GenAI vendor advisory

Enterprise GenAI procurement is roughly where enterprise SaaS was in 2014. Pricing is opaque, contracts are vendor-favorable, governance is informal, and the spend is growing 80 to 150 percent year over year at most enterprises. Most CIOs are signing AI commitments with structural risk they would not accept on any other software category.

What we cover

The GenAI vendor practice covers the seven enterprise AI relationships that matter today: OpenAI (direct and via Microsoft Azure), Anthropic (direct and via AWS Bedrock or Google Cloud), Google (Gemini, Workspace AI), Microsoft Copilot (M365 and Studio), AWS Bedrock as a multi-model platform, and the broader GenAI ecosystem of GitHub Copilot, Glean, Writer, and Jasper.

Where the negotiation sits

Most enterprises sign their first AI contracts at the consumption tier the vendor proposes, with whatever data terms the vendor's standard contract carries. That works at small scale. It stops working at $1M annual spend. The negotiation we run typically delivers four things:

• Right-sized commitment. AI consumption is volatile. The contract should have flex, ramp, and underspend protection rather than flat commit.
• Data and IP terms. Training data carve-outs, model output ownership, and the data-residency clauses that matter for regulated industries.
• Price protection. AI list prices have moved significantly. Some up, some down. Contracts need uplift caps and downgrade protection.
• Exit and portability. The ability to migrate to a different model provider without losing prompts, fine-tunes, or operational continuity.

The vendors

OpenAI direct contracts are now substantially more enterprise-grade than the 2023 versions, but data terms still need negotiation. The pricing tiers between Plus, Team, Enterprise, and the underlying API are not always rationalized against actual usage.

Anthropic enterprise contracts are tighter on data terms by default and looser on commercial flexibility. The negotiation usually focuses on commit structure rather than IP terms.

Google AI is bundled in complex ways with Workspace and GCP. The bundle math is rarely intuitive and frequently produces over-spend.

Microsoft Copilot is the most expensive AI commitment most enterprises will sign. The 100 percent attach play at EA renewal is the largest 2026 commercial pressure point. Decoupling Copilot from EA and pilot-sizing it is the single highest-impact negotiation move in the Microsoft world.

AWS Bedrock bundles multiple models behind a unified API and consumption model. Useful as a platform play. Pricing depends heavily on which models you consume and how that interacts with your EDP commit.

How engagements work

Three formats. Pre-deal advisory, four to eight weeks before signing, scopes the contract structure and pricing benchmarks. Renewal advisory, six months before contract expiry, runs the full negotiation. Portfolio advisory, ongoing, covers governance across multiple AI vendor relationships. The portfolio model is increasingly popular as enterprises move from one or two AI vendors to five or seven.

Read the Enterprise AI Licensing Guide 2026 for the full landscape view, the AI Platform Contract Playbook for the negotiation framework, or request a confidential GenAI review if you have a specific contract on the table.