What an AI actually finds in a contract
Tuned on your own documents, the model reads an agreement in seconds and highlights the places where companies usually get caught:
- •Unfavourable terms: unilateral termination, penalties and liquidated damages, choice of forum, automatic renewal,
- •Imbalance of liability — where your obligations are harsher than the other side’s,
- •Vague wording around deadlines, payment schedule and acceptance,
- •Differences between the new draft and your version — exactly what the counterparty changed,
- •Missing clauses that this type of agreement normally contains.
What it does beyond reading the text
| Task | How it helps |
|---|---|
| Redline comparison | shows what changed between versions, so nobody diffs by hand |
| Counterparty check | registry status, signs of liquidation or insolvency, disqualified directors |
| Search across the contract archive | “where do we have the 0.1% penalty clause?” — answered with a link to the document |
| Statement of disagreements | a draft built from your own templates on the basis of the risks found |
What an AI cannot do
The honest limits matter as much as the capabilities:
- •Make the decision or carry the responsibility — that stays with the lawyer,
- •Handle bespoke, multi-layered deals with the same confidence as standard agreements,
- •Know your unwritten practice if it was never loaded into the system,
- •Guarantee complete coverage — which is why a person checks the output and why deployment starts with a run across your archive.
How it gets deployed
The AI Lawyer is tuned to your sector: your templates, your internal rules and your positions on contested clauses go in, and from then on it checks documents by your own standards. To keep trade secrets and personal data safe, processing runs on your server or on our site — see the AI Lawyer page, or the on-premise AI setup for law firms.
This is an overview, not legal advice — the final legal assessment comes from a qualified lawyer.
Frequently asked questions
Will an AI replace our lawyers?
No, and it should not. It takes the routine — the first pass, redline comparison, searching the archive, drafting. The decision and the responsibility stay with a human lawyer, who now spends the time on the contested points instead of proofreading every clause.
How accurately does it find risks?
On standard agreements — supply, services, lease, works — accuracy is high once the system has been tuned on your own documents and your own negotiating position. Before go-live you run it across your contract archive and check the output against a lawyer.
Do our contracts leave the building?
Not when the model is installed on your server: processing stays inside the perimeter. If a top-tier cloud model is genuinely needed for a complex analysis, the text is de-identified first — names, amounts and registration details are stripped and double-checked before anything goes out.