Service 06 · AI & SaaS Procurement

AI contracts are not SaaS contracts.

Microsoft Copilot, OpenAI Enterprise, Anthropic, Google Gemini, AWS Bedrock — AI procurement introduces usage-based pricing, IP indemnification gaps, training data clauses, and model deprecation risk that procurement teams aren't yet calibrated to. We sit on the buyer side and harden the terms.

Contact Us →Download Playbooks
340+
Procurement engagements
$1.8B+
Documented client savings
11
Vendor practices
Gartner
Recognised practice
What we negotiate

AI clauses under scrutiny.

01
Usage-based pricing structure
AI pricing is rarely flat seats. Token-based, query-based, capacity-based, mixed-mode — each pricing model carries forecast risk and overage exposure. We model usage trajectories and structure the commitment to protect against AI workload volatility.
02
IP and indemnification
Who owns outputs? Who indemnifies for training-data infringement claims? How does indemnification scale with usage? Microsoft's Copilot Copyright Commitment, OpenAI's IP indemnification, Anthropic's customer protections — each is different. None covers everything.
03
Data residency and training rights
Where does input data go? Does the vendor train on customer data? What logs, what telemetry, what derived data? We negotiate residency, training opt-outs, and data minimisation clauses explicitly.
04
Model deprecation and substitution
The model version you contracted for may be deprecated in twelve months. We negotiate model-substitution rights, performance benchmarks for replacement models, and price-hold clauses across model generations.
05
Security and audit
SOC 2, ISO 27001, FedRAMP, HIPAA, GDPR — AI vendors are still building certification surface area. We negotiate audit rights, sub-processor visibility, breach-notification SLAs, and architectural review hours.
06
Exit and portability
AI lock-in is real — fine-tuned models, retrieval indexes, prompt libraries, embedded workflows. We negotiate data and artefact portability rights, model-export commitments, and reasonable transition assistance.
Why AI procurement is different

The procurement playbook doesn't yet exist.

01
Pricing models are unstable
Token pricing, agentic action pricing, capacity reservations, mixed-mode bundles — vendors are still experimenting. The contract terms you accept today set anchor points for the next renewal, regardless of how the pricing evolves.
02
Indemnification gaps are real
Most enterprise AI contracts indemnify the customer against IP infringement only under narrow conditions: customer-supplied data, no fine-tuning, specific model versions, capped at fees. Read the carve-outs before signing.
03
Sub-processor sprawl
An AI vendor's stack often layers model providers, infrastructure providers, evaluation tooling, and retrieval services. Each is a sub-processor. Procurement frameworks need to flow down obligations across the full chain.
04
Regulatory drift
EU AI Act, US state-level frameworks, sector-specific guidance — the regulatory floor is rising. We negotiate forward-looking compliance commitments and amendment rights when regulations change mid-term.
Inline · talk to a senior advisor

Not sure where to start?

Contact Us →See outcomes
Research before you procure

AI and SaaS briefings.

All papers
Briefing
AI Vendor Contract Red Flags
Usage pricing, IP, data rights, indemnity
Guide
Microsoft Copilot Enterprise Guide
E5 prerequisites, Copilot licensing economics, rollout strategy
Guide
SaaS Spend Optimization Guide
Rationalising the long tail of SaaS subscriptions
Guide
CIO Contract Governance Guide
Portfolio-wide contract policy and renewals discipline
Recent outcomes

Where our work paid for itself.

All cases
Financial services office
Salesforce · Financial
$3.2M
Salesforce Renewal
Einstein and Data Cloud right-sized, price-hold secured.
Read case study
Insurance office
ServiceNow · Insurance
$2.1M
ServiceNow Optimization
ITSM and Now Assist scoped without over-buying.
Read case study
Corporate office
Microsoft · Telecom
$8.7M
Microsoft EA Optimization
Copilot pilot rolled in without expanding E5 footprint.
Read case study
Questions

Frequently asked, frankly answered.

Q1
Do you cover AI platforms beyond Microsoft?
Yes. We've advised on Microsoft Copilot, OpenAI Enterprise, Anthropic, Google Gemini and Vertex, AWS Bedrock, ServiceNow Now Assist, Salesforce Einstein, and a range of vertical AI vendors.
Q2
Can you advise on procurement framework design?
Yes. Several CIO clients engage us to design the procurement framework before they begin signing AI contracts — risk taxonomy, contract templates, escalation triggers, governance policies.
Q3
What about open-source LLM deployments?
Self-hosted Llama, Mistral, and similar deployments shift the contract surface to infrastructure, hardware, support, and indemnification on weights. We advise on the procurement framing of self-hosted strategies.
Q4
How do you handle usage forecasting?
We model forecast bands, peak-load assumptions, growth scenarios, and adoption rate. The commitment should be sized to a credible adoption curve, not to the vendor's growth assumptions.
Q5
Does this overlap with cloud advisory?
Yes — significantly. AI workloads sit on hyperscaler infrastructure, and AI commitments often interact with EDP, MACC, or CUD obligations. We coordinate the two engagements when both are in scope.
Q6
How do you bill?
Fixed-fee, scoped by deal size and complexity. No contingency, no percentage of savings — see /pricing/.

AI contract on the table?
Read the indemnification first.

Token pricing, IP gaps, training data clauses, model deprecation. We harden every line before you sign.

Contact Us →Download Playbooks

The Compliance Brief

Weekly compliance intelligence for IT leaders.