Extract the logic. Keep your options open.
Turn opaque spreadsheet and legacy-platform models into source-cited logic maps, reviewable specifications and executable behaviour tests. Use the evidence to govern the model where it is—or as the controlled baseline for a later migration.
Results from the public logic-extraction and migration demonstration.
Critical logic should not live only in formulas and memory
In mature models, rules are spread across formulas, tables, switches, macros and historical workarounds. Documentation often describes the intent without capturing the exact implemented behaviour.
That opacity slows review, onboarding and controlled change even when the platform stays. During migration, the same knowledge gap becomes a delivery and validation risk.
Turn behaviour into evidence
The harness turns an opaque model into a sequence of comprehension, specification and acceptance gates.
Census the source
Map every formula and calculation node, including unused and dead calculations.
Freeze the baseline
Extract golden values mechanically from the current model with engine, scenario and period recorded.
Test comprehension
Write a source-cited specification, then ask a fresh agent to predict intermediate values without seeing the answers.
Codify the logic
Create a reviewable functional specification, dependency map and executable reference tests without requiring a new platform.
Stress the branches
Exercise table boundaries, switches, partial years, negative rates and original error propagation.
Gate the evidence
Block completion until every declared comparison passes and each ambiguity, deviation or override is recorded.
A dossier, not a reconciliation memo
The evidence survives the build session and can be reviewed independently.
Source-to-module traceability
Specifications and implementation point back to the source cells and rules they reproduce.
Scenario and branch coverage
Every declared scenario and intermediate result is compared against its recorded baseline value.
Error fidelity
An original #N/A remains an #N/A; returning a polite zero is treated as a failure.
Reusable control baseline
The same evidence can assess current-platform changes, an independent rebuild or a future migration.
Extracted does not mean approved
The process demonstrates that the extracted specification and tests represent the source model's behaviour under the named engine, scenarios and declared tolerance. It does not prove that the original actuarial assumptions or logic were appropriate.
What this does not claimWhat the engagement leaves behind
Engagement outputWhat stakeholders usually need resolved
Questions & boundariesDo we need to migrate to get value from this?
No. The logic map, sourced specification and test baseline improve documentation, review, onboarding and change control on the current platform. Migration is one possible later application.
Why test behaviour before improving the model?
A tested baseline separates what the model does today from any proposed improvement. Each later change can then be attributed, challenged and approved on its own merits.
Does logic extraction depend on AI?
No. AI can accelerate inventory and specification, but acceptance comes from source traceability, deterministic comparisons, declared tolerances and actuarial review.
What is needed to start?
A source-model inventory, representative inputs, authoritative outputs, its current execution environment and access to people who can resolve intentional behaviour, known defects and judgmental rules. A target platform is only needed if migration is in scope.
Related control patterns
Plain language in. Governed actuarial work out.
Give actuarial teams a natural-language interface without turning production into a chat. Skills define the procedure, a harness executes and records it, and the approved engine performs the calculations.
Explore this solutionSolution · Documentation & regulatory reviewDefine good. Then make the work pass.
Turn actuarial and regulatory expertise into evidence rules, procedures, checks and stopping conditions that AI must follow. The report becomes a review interface, not an unsupported draft.
Explore this solutionNeed to understand a model before changing—or moving—it?
Start with the current model, the questions its documentation cannot answer and the evidence your reviewers need.