How do you think AI will ultimately reshape legal fees and the billable hour? How should it?

Law firms will do what is in their best interests from a revenue perspective. For at least twenty years or more, there has been talk of "value billing" within legal circles, yet the billable hour persists in many (most?) cases. If AI achieves the breathless predictions being made about transforming how legal work is done, I expect more firms to switch to value, rather than hourly, billing models because it finally makes sense from a revenue perspective. That will open the unresolved (read: difficult) questions around how you actually determine value to the client, because that is a two-way discussion where there should be some meeting of the minds between lawyer and client about a subjective valuation. In my mind, that is why the default continues to be hourly billing - that is an objective, one-way telling by the lawyer to the client as to the value of the performed work.

Clients are equally to blame in this dance because we generally have a poor ability to assess the value of a lawyer's work to our organization. This is because the work lawyers do is often defensive or preventative in nature, or it is reactive to some undesired event that occurred.

We’ve been having this same discussion here for a few months.
We’ve identified a need to strengthen our Billing Guidelines around the use of AI by our law firms including what is and is not acceptable.
Our expectation is that if the firm uses AI to reduce time spent on routine tasks, such as drafting Complaints, filings, etc. and communications, that they are (1) responsible for ensuring accuracy of those deliverables, (2) reducing the amount of time billed on those tasks by the partner, associate or paralegal, and (3) firms are not permitted to submit time (or costs) associated with the AI’s “work”.
We’ve seen firms begin to roll out the use of AI in the e-discovery managed reviews which has reduced those fees, but this is solely reliant on the firm’s truthfulness around how much time was saved by using the AI instead of human reviewers.