The fraud you are pricing in is the fraud you have not detected.
Every basis point of undetected fraud is a basis point of loan pricing your good borrowers subsidise. We build fraud brains that combine identity, device, graph, behavioural and post-drawdown signals into a single defence — for consumer lenders, SME lenders, bridging shops and BNPL alike.
Why the fraud you are pricing in is the fraud you have not detected.
Most lenders under £500m book operate with a fraud stack of Sumsub or Onfido for ID, a set of internal rules for velocity and duplicates, and a manual fraud team for review. The good news is the obvious fraud gets caught. The bad news is the sophisticated fraud — synthetic identities aged over 12 months, coordinated bust-out rings across guarantors, first-party fraud with real IDs and manufactured cashflows, mule networks that pass every check individually — walks straight through and lands on the loss line.
The pricing effect is compounding. Every basis point of undetected fraud is a basis point of pricing your good borrowers pay. The lenders that have shipped a proper fraud brain over the last three years have moved this number 20–50 basis points — real money on a mid-sized book, transformative on a subprime one.
What good fraud detection looks like in 2026.
Not another vendor bolted onto Sumsub. A firm-specific brain that produces:
- A per-application fraud score combining ID, device, IP graph, behavioural, cashflow-pattern and application-consistency signals — scored in seconds.
- A graph engine that surfaces coordinated rings across guarantors, addresses, devices, employer references and payment accounts — the fraud your rule engine cannot see.
- First-party fraud detection tuned to the specific patterns your book is bleeding on (cashflow manufacturing, employment falsification, guarantor collusion, address hopping).
- Post-drawdown fraud surveillance — bust-out detection, mule-net signals, sudden behaviour changes.
- A fraud operator UI that ranks cases with reasoning, not a wall of alerts.
What KJ Capital ships.
A bespoke fraud brain that sits above your existing ID/AML vendors (Sumsub, Onfido, SEON, ComplyAdvantage) and your LMS. We deliver:
- A production per-application fraud score trained on your book, evaluated against a hold-out cohort with published precision, recall and cost-weighted metrics.
- A graph engine spanning applications, guarantors, addresses, devices, employers and payment accounts.
- First-party fraud detectors calibrated to your specific loss patterns.
- Post-drawdown surveillance signalling on active book.
- A fraud operator UI plus a review-pack generator for the compliance-adjacent workflows.
What lenders actually see after go-live.
What buyers ask us about this build.
Q01Do you replace Sumsub / Onfido / SEON?+
Q02How do you handle GDPR and graph data?+
Q03Can this help with BNPL / subprime volumes?+
Q04How does this integrate with post-drawdown collections?+
Q05Price?+
What lenders ship alongside this one.
AI credit decisioning
Credit decisioning is the single most defensible AI investment a lender makes. We build firm-specific scorecards that combine open banking, bureau, alt-data and application signals — with explainable adverse-action notices, live A/B champion-challenger, and full audit trail for the FCA and your risk committee.
Application & origination
Every abandoned application is an acquisition cost wasted. We build AI-native application flows that pull open-banking, run KYC/KYB across Sumsub/Onfido, extract documents automatically, decision in seconds and hand a fully-packaged case to your underwriter or straight to drawdown.
Collections & arrears
The best collections shop is the one that never has to collect. We build pre-arrears signal engines, Consumer Duty-safe outreach, forbearance and restructure workflows, and vulnerable-customer handling that FCA supervisors respect and roll rates rebel against.