A mid-market financial-services lender operating in a heavily regulated jurisdiction, originating secured and unsecured loans across multiple products. Roughly 200 staff across origination, credit, compliance, and operations.
Loan origination was a chain of manual hand-offs: due diligence in spreadsheets, credit decisions in email threads, compliance checks re-keyed between systems, and disbursement gated by manual sign-offs. Cycle times were measured in days, and every exception required a human to chase context across four teams.
We rebuilt the lending operation as an AI-native workflow. Coordinated agents run due diligence, assemble the credit file, apply policy, and route disbursement, coordinating across every stakeholder through the Negotiate protocol. The same Kolega agent that powers our products runs underneath, so exceptions are handled in context instead of escalated blind.
Connected to the existing core banking ledger, the KYC/AML provider, the document store, and the e-signature platform, so the workflow runs on the systems of record already in place, with no rip-and-replace.
Every agent action is logged to an immutable audit trail. Policy is expressed as enforced rules, not guidance; high-risk decisions route to a human approver; and all code shipped through the build was scanned and remediated with Kolega DevSec before release. Controls aligned with SOC 2 and ISO 27001.
A 300% increase in operational efficiency on lending throughput, measured over the first two quarters in production against the prior twelve-month baseline. Cycle times dropped from days to hours, and the compliance team moved from re-keying data to reviewing exceptions.