Enterprise software pricing is shifting from per-seat models towards usage- and outcome-based approaches.[1]
This shift is being accelerated by AI features that are designed to increase efficiency: if AI reduces the need for human involvement, per-seat pricing can become misaligned with a service’s value proposition. At the same time, vendors face a trade-off. Moving away from per-seat pricing can mean sacrificing predictable recurring revenue patterns in favour of pricing models that are harder to forecast.
Usage-based pricing also introduces new alignment. If a customer uses the service more and obtains more value from it, the vendor earns more fees. But it also introduces new risks, particularly for customers who are accustomed to stable, seat-based bills. Sudden spikes in usage can create “price shock” moments, making it difficult to forecast how much an AI service will cost over its committed term.
Alongside the pricing-model transition is volatility in the underlying cost of AI processing (inference).[2] If the costs of generating model outputs are falling quickly, customers making long-term pricing commitments will often want to ensure they benefit from that downward cost curve rather than locking in static rates that ignore improvements in efficiency and compute economics over the medium term.
Contracts need to evolve accordingly. The key drafting challenge is to make usage measurable, controllable, and auditable. That typically means:
- Defining the pricing unit clearly (for example: tokens consumed, tickets resolved, qualified leads generated, images produced).
- Choosing a structure (pure consumption, tiered bundles, or hybrid models such as a fixed platform fee plus usage-based AI overages).
- Implementing spending controls (hard or soft caps, alerts, throttling, or other mechanisms to prevent runaway costs).
- Addressing cost pass-through (whether and how changes in underlying input costs affect pricing).
- Building measurement and dispute mechanics (logging, auditability, and a process for challenging invoices where usage data is contested).
All this means pricing clauses look set to get more complex in the AI era. Pricing will come to shape adoption behaviour, risk exposure, and the long-term economics of the relationship in new ways. The organisations that manage this change well will be those that treat pricing design and contractual controls as part of the same system. We are looking at this theme and some other implications of the “AI Platform Shift” for enterprise software in our ongoing “AI Contracting Playbook” webinar series. The first episode took place in mid-February 2026. Please get in touch at chris.kemp@kempitlaw.com for a copy of the slides and recording. The next one will be on Wednesday 1 July 2026. You can sign up here.
[1] Some commentators predict that by the end of 2026, hybrid pricing models combining usage- and outcome-based elements will capture a majority of enterprise software revenue. See AlixPartners, 2026 Enterprise software technology predictions report here.
[2] By some measures, the cost of AI inference is plummeting. For example, research from AI research consultancy Epoch AI suggests that the inference costs of high-performing models has fallen around 900 times between 2023-24. See here.
