Back in May '26 Kirkland & Ellis announced a $500m investment over the next 3-4 years to build a proprietary AI platform based around its own knowledge base and in partnership with Palantir Technologies, and then embed that within various client workflows In August '26, I was struck by Thomson Reuters announcing that they had spent $40m using an open source model applied to Westlaw, Practical Law and so on At 10,000 ft it was the gulf between those numbers that caught my eye,

The way in which firms generally start to formalise central roles and responsibilities around pricing creates a visibility gap, and false sense of confidence around where profit leakage happens For the most part, firms start with a single hire, or small pricing team, to "see how it goes" And how it goes, inevitably, is that the pricing team is immediately constrained by that limited resource. Naturally they therefore prioritise that resource for impact - e.g. setting policy,

The landscape around how AI vendors charge for use is becoming more complex On the basis that most firms using AI likely find themselves: (a) in a transitionary period moving from fixed AI subscriptions to consumption based charges, (b) with a mix of fixed and hourly rate arrangements with their own clients, and are (c) struggling to move away from cost- or competition- based pricing towards value based pricing The matrix below shows a simplied view of what you might consider