When Legacy Has No Source Code: Recovering Black-Box Applications with AI and Human Engineering

A business-critical application is still running, but the people who understood it have moved on. Documentation is missing or obsolete. Current-state architecture is unclear. Dependencies are buried across aging platforms. In the worst cases, there is not even usable source code to inspect. Yet the system still carries rules, calculations, workflows and exceptions the business depends on every day.

This is where many transformation efforts stall. Leaders know the platform must change, but they cannot risk breaking behavior they can no longer fully explain.

Publicis Sapient approaches this problem differently. With Sapient Slingshot and experienced engineering oversight, we help organizations recover what their systems know before new code is produced. That means uncovering hidden business rules, tracing dependencies, generating specifications and rebuilding confidence in the modernization path. AI is not used as a shortcut around understanding. It is used to make understanding possible again.

The real problem is lost system knowledge

In undocumented legacy environments, the technical challenge is only part of the issue. The deeper problem is that business logic has become trapped inside software artifacts the organization can no longer interpret easily. Rules live in old code, obscure workflows, tightly coupled integrations and decades of layered changes. What once existed as shared institutional memory is now fragmented across repositories, tickets, spreadsheets and tribal knowledge.

Traditional modernization methods struggle here. Rewrite-from-scratch efforts rely on assumptions. Manual discovery takes too long. SMEs are scarce, expensive and often uncertain themselves. Teams end up choosing between slow, expensive archaeology and risky reinvention.

That is why black-box recovery requires a different starting point. Before design, before code generation and before migration planning, teams need a reliable way to turn hidden system behavior into usable knowledge.

From code to specifications to confidence

Sapient Slingshot is designed to modernize legacy systems by automating and accelerating the full software development lifecycle while preserving critical business logic. In black-box scenarios, that starts with discovery.

Slingshot can ingest legacy applications, analyze their structure, trace data entities and logical flows, and surface multi-level dependencies that would be difficult to map manually. Instead of jumping directly from old systems to new code, it inserts a specification layer in between. That specification becomes a testable source of truth for what the system does, how it behaves and what must be preserved.

This is a critical difference. Many modernization tools focus on conversion speed. Publicis Sapient focuses first on comprehension. By translating opaque systems into business and functional specifications, teams can validate logic earlier, reduce guesswork and modernize with stronger control.

For organizations facing undocumented estates, that recovered knowledge is often the most valuable modernization asset of all.

AI that works with engineers, not around them

Black-box modernization is a high-anxiety problem because the cost of false confidence is high. That is why Publicis Sapient’s model keeps humans in control.

Sapient Slingshot uses AI agents, prompt libraries, enterprise context and specialized modernization workflows to accelerate discovery and transformation. But generated outputs are visible, reviewable and traceable. Engineers inspect specifications, validate recovered logic, assess edge cases and guide modernization decisions. Product and business stakeholders can engage earlier because the system intent is expressed in clearer artifacts rather than buried in technical complexity.

This human-in-the-loop model matters. It turns AI from an opaque generator into a governed engineering capability. Teams do not have to accept mystery outputs on faith. They can review what was extracted, test what was inferred and confirm that critical behavior is preserved before moving downstream.

In other words, AI helps surface what the organization can no longer easily explain about its own systems, while experienced engineers ensure that explanation is good enough to trust.

Dependency tracing before disruption

One reason legacy programs derail is that dependencies are discovered too late. A seemingly simple change reveals a hidden integration. A report depends on logic no one documented. A workflow touches adjacent systems in ways nobody mapped. By the time teams realize the true impact, timelines slip and confidence drops.

Sapient Slingshot helps address this by carrying enterprise context forward across the lifecycle. Its enterprise context graph connects repositories, specifications, journeys, data and dependencies into a living system of understanding. That continuity helps teams move from fragmented project memory to a more persistent, explainable view of how the application estate works.

For engineering leaders, this changes the posture of modernization. Instead of planning around partial knowledge and hoping testing catches the rest, teams can begin with a more structured picture of the current state. That improves sequencing, reduces rework and strengthens the business case for modernization.

From recovered logic to modern code

Once legacy behavior has been translated into specifications, modernization becomes far more manageable. Specifications can drive design, code generation, testing and deployment with traceability carried forward. That reduces the risk common in rewrite and replatform efforts, where undocumented logic is often lost in translation.

Publicis Sapient uses this approach across multiple modernization archetypes, including mainframe, back-end, front-end and tightly coupled enterprise platforms. In COBOL modernization scenarios, Slingshot can break down large codebases, map dependency trees, generate detailed business and functional specifications, and then convert those verified requirements into modern cloud-native architectures such as Java microservices. Security and testing are built into the workflow so quality, compliance and deployment readiness are part of the transformation process rather than an afterthought.

Proven in high-risk legacy scenarios

This approach is especially valuable when the starting point appears least recoverable. Publicis Sapient used Slingshot, paired with human oversight, to help RWE revive a 24-year-old application with no source code or documentation in two days. In another large-scale healthcare modernization effort, Slingshot helped uncover hidden business rules and dependencies across more than 10,000 COBOL and Synon mainframe screens, supporting faster and safer migration.

These examples point to the same principle: legacy systems do not have to be fully understood manually before transformation can begin, but they do need to become explainable before change can be trusted.

Modernize what you cannot easily explain

For many enterprises, the biggest modernization blocker is not technical debt alone. It is the fear of changing systems nobody fully understands anymore.

Publicis Sapient helps reduce that fear by combining Sapient Slingshot’s AI-powered discovery, code-to-spec and modernization capabilities with the judgment of experienced engineers. Together, they create a path from opaque legacy behavior to explicit specifications, from hidden dependencies to visible architecture, and from modernization anxiety to delivery confidence.

That is the real opportunity in AI-assisted modernization. Not blind acceleration. Not code generation for its own sake. But the ability to recover the logic, intent and operational knowledge your business still depends on—so transformation can move forward with speed, traceability and control.

When legacy systems have become a black box, the first step is not rewriting them. It is making them understandable again.