AI-Generated Code in the Codebase: A New Diligence Question

Somewhere between 20 and 50 percent of the code in a typical engineering organization is now written with AI assistance. For a target company acquired today, a meaningful share of the codebase was produced by a developer accepting a suggestion rather than writing a line.

Somewhere between 20 and 50 percent of the code in a typical engineering organization is now written with AI assistance. For a target company acquired today, a meaningful share of the codebase was produced by a developer accepting a suggestion rather than writing a line.

This is not inherently a problem. AI-assisted development has made real teams faster and, in some cases, better. But it changes what diligence needs to look at, and most technical diligence processes have not caught up.

Why It Matters to a Buyer

Code volume stopped being a proxy for team capability. Commit counts, lines shipped, and velocity metrics have historically served as rough indicators of engineering throughput. When a significant portion of output is AI-generated, those numbers measure tool adoption as much as team strength. A team that looks productive on paper may be producing more code than it can actually maintain.

Maintainability risk looks different. AI-generated code tends to work. Whether it fits the existing architecture is a separate question. Common patterns include duplicated logic that reimplements existing utilities, inconsistent error handling, and dependencies introduced without review. Individually minor, these accumulate into the kind of technical debt that is hard to see in a code sample and expensive to unwind.

Comprehension gaps create key-person risk in a new form. Traditional key-person risk means one engineer understands a critical system. The emerging version is that nobody fully understands it. When code was generated, reviewed quickly, and shipped, institutional knowledge may be thinner than headcount suggests. Post-close, that shows up as slow incident response and cautious refactoring.

IP provenance deserves a direct question. Depending on the tools used and how they were configured, there are legitimate questions about training data and code provenance. Most enterprise AI coding tools now offer indemnification and controls, but not all teams have enabled them. For a buyer, the question is whether the target can document which tools were used, under what license terms, and with what policies in place.

What to Ask During Diligence

Ask directly what AI tools the engineering team uses, when adoption started, and whether usage is governed by policy. The answer is informative regardless of what it is. Generally, teams with thoughtful governance tend to have thoughtful engineering practice.

Ask how AI-assisted code is reviewed. The strongest signal is whether review standards changed when AI adoption scaled. Teams that maintained rigorous review have a defensible position. Teams that let review loosen to capture velocity gains have accumulated risk they may not have measured.

Ask about test coverage trends over the adoption period. If code volume grew substantially while coverage stayed flat or declined, that gap is a reasonable proxy for unmanaged risk.

Ask engineers to explain recently shipped features. Not the architecture in the abstract, but specific implementation decisions. Depth of answer reveals whether the team owns its codebase or merely operates it.

The Practical Takeaway

AI-assisted development is now standard practice, and buyers should expect to find it. The diligence question now is if they use the tools with discipline.

That distinction will increasingly separate targets that scale cleanly post-close from those that do not.

 

BTA provides technical due diligence and AI strategy support for private equity, investment banking, and strategic buyer teams. Contact us to learn more.

Confidence in M&A and Strategy leads to effective business results.

BTA partners with investors and operators to assess technology risk, execute AI-driven value creation, and provide hands-on technology leadership through critical stages of growth and M&A. Contact us today to learn more.