Your law firm AI business case is a hypothesis. Keep testing it.

At a recent roundtable with value advisors, we kept coming back to the same point: a business case isn’t a forecast you sign and file. It’s a hypothesis, built with the firm. This is the problem, this is what the solution changes, this is the impact we expect. Then you test it.

That test used to happen slowly: set the case, then review it a year or two later. AI changes quickly enough that a review on that schedule comes too late to correct anything.

In practice, the test rarely happens at all. A 2026 Blickstein Group survey found 66% of law firm COOs say their firm doesn’t formally document efficiency gains from AI. Without that record, there is nothing to test the original case against, and no evidence of law firm AI ROI. 

How to test AI ROI before and after go-live

Built jointly with the firm, the case stops being a static document and becomes a shared plan for reaching an outcome. Before signature, we test the hypothesis against the firm’s own baseline. After go-live, we test it again, with finance and the executive sponsor in the room.  

Where a measure is on target, we know where the value landed. Where it falls short, we agree what to do about it. Either way, the review exists to maximize the return together. 

Time saved is the easy one to quantify, and a good early win. But it’s now table stakes, and it isn’t enough to justify further investment. Workday found that for every 10 hours AI gives back, about four go to correcting its output. Time saved answers the individual’s question. The CFO’s question is firm-level: what changed in realization, write-downs, and billing cycle time, and what does that do to law firm profitability? 

If we forget our people, we won’t see the benefits of AI.

Value starts at capture and has to survive billing

In a law firm, the answer to that question starts earlier than most business cases look, because every invoice begins as a time entry. If the entry is wrong, everything downstream inherits the error. Time data is the most complete record of what a firm does and what it earns. That makes it the data foundation for law firm profitability, pricing, and staffing decisions, and an AI business case is only as reliable as the entries beneath it. That’s why ROI claims built on full automation are the hardest to defend: they model the machine and leave out everyone around it.

Capturing more time is the first half of the job. The second half is harder. Does the entry follow the client’s billing guidelines, reach the billing team clean, and survive the client’s review? An error nobody catches at entry doesn’t disappear. It travels, and it costs more at every step, until a partner is working out why an invoice was rejected. 

That’s the human element firms forget. If we forget our people, we won’t see the benefits of AI. People aren’t a brake on automation. They make the number hold. A timekeeper confirms their own time. A partner reviews exceptions only. Finance looks at the number again after go-live. 

What this looks like in Intapp Time with Celeste

Time with Celeste is built around that thinking: revenue and engagement intelligence for firm outcomes, with governance built in. It covers the path from time entry through prebill: 

• It captures work as it happens and drafts the entry, including which AI tools were used on a matter. The timekeeper confirms, so the human still owns the record. 
• It checks firm and client rules as the entry is written, while the timekeeper still remembers the work. 
• It checks those rules again at prebill from the same source, so an entry that passes at initial entry isn’t re-argued at billing. 
• It makes each step measurable, from adoption and same-day recording through compliance, rejections, and write-offs. 
• It shows which AI tools are used on which matters and what effect they have on hours and realization, so a firm can test its AI business case against its own data. 

Accountability for value and outcomes is shared between the firm and its technology partner. The result depends on how the firm adopts, checks, and enables the technology, and the business case is where that gets agreed. 

Before you sign, ask

Whatever business case you’re shown, ours included, ask these questions: 

• Who built it with you, and what did you agree to measure before go-live? 
• Who owns the number after go-live, by name? 
• Where does the number stop: saved, captured, billed, or collected? 
• When the number comes in below plan, what happens next? 

Let the case move with AI. Keep it anchored to the outcome.

See how Time with Celeste captures and checks time in the interactive tour.