reason for narrowness. A tightly scoped check against stated criteria is easier to trace than a broad system making open-ended calls. Traceability and human accountability are presented as the safeguard against opaque “black box” decisions. It is a test from an interviewee, not official regulatory guidance.
Rank #2
Build or buy: where the argument lands
Modern Lender attributes this to Kotur: “Most teams don’t regret the initial build. They regret year two.” The point is that this is an ongoing operations decision. Curvestone’s own adapted article says internal systems need continuing maintenance as documents, regulation and models change. It also estimates that an internal system takes “12 to 18 months” to reach production-grade. That is the vendor’s estimate, not a verified sector average.
Recommended Free Tools
The sources support these axes for comparing the two routes, though they offer no neutral cost model or head-to-head trial:
| Axis | What to ask |
|---|---|
| Control | How far can you customize policy and lender-specific criteria? |
| Time and staffing | What specialist skills and months are needed to reach production? |
| Maintenance | Who monitors and updates as data formats, rules and models change? |
| Integration | How does it fit your existing systems? |
| Evidence and explainability | Does every result retain reasoning and source evidence? |
| Human oversight | Can reviewers approve, override and escalate? |
| Evaluation | Can you test performance on your own real cases before committing? |
How far to trust the claims
Treat the interviews as informed advocacy from a vendor chief executive. The Modern Lender piece and Curvestone’s own article are closely linked, since the company says its article is adapted from the interview. The Intermediary interview adds overlapping themes. The useful takeaways are the questions to put to any supplier. Is the scope narrow and clearly defined? Is a named person accountable for each judgment? Can you reconstruct why a result was reached? Can you measure it on your own cases?
Quick Recap
Rank #4
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
