Haihtuva › Guides
AI due diligence: how it works and what to check
AI due diligence uses language models to read the documents of a transaction and help the investigating party find what matters faster: risks, gaps, inconsistencies and open questions.
What the AI does in Haihtuva
- Outside-in first: the target's website and fixed, credible public sources (for life science, for example Europe PMC and ClinicalTrials.gov, plus the laws and regulations that apply) are compared with what the data room says.
- Document review: every released document is reviewed for risks, each with the page and a short quote it comes from. The newest document on a topic leads; other parties' material is left out of the target's analysis.
- Questions from the data room first: standard due-diligence questions are answered from the documents; only what the data room does not answer goes to the seller, deal-breakers first.
- Reports: findings by area, risks with severity, follow-ups and a due-diligence report, saved to the buyer's own storage if it wishes.
What people still do
AI output can be wrong or incomplete. Every finding cites its source so a person can check it. Decisions, valuation and deal terms stay with people and their advisers; the AI does not recommend whether to invest or at what price.
Confidentiality
The analysis runs in a sealed Azure confidential virtual machine in the EU. The approved model, region, retention and logging settings are written into the room's Security Specification, which both sides approve before any document is read.