Forty-year-old logic, explained in plain English before you touch it.
Turns the explainability engine on old code instead of models or agents: reads a decades-old system, explains what it actually does and why, and produces a dependency map before anyone attempts a migration. Roughly 220 billion lines of COBOL still run production systems in banking, government, and healthcare — and the average COBOL programmer is 55, with about a tenth of that workforce retiring every year.
Reads decades-old systems and explains what they actually do, not just what the code syntactically says.
Produces a map of what a legacy system reads from and writes to before anyone attempts to touch it.
Surfaces the specific logic most likely to break or need manual review during a migration attempt.
Planned design — this pipeline does not exist yet.
Reads a legacy codebase — COBOL and similar decades-old systems are the named starting point.
Applies Decision Intelligence's explainability approach to code instead of model decisions, producing a plain-English summary.
Documents upstream/downstream dependencies and flags logic most likely to need careful handling during migration.
Routine by routine, as LegacySense reads more of a codebase, the map fills in.
Illustrative shape of the output only. No codebase has been mapped — the map, the explanations, and the "explained / unmapped" states shown here are a design target, not a screenshot.
The reasoning approach this concept is designed around — a path from cryptic code to a plain-English read, one step at a time.
Concept only — no reasoning path has been built or tested.
The core reframing: the same 'why did this happen' approach Decision Intelligence applies to model decisions, applied to what a legacy system's logic actually does.
Points to a specific, verifiable pressure: a large volume of COBOL still runs production systems, and the workforce that understands it is aging out.
The map comes before the migration attempt, not after something breaks — the goal is to de-risk a migration, not clean up after one.
Depends on Decision Intelligence's explainability core existing first — LegacySense is an application of that engine, not a separate one.
Nothing below is built. This describes the intended design and its dependencies — laid out here as a translation view, since that's the actual job: old logic in, plain English out.
PERFORM VARYING WS-IDX
FROM 1 BY 1
UNTIL WS-IDX > WS-MAX-REC
IF WS-REC-FLG(WS-IDX) = '1'
COMPUTE WS-INT-AMT =
WS-PRIN * WS-RATE-TBL(WS-IDX)
ADD WS-INT-AMT TO WS-BAL
END-IF
END-PERFORM."For every active record, calculate interest using that record's entry in the rate table, then add it to the running balance."
Shape only — which routines a piece of legacy logic calls. Not yet designed.
Named starting point: COBOL and similar legacy production code.
Produces the map before anyone attempts a migration, not after something breaks.
Applies the same explainability engine to code instead of models.