Artificial intelligence is already reshaping how deal teams move through diligence, but the most dangerous moment in that shift isn't a bad AI output — it's reaching for the capability before the governance framework is in place. This episode of VDR.ai cuts through the enthusiasm and the anxiety to lay out a practical, sequenced framework: four questions every deal team must answer before a single document gets near an AI tool.

The episode walks through each question in the order it needs to be resolved, explaining why the sequence itself is part of the discipline:

• Where does the data go? Consumer-grade AI tools routinely retain inputs for model training — a practice that is incompatible with confidential deal material. The episode explains why a zero-data-retention AI commitment needs to be in the contract, not the FAQ, and why anything less puts NDAs and fiduciary duties at risk.
• Where does the model actually run? There is a meaningful distinction between a contractual assurance and a technical one. For deals touching regulated industries, government contracts, or cross-border data residency rules, sovereign or on-premise deployment may be the only permissible option — and that determination belongs at deal setup, not after ingestion.
• Who can see what the AI produces? Clean-team walls that hold at the document level can collapse at the output layer if AI summaries surface insights to the wrong eyes. The episode makes the case that granular permissions must extend to AI-generated outputs, not just the underlying documents — and that tools incapable of supporting that role-based logic are architecturally incompatible with clean-team deals.
• How do you cite the AI's work? When a finding flows into an IC memo or a purchase agreement representation, "the AI said so" is not a defensible answer. The episode frames AI-assisted review as a navigation layer: the model directs attention, a human confirms and cites the primary source, and the audit trail stays intact.
• The teams that win aren't the ones who automate the most — they're the ones who answered these four questions first and built their workflows inside those guardrails.

Practitioners looking to go deeper on AI-assisted review can explore agentic due diligence on the VDR.ai platform, and the M&A due diligence guide offers a broader reference for structuring the full diligence process. For a close look at why the records generated during a deal matter as much as the deal itself, the companion episode Audit Trails in a Data Room: The Record You Build While You Work is the natural next listen.

VDR.ai