Module 12 · Tier 8 — Transformational Leadership

AI Partnership & Procurement

How to evaluate AI vendors, structure partnerships, and avoid the procurement decisions that constrain your organisation for years.

12 min Tier 8 — Transformational Leadership

Why This Matters

AI procurement is unusual compared to other enterprise technology procurement because the market is developing so fast that today's leading solution may be significantly behind in 18 months, because the switching costs for deeply integrated AI systems can be very high, and because the gap between what vendors claim and what actually works in your specific context can be large. The organisations navigating AI procurement well share a specific discipline: they buy for current need with an explicit plan for flexibility, and they never let vendor enthusiasm substitute for their own evaluation.

AI procurement decisions fall into three categories, each with different evaluation criteria:

Foundation model access: Licensing access to large language models or other foundation models (OpenAI, Anthropic, Google, Mistral, and others). Evaluation criteria: capability on your specific tasks, pricing model, data handling and privacy terms, reliability, and vendor stability. Lock-in risk: moderate — APIs can usually be swapped, though prompt engineering and integrations add friction.

AI applications: Purpose-built AI software for specific functions (AI-powered CRM, legal AI, HR AI, coding assistants). Evaluation criteria: fit for your specific workflow, integration with existing systems, vendor track record, customisation depth, total cost of ownership. Lock-in risk: high — deep workflow integration makes switching expensive.

Full access is for AIQ members

Unlock all 56 lessons, the certificate pathway, and the SociA|~ community.

56 lessons across all three course tracks
AIQ certification on completion
SociA|~ Society community access

Unlock full access →

The Concept

The Procurement Landscape

AI procurement decisions fall into three categories, each with different evaluation criteria:

AI consulting and implementation: External expertise for strategy, build, or change management. Evaluation criteria: track record on comparable organisations, methodology transparency, knowledge transfer versus dependency, pricing structure. Lock-in risk: varies — proprietary methodologies and knowledge concentration create dependency that a good contract can mitigate but not eliminate.

The Evaluation Framework

Before any significant AI procurement decision, answer:

Does this actually work for our specific use case? Not in a vendor demo environment with prepared data, but with your actual data and workflows. Require a proof-of-concept with real conditions before commitment.
What are the total costs? License fees are usually the minority of total cost. Integration, training, change management, maintenance, and the ongoing cost of the capability-gap between what the vendor does and what you actually need are often larger.
What happens if this vendor fails or changes? Vendor consolidation in AI is ongoing. What is your exit strategy if this vendor is acquired, changes pricing significantly, or discontinues the product?
Who owns the data and the models? Specifically: does the vendor use your data to train their models? Does anything you build with this platform belong to you or to them? These contractual terms vary significantly and can be negotiated.
What does this commit us to beyond the immediate need? Some AI systems, once integrated, significantly shape what becomes possible downstream. Evaluate not just the immediate value but the strategic options this decision preserves or closes.

Partnership Structures That Work

The most effective AI partnerships share three characteristics:

Explicit knowledge transfer: The contract specifies what capability your team will have at the end of the engagement, not just what the vendor will deliver. Engagements that don't include knowledge transfer create perpetual dependency.

Performance measurement: Agreed metrics before the engagement starts, with consequences if they are not achieved. Vendors who resist performance metrics are usually telling you something about their confidence in their own claims.

Flexibility provisions: Structured off-ramps at agreed milestones. The ability to exit or reshape a partnership at 6-month intervals costs something in negotiation but significantly reduces the risk of being locked into an approach that isn't working.

The vendor evaluation protocol

A structured evaluation for an AI application procurement decision:

Week 1-2: Shortlist and RFI. Issue a request for information to 4-6 vendors. Ask specifically: their data handling terms, reference clients in your sector, pricing model, and how they handle model updates (do updates change behaviour without notice?)

Week 3-4: Demos with your own data. Require all shortlisted vendors to demo using a sample of your actual data and a representative workflow. Vendor-curated demos reveal capability; your-data demos reveal fit. The gap between these two is often large.

Week 5-6: Reference checks. Speak with three current customers at comparable organisations. Ask specifically: what did not work as promised, what was the implementation experience like, and would they make the same decision again?

Week 7-8: Contract negotiation. Specifically negotiate: data ownership terms, knowledge transfer obligations, performance metrics with consequences, and a 6-month exit provision at fair cost.

This process adds 6-8 weeks to procurement timelines. It consistently prevents expensive multi-year mistakes.

Hands-On Exercise

Evaluate a current or upcoming AI procurement decision

ClaudeAny AI assistant

Apply the evaluation framework to an AI procurement decision you are currently facing or have recently made: 1. Run the five evaluation questions against the decision. Which questions were (or were not) answered before commitment? 2. For an active vendor relationship: what are your current contractual terms on data ownership and model training? If you don't know, find out. 3. What is your exit strategy for your two most significant AI vendor relationships? If there isn't one, what would it take to create one? 4. If you are about to make a new AI procurement decision: design the proof-of-concept that tests this vendor with your actual data before commitment.

Data ownership terms are almost always negotiable, rarely reviewed, and occasionally significant. If you don't know what your current contracts say about your data, the answer is probably not what you'd want it to be.

Active Recall

Before moving on — close this lesson and answer these from memory. Then come back and check. Testing yourself (not re-reading) is how this sticks.

1 What are the three categories of AI procurement and their respective lock-in risks? Which category carries the highest long-term switching cost?

2 What are the five evaluation questions for any significant AI procurement decision? Which is most frequently skipped, and what are the consequences?

Reflection

For your most significant current AI vendor relationship: do you have explicit knowledge transfer terms, performance metrics, and exit provisions? If not, what would it take to renegotiate or restructure the relationship?

Key Takeaway

AI procurement is distinctive because market pace makes today's leader potentially tomorrow's laggard, integration depth makes switching expensive, and vendor claims often don't survive contact with real organisational data. The disciplines that prevent expensive mistakes: proof-of-concept with your own data, explicit knowledge transfer terms, performance metrics with consequences, and structured exit provisions.