When Technology Due Diligence Comes Too Late: A Cautionary Tale

The conversation about technology due diligence usually starts too late.

Not because investors don’t understand its value — most do, in principle. It starts too late because deal timelines compress, because management teams are persuasive, because the commercial thesis feels solid and the technical risk feels like detail. And so the TDD gets commissioned after heads of terms are signed, or scoped down to fit a three-week window, or skipped entirely on the basis that the platform “seems straightforward.”

What follows is rarely a sudden catastrophe. It is almost always a slow accumulation of costs that were knowable before close — and weren’t.


The pattern that plays out most often

The most common version of this story does not involve fraud or deliberate concealment. It involves a management team that genuinely believes its platform is in better shape than it is, and a buyer who did not commission the independent assessment that would have produced a different view.

The technical debt that “we’ve been planning to address” has been on the backlog for three years. The key engineer who “definitely plans to stay” leaves within six months of close, taking institutional knowledge that was never documented. The security certification the company holds was last assessed eighteen months ago, and the platform has been substantially rebuilt since. The AI capability that featured prominently in the investment thesis turns out to be a wrapper around a third-party model with no proprietary data, no fine-tuning, and terms of service that restrict the commercial use the buyer had assumed.

None of these are unusual findings. All of them are findings that a competent TDD, commissioned at the right time, would have surfaced.


Three scenarios where late due diligence consistently fails buyers

When TDD is scoped to fit the timeline rather than the risk. A compressed timeline does not reduce the risk in the asset — it reduces the visibility into it. A TDD scoped to two weeks when the asset warrants five will produce findings on the dimensions that are fastest to assess, and silence on the ones that take longer. Security assessments that should include penetration testing get replaced by documentation reviews. Codebase analysis that should cover the full repository gets sampled. The report looks complete. The coverage isn’t.

The correct response to a compressed timeline is not a compressed TDD. It is a risk-prioritised assessment that is explicit about what was and was not covered — so that the buyer knows where the residual uncertainty sits and can structure accordingly.

When the commercial team has already committed before the technical team has assessed. This is the most structurally difficult version of the problem. Heads of terms have been agreed, a price has been anchored in both parties’ minds, and the TDD is now expected to confirm the thesis rather than test it. The findings that emerge — technical debt that will cost £200k to remediate, a security posture that requires significant investment before enterprise customers will pass their vendor assessments, an AI capability that is far thinner than represented — arrive at a moment when acting on them feels like renegotiating in bad faith.

The findings are real. The window to use them has substantially closed. Price adjustments are harder to negotiate after commercial terms are agreed. Escrow provisions that would have been standard pre-HoT become contentious. Post-close conditions that should be contractual become informal commitments that may or may not be honoured.

When AI and security risks are treated as secondary scope. In the current market, a significant proportion of acquisition targets include AI-dependent features as a material part of their value proposition. Security posture has become a gating issue for enterprise customers and a regulatory obligation in many sectors. Both deserve primary scope, not an add-on assessed in the final days of a compressed timeline.

The specific risks are well-documented but still routinely underestimated. Training data collected without appropriate consent transfers its liability to the acquirer. A model that has never been tested for adversarial robustness may behave unpredictably under inputs that a motivated adversary can craft in hours. EU AI Act compliance obligations for high-risk AI systems do not pause for a change of ownership. A security gap that the seller managed through obscurity becomes a known attack surface the moment the acquisition is public.


What the timing of due diligence actually determines

The timing of technology due diligence does not determine what problems exist in the asset. It determines whether those problems are discovered before close, when they can inform deal structure, or after close, when they become the buyer’s problem to fund.

This is a simple but frequently underweighted point. The cost of a TDD commissioned at the right time — before heads of terms, or at least before commercial terms are fully anchored — is fixed and known. The cost of remediating findings that were discoverable but undiscovered is variable and almost always higher than it would have appeared in a pre-close assessment.

The cautionary tale is not one deal. It is a pattern that repeats across markets and deal sizes, with consistent elements: a compressed timeline, a confident seller, a buyer who prioritised speed over certainty, and a post-close period spent funding work that should have been priced into the transaction.

The structural fix is straightforward: commission technology due diligence earlier, scope it to the risk rather than the timeline, and treat AI readiness and security as primary dimensions rather than supplementary ones.

At VeryDiligent, we work with buyers who want to get ahead of this pattern — including in situations where the timeline is genuinely constrained. A risk-prioritised assessment with explicit coverage boundaries is almost always better than a compressed one that appears complete. If you are approaching a transaction, we would be glad to discuss what proportionate looks like for your situation.


Related reading: What PE Firms Should Demand from Technology Due Diligence | Scoring Technical Maturity: A Practical Model | The Rise of Deep Tech and Why Traditional TDD Frameworks Fail