AI contract review software, hosted in Germany

AI contract review that applies your legal team's rules — not a generic model's opinion.

Treasure Contracts checks every contract against your playbook and shows a traffic light per criterion, with a compliant wording suggestion for each finding.

Sales, procurement and legal get the same answer to the same question. Every review is repeatable, runs again on each new version and is stored with its rationale in the audit trail.

Review my own contract live See the Proof Sprint

ISO 27001 · Hosted in Frankfurt · GDPR · 4-week right of withdrawal

In short

How does AI contract review work?

AI contract review reads a contract, compares each clause with defined criteria and flags risks. In Treasure Contracts, those criteria come from your own playbook: targets, no-gos and your party perspective. Each criterion gets a traffic light — compliant, moderate or critical — with a compliant wording suggestion. The review also flags general risks, internal contradictions and deviations from your reference template.

Key facts

Result per playbook criterion as a traffic light: compliant, moderate or critical
Review reruns automatically on every new version, so counterparty edits are checked against your playbook again
Platform on AWS in Frankfurt, AI processing in EU regions, no training on customer data

Updated: 26 September 2026 · By Tobias Warken, Moritz Biersack and Dr. Annalisa Steiner (founding team, Treasure Contracts)

The problem

Contract review rarely fails on legal knowledge. It fails on capacity and consistency.

The same NDA, the same supplier terms, the same customer paper — reviewed again and again by the same few people, with slightly different results each time.

Every review starts from scratch

Even for a standard NDA, someone reads all ten pages, looks up the fallback position and writes the same comments as last week.

Results depend on who reviews

One colleague accepts a liability cap, another escalates it. Without written rules, the business gets inconsistent answers and loses trust in the process.

Counterparty edits slip through

The second or third draft comes back with a small change in the indemnity clause. Nobody compares it line by line, and the risk ends up in the signed version.

Chatbot reviews are hard to trace

Someone pastes a contract into a public AI chat. The answer sounds plausible, but the standard depends on the prompt, there is no per-contract audit trail like in a review system, and data handling depends on the provider's terms.

Your outcome

Standard contracts pre-reviewed against your rules. Legal decides only the real exceptions.

The point of AI contract review is not to replace legal judgement. It is to apply that judgement consistently, at scale, without legal reading every page.

Your outcome

An incoming NDA or supplier contract is pre-reviewed against your playbook, with a rationale and wording suggestion per criterion, before it reaches a lawyer.

Why it works

The rules come from your legal team, not from a generic model. Playbooks can be tested with golden-set tests on synthetic documents before going live.

How fast

You see the first review against your criteria in a 30-minute demo on your own contract. In the Proof Sprint, two contract types run live within ten working days.

How little effort

We derive the playbook with you from 10–20 historical contracts. Your team then uploads contracts or simply forwards them to a dedicated email address.

Features

What AI contract review in Treasure Contracts covers

All features listed here are live in the product today.

Playbook check with traffic lights

Criteria, targets, no-gos, party perspective and reference documents define the review. Each criterion is rated compliant, moderate or critical, with a short rationale.

Risks, contradictions and deviations

Beyond your criteria, the AI flags general risks, contradictions within the document and deviations from your reference template or standard clauses.

Compliant wording suggestions

For every finding, Treasure proposes wording that meets your playbook. Reviewers can add their own risks and give feedback on each result — human in the loop.

Re-review on every version

Each version is stored immutably and compared side by side as a redline. The review reruns automatically, so changes from the other side are checked again.

Review by email

Forward a contract to a plus address such as an NDA review inbox, and it goes straight to the right playbook. Sender whitelists and DMARC/SPF checks protect the inbox.

HubSpot and Slack

Quote webhooks from HubSpot trigger a review and report the result back to the deal. Slack notifications tell the team when a review needs attention.

Playbook with review options and criteria

How it works

How AI contract review gets into your daily work

Demo on your own contract
Bring a real contract. In 30 minutes with the founders, we show how Treasure reviews it against your criteria.
Build the playbook
From your guidelines and 10–20 historical contracts, we derive criteria, targets and no-gos. Legal confirms every rule.
Test before go-live
Golden-set tests with synthetic documents show whether the playbook catches what it should, before real contracts run through it.
Live review in the Proof Sprint
Two contract types run live with risk rating, redlines and audit trail, followed by acceptance by your legal team and an ROI evaluation.

Comparison

Manual review, a general AI chat or Treasure Contracts?

Three ways to review a contract — and what they mean for consistency, speed and control.

Criterion	Manual review	General AI chat	Treasure Contracts
Standard applied	Reviewer's experience	Generic model, depends on the prompt	Your playbook, confirmed by legal
Result format	Comments in the document	Free text	Traffic light per criterion with wording suggestion
New versions	Compared manually	Pasted again; the standard depends on the prompt	Automatic re-review and redline comparison
Documentation	Email threads	Chat history; no per-contract audit trail like in Treasure	Audit trail with version and rationale
Data location	Your systems	Depends on the provider and plan	AWS Frankfurt, AI in EU regions, no training

Data handling of general AI assistants depends on provider, plan and settings.

In detail

AI contract review: what separates a useful tool from a demo

The playbook is the product, the model is the engine

Any modern language model can summarise a contract. What it does not know without your input is which liability cap your company accepts, which governing law you insist on, or when a confidentiality term becomes a deal-breaker. That knowledge lives in your legal team. Treasure Contracts makes it executable: criteria, targets, no-gos and the perspective of your party are stored in a playbook, supplemented by reference documents. Every AI contract review follows that playbook, so results are consistent and explainable rather than dependent on phrasing.

Traffic lights make decisions possible for non-lawyers

A long list of AI comments does not help a sales manager decide whether to escalate. A traffic light per criterion does. Compliant criteria need no action, moderate ones point to a suggested wording, critical ones go to legal. Because each finding comes with a rationale and a compliant alternative, the business can resolve routine points itself, and legal sees only the cases that genuinely need judgement. The dashboard shows open reviews and critical findings at a glance.

Why re-reviewing every version matters

Most contract risk does not sit in the first draft you send. It sits in the redline that comes back. Treasure stores every version immutably, compares two versions side by side and reruns the review automatically on each new one. A quiet change to an indemnity clause is therefore checked against the playbook again, just like the first draft, and the audit trail shows which version was reviewed, with which result and which feedback from your reviewers.

Measuring quality before trusting it

Legal teams are right to be sceptical of AI output. Treasure addresses that with golden-set tests: synthetic documents with known issues are run against a playbook to check whether it flags what it should. Reviewers can add missing risks and give feedback on each result, which keeps a human in the loop. Treasure does not promise legally flawless results; it promises a documented, repeatable first review that your team controls.

Your first step

Prove it first. Decide second.

You don't have to take our word for it. You see Treasure on your own contracts — first in a free 30-minute session, then in a 10-day Proof Sprint with measurable results.

For AI contract review, that means: we take your two most frequent contract types, such as NDAs and supplier agreements, turn your review rules into a playbook and show on real contracts how the traffic lights compare with your lawyers' judgement.

Start free: your contract, reviewed live

30 minutes with the founding team. Bring a real contract and we show Treasure live on your clauses. No slides, 24-hour response guarantee — including a free AI Readiness Assessment.

Pick a time

What you get in the 10-day Proof Sprint

Baseline for cycle time, volume and legal effort
Testable rules derived from your existing contracts and decisions
Live test on 10–20 historical contracts with risk rating, redlines and audit trail
Sign-off of the rules and jointly defined success criteria
ROI analysis and a concrete 30-day plan for your first live process

€4,900 net, one-off · fully credited towards implementation if you sign within 30 days

Results guarantee: If Treasure misses the acceptance criteria agreed in writing up front, we keep working for up to 20 additional business days at no cost. If that is still not enough, we refund the sprint fee. This requires your timely cooperation; details are in the sprint offer.

Right of withdrawal: 4 weeks unconditional right of withdrawal with a full refund.

We run at most 3 Proof Sprints per month — we confirm the next available start date in the first call.

See pricing & Proof Sprint

FAQ

How accurate is AI contract review?

Accuracy depends above all on the quality of the rules. Treasure reviews against a playbook your legal team has confirmed, and playbooks can be tested with golden-set tests on synthetic documents before going live. Reviewers give feedback on each result and can add risks. Treasure delivers a documented first review; the final decision stays with your team.

Which contracts can AI contract review handle?

Treasure processes PDF, Word and email files including attachments, with OCR for scanned documents. Common starting points are NDAs, supplier and customer agreements, SaaS contracts and data processing agreements. You can define your own contract types and a dedicated playbook for each.

Why not just use ChatGPT or Claude for contract review?

A general AI chat gives an answer that depends on the prompt, without a fixed rule set confirmed by legal and without a per-contract, per-version audit trail like Treasure's. Treasure applies the same legal-approved rules to every contract, reruns the review on each new version and stores everything in the audit trail. With the Treasure MCP server (AI tool integration, €50/month, included from Professional), you can still trigger reviews from ChatGPT or Claude.

Is our contract data safe and GDPR-compliant?

The platform runs on AWS in Frankfurt, AI processing happens in EU regions, and customer data is never used to train models. Tenants are separated at database schema level. Two-factor authentication, Microsoft Entra ID SSO and a CSV-exportable audit trail are available, and information security is ISO 27001 certified.

What does AI contract review with Treasure cost?

Plans start at €590 per month (Starter), followed by Professional at €1,790 and Business at €3,590, each billed annually and net; Enterprise is on request. The 10-day Proof Sprint costs €4,900 and is fully credited towards implementation if you book within 30 days.

How much work is it to set up a playbook?

Less than writing one from scratch. In the Proof Sprint, we derive the rules with you from 10–20 historical contracts and your existing guidelines, and legal confirms each one. If the acceptance criteria agreed in writing beforehand are not met, we work up to 20 more working days at no cost and otherwise refund the sprint price; your cooperation is required.

Bring your next contract.

In 30 minutes you will know whether Treasure can apply your rules — on a real contract, with the founders.