cs.SESep 19, 2026

From Code to Requirements: Agentic Reverse Engineering of Business Rules at Enterprise Scale

Authors: Garima Agrawal, Prasun Das, Priyanka L, Minisha N, Akshay Sarvade, Hemath Manivanan, Sravani Joshna, Prasad Kalyansundaram

Organizations: Cognizant, Plano, Texas, USA

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Business requirements for enterprise software systems are rarely captured in structured form; the logic resides instead in source code, configuration files, and institutional memory. When these systems must be migrated, extended, or audited, the absence of formal requirements artifacts forces teams into expensive, knowledge-losing manual reverse engineering. This paper presents an agentic framework that autonomously generates Business Requirements Documents (bRDs) through reverse engineering of undocumented enterprise software. Seven specialized agents collaborate to discover user journeys, extract business rules, and synthesize a locked bRD from code, test suites, configuration, and available documentation, supported by static analysis tools, guided by embedded expert practitioner cognitive models, and overseen by human reviewers at controlled escalation points. Measured on actual execution traces across a real enterprise deployment, the framework generates comprehensive bRDs in under nine minutes per service, with extracted rules independently corroborated against real production defect records. In a comparable prior migration to the same target architecture, delivery required over two years; on the present programme the framework achieves a cost reduction exceeding 98% against a baseline derived from actual repository metrics using IFPUG complexity models and industry benchmark labor rates.

Figures & tables

Figure 1: Seven-agent bRD generation pipeline. Agents 2+3 and Agents 6+7 execute in parallel. All agents read from and write to the shared state store. The knowledge layer compounds organizational knowledge across successive pipeline runs.
Figure 2
Figure 3
AgentInput SourcesStatic ToolsLLM RoleOutput
A1 DiscoveryRepos, docs, ticketsFile scanner, dependency mapperScope synthesisScope doc, artifact catalog
A2 JourneyCode, test suites, API specs, docsAST parser, test suite minerJourney typing, conflict resolutionJourney inventory, coverage matrix
A3 RulesSource code, config files, docsPattern matcher, AST parserRule translationRules catalog, regulatory flags
A4 GapJourney inventory, rules, target specsSchema diff, API parserGap classificationGap register, risk scores
A5 SynthesisAll agent outputs, knowledge layerVector storeConflict resolution, bRD assemblyLocked bRD (human + machine)
A6 ACLocked bRD, test suiteRegression comparatorAC authoringAC document, Gherkin feature file
Table 1: Agent pipeline: input sources, static tools, LLM role, and output artifact.
MetricTotalAvg/Service
Business journeys28020.0
Business rules1,612115.1
Test cases mapped32723.4
Journey range3–42—
Rules range10–187—
Tests range3–42—
Table 2: Coverage footprint across 14 measured services.
Figure 6
Figure 7
MetricMinMax
Production LOC556107,810
Test LOC082,378
Class count221,791
Method count4715,228
IFPUG CI0.0010.957
Manual hours (repo)483,896
Table 3: Repository sample characteristics (19 of 39 repositories; identifiers anonymized).
MetricMinMaxAvg/Svc
Input tokens5,50045,23029,482
Output tokens3,30028,15018,211
Total tokens8,80073,38047,694
Cost – Haiku ($)0.0140.1040.066
Generation (min)3.0012.458.78
Table 4: Agentic token measurements across 14 services (LangSmith traces).
MetricManual (derived)Agentic (measured)
Scope39 repositories18 services, 39 repos automated
Total human hours34,981 hrs∼\sim 18 hrs review
Person-months230 PM<< 1 PM recurring
One-time setupN/A∼\sim 3 months (non-recurring)
Team of 5 duration∼\sim 40 months<< 1 week
LLM inference costN/A$1.18 (Haiku benchmark)
Table 5: Programme-level comparison: manual derived vs. agentic measured (39-repository programme scope).

Explore similar work

CardsList

  1. Reversa: A Reverse Documentation Engineering Framework for Converting Legacy Software into Operational Specifications for AI Agents

    May 18, 2026Sanderson Oliveira de Macedo, Ronaldo Martins da CostaSoftware EngineeringSoftware Engineering Agents

    Federal Institute of Goias - Brazil · Federal University of Goias - Brazil

  2. ReqFusion: A Multi-Provider Framework for Automated PEGS Analysis Across Software Domains

    Mar 24, 2026Muhammad Khalid, Manuel Oriol, Yilmaz UygunRequirements EngineeringAutomated Software Engineering

    Constructor University Bremen, 28759 Bremen, Germany

  3. An Industrial-Scale Retrieval-Augmented Generation Framework for Requirements Engineering: Empirical Evaluation with Automotive Manufacturing Data

    Mar 20, 2026Muhammad Khalid, Yilmaz UygunRequirements EngineeringRAG Evaluation

    Constructor University Bremen, Bremen, Germany