Create one commercial inventory
- List every AI tool with a recurring, usage-based or committed-spend component.
- Capture owner, department, contract term, renewal date, committed amount and current monthly run rate.
- Separate experiments from business-critical workloads.
Track the unit that drives cost
- For seat tools, track paid seats and active users.
- For consumption tools, identify the measurable driver: tokens, requests, credits, storage, compute or another unit.
- Compare committed volume with actual usage and expected growth.
Stop duplicate buying before it compounds
- Map overlapping copilots, writing tools, research tools, meeting tools and developer assistants.
- Ask whether a new purchase is additive or simply another interface for capabilities already owned.
- Require a business owner and renewal owner for every material tool.
Negotiate the commercial model
- Review minimum commitments, overage pricing, price protections, renewal escalators, ramp schedules and unused-commitment treatment.
- For fast-growing use cases, model multiple usage scenarios before agreeing to volume.
- Avoid locking a pilot economics assumption into a large annual commitment without evidence.
Related Calder guidance
Bottom lineAI spend control is mostly an ownership, visibility and commercial-discipline problem before it becomes a tooling problem.
Have a material renewal coming up?
Use the framework to define the buyer position before the supplier controls the timeline.
Discuss an engagement