Run credit decisions from policy, change them yourself
Turn a credit policy PDF into tested decision logic you can run yourself.
Transcript
This is a credit policy. Three pages of rules, scoring and pricing, and written by a risk team, not an engineer.
Watch what happens when you hand it to Decisimo.
It reads the document and pulls out every requirement, then works out which ones are rules, which is a scorecard, and which is a pricing tree.
Then it tells you exactly what it will build, and draws the flow before building anything.
Every value is checked back against your document. It cannot invent one.
And nothing is built until you approve it.
Now it builds the real, executable logic.
Open any one of them and the real rules are there, with their real conditions. Age. Sanctions. Affordability. Every one of them editable.
And the pricing tree carries every branch the document described. Loan type, then product category, then amount, with every rate exactly as written.
Every artifact created, from a single PDF. Versioned, testable, ready to review.
With a live endpoint, an API key, and test data drawn from the logic itself.
From a policy document to a running decision flow. No code.
Decisimo. Decision logic that's easy to build, and easy to change.
AI drafts from policy, a person approves each change
- Policy to flowA PDF becomes proposed rule sets, decision tables and lookups
- Grounded valuesUngrounded values go back for correction against the document and data.
- Review each changeA person sees the diff and approves each proposed edit
See the effect of a change before release
- Unit testsDecision tables suggest row tests, and scorecards suggest one per bin
- Regression suitesRe-run saved cases and compare current logic with proposed logic
- Release diffA release has a computed diff, and an approval sequence can be set
Explain any decision months later
- Execution traceSee each action’s order, inputs, outputs and duration
- Rule and scoreSee which rule fired and each predictor’s contribution
- Revision historyThe version remains readable, and the audit log records who viewed, edited or deployed it
The same artifacts run decisions in milliseconds or over weeks
- Real-time decisioningA request arrives, the decision flow runs, and the answer comes back
- Case workA case file reruns logic when new evidence makes an earlier conclusion out of date.
Where data, logic and credentials live
- Decision resultsWrite them to a table in the lender’s own Postgres
- Decision logicExport it as a Markdown file
- Provider credentialsIn an encrypted vault against the lender’s own provider contracts
Capabilities in one table
| Area | Capabilities |
|---|---|
| Decision artifacts |
|
| Data |
|
| Flow |
|
| Testing and assurance |
|
| Governance |
|
| Execution |
|
| Explainability |
|
| Security |
|
| AI |
|
| Case work |
|
Questions we get asked
What is a decision engine?
A decision engine applies defined logic to input data and returns a decision. Decisimo runs rule sets, decision tables, scorecards and other logic as a decision flow.
Do we have to use AI to build decision logic?
The same editors let your team build every artifact by hand.
What stops AI from inventing a threshold?
A grounding check compares proposed values with the policy document and existing data. A person reviews and approves each change.