AI Tool Contract Review Checklist

The biggest risk in AI contracts isn't the first-year price. It's what happens at renewal and what happens when usage exceeds your plan.

• Pricing model is clearly defined. Is this per-seat, per-user, credit-based, consumption-based, flat fee, or a hybrid? If it's credit-based, is there a clear conversion rate between credits and actual usage?
• Annual uplift is capped in writing. What's the maximum price increase at renewal? Push for 3–5%. If the vendor won't cap it, request a most-favored-nation (MFN) clause guaranteeing you won't pay more than comparable customers.
• Overage charges are documented and capped. If you exceed your plan (credits, API calls, storage, seats), what happens? Are overages billed automatically, or is there a notification and approval step? Is there a spend cap you can set?
• AI features vs. base product pricing is separated. Is the AI functionality included in your current tier, or is it a separate add-on with its own pricing? If it's included now, can the vendor unbundle it later and charge separately?
• Credit-to-dollar conversion is transparent. For credit-based models: how much does one credit cost? What actions consume credits? Can credit costs change mid-contract? Is there a dashboard to monitor credit usage in real time?
• Pricing model changes require notice. If the vendor transitions from per-seat to consumption pricing (or vice versa), does the contract require advance written notice and your consent? What's the notice period?
• Discount structure is unconditional. If you're receiving a discount, is it tied to purchasing AI add-ons or committing to a longer term? Conditional discounts should be evaluated separately from the base renewal.
• Benchmarking rights are included. Does the contract allow you to benchmark pricing against comparable vendors or industry standards? This is especially important for credit-based models where there are no public price lists to compare against.

AI tools interact with your data differently than traditional SaaS. The contract needs to reflect that.

• Training data usage is addressed. Does the vendor use your data to train, fine-tune, or improve their AI models? Is this opt-in or opt-out? Is the opt-out available at the contract level, the account level, or the individual user level?
• Opt-out is contractually binding. If you opt out of training data usage, is that commitment written into the contract — or just a settings toggle that the vendor can change in a terms-of-service update?
• Output ownership is clearly assigned. Who owns the content, analysis, or recommendations generated by the AI using your data? Does the vendor retain any rights to AI-generated outputs?
• Data handling on cancellation is defined. What happens to your data if you cancel? Is it deleted within a specific timeframe? Can you export it in a standard format? Is there a transition window?
• Data handling on acquisition is defined. If the vendor is acquired, what happens to your data and your contract terms? Do you have the right to terminate if the acquiring entity has different data practices?
• Data residency and geographic restrictions are documented. Where is your data stored and processed? Does it cross borders? If you have regulatory requirements (GDPR, CCPA, SOC 2, HIPAA), are they addressed explicitly?
• AI content indemnification is included. If the AI generates content that infringes on a third party's intellectual property, who's liable? Does the vendor indemnify you for claims arising from AI-generated outputs?
• Sub-processor transparency is addressed. Does the vendor use third-party AI models or infrastructure (e.g., OpenAI, Anthropic, AWS Bedrock)? If so, are those sub-processors named in the contract, and are they subject to the same data protection terms?

AI features are newer, less proven, and more likely to change. The contract should account for that.

• SLA covers AI features specifically. Does the uptime and performance SLA extend to AI-specific functionality, or only to the base platform? If AI features go down, is that covered?
• Service credits are automatic. If the vendor misses SLA targets, do service credits apply automatically — or do you have to request them?
• Audit rights are included. Can you audit or request documentation on how the AI features use your data, what models are being used, and whether your opt-out preferences are being respected?
• Material change notification is required. If the vendor changes the underlying AI model, switches providers, or materially alters the functionality of AI features, are they required to notify you in advance? What's the notice period?
• Feature deprecation protections exist. If the vendor discontinues an AI feature you rely on, what are your remedies? Can you terminate early? Is there a minimum notice period before deprecation?
• Accuracy and reliability standards are addressed. For AI tools where output quality matters (analytics, forecasting, content generation), are there any performance benchmarks or quality commitments in the contract?

AI adoption is unpredictable. Your contract should give you room to adjust.

• Mid-term seat or usage adjustments are allowed. If adoption is lower (or higher) than expected, can you adjust your plan mid-contract? Are there penalties for reducing usage?
• Short-term or pilot options are available. If this is a new AI tool, is there a pilot structure with a shorter commitment and the option to expand? Have you negotiated a walk-away clause if the pilot doesn't meet defined success criteria?
• Cancellation notice period is reasonable. What's the required notice period to cancel or non-renew? Is it 30, 60, or 90 days? Is the notice window clearly documented and not buried in an appendix?
• Auto-renewal terms are acceptable. Does the contract auto-renew? If so, what's the notice window to opt out? Is the auto-renewal at the current rate, or at a new rate the vendor sets?
• Data export and transition provisions exist. If you leave, can you export all your data in a standard format (CSV, JSON, API)? Is there a transition window (30–90 days) after cancellation where you still have access?
• Termination for convenience is available. Can you terminate the contract early without cause? If so, what's the penalty (if any)? This is especially important for first-time AI tool purchases where long-term fit is uncertain.