81% of AI Projects Pay Off for the Firms That Measure. Intake Needs One Number

Insights & Use CasesBlog
81% of AI Projects Pay Off for the Firms That Measure. Intake Needs One Number

Three Gartner analysts published an unusually blunt piece on 5 November: the money going into AI for customer service is outrunning anyone’s ability to say what it returns. Service and support organisations put 12.7% of their functional budgets into AI, with median spending approaching $6 million in 2025, and across the use cases Gartner evaluated, roughly a quarter produced positive returns, a quarter produced negative ones and 11% broke even. Then the figure that should stop you mid-sentence: 42% of those leaders did not know what they spent, what they got back, or both. Seventy-eight per cent are raising the budget in 2026 regardless.

Set that against a separate Gartner survey of 1,303 functional leaders, which found what discipline is actually worth. The high performers — defined as the ones who track every initiative’s return, review it regularly and shut down what underperforms — reported positive returns on 81% of their AI projects, while the laggards could not state the return on 29% of theirs. The distance between a quarter and four fifths is not the technology. It is bookkeeping.

The measurement problem is an enterprise problem

A $6 million programme spread across deflection, agent assist, summarisation, routing and knowledge search is genuinely hard to attribute. Nobody can isolate which model call saved which minute of whose day, and the bill moves underneath you: Gartner notes one pricing change to GitHub Copilot in June multiplied the average client’s monthly bill by three to ten times. “Improvements in model efficiency do not guarantee reductions in enterprise AI spending,” the analysts write.

A seven-attorney firm has none of that complexity. Your AI has one job, one bill and one number behind it — and unlike a retailer counting deflected tickets, you already know what a single conversion is worth, because you know your average matter value.

The number is the consultation, not the conversation

So do not measure containment, resolution rate or minutes saved. Measure consultations booked from enquiries that previously got nothing.

The shape of legal intake makes that number easy to isolate. Prospective clients ring during a hearing, at nine in the evening, on a Sunday after an accident — and they ring several firms, generally retaining whichever one answers first. An unreturned call on Friday afternoon is usually a signed client somewhere else by Monday. The questions underneath are repetitive: do you handle this case type, what does it cost, how long does it take, do you speak Spanish. None of it is advice, all of it decides whether the caller keeps dialling. That is the part of intake at a small or mid-size practice that scales badly with headcount, because you cannot staff a receptionist for Sunday night on the chance the phone rings.

Where to start before you sign anything

Log one month. Every call, form and text that arrived outside staffed hours or while the office was busy, and how many got a human reply the same day. Multiply the unanswered ones by your own conversion rate and average matter value. That is your baseline, and it costs nothing but attention.

Then scope narrowly. One job: answer every call and message, qualify it, route by practice area, and send a booking link by SMS while the caller is still engaged — the same agent working the phone and the chat together, rather than five tools you cannot attribute. Track exactly one number against the baseline, monthly.

Finally, take the repricing warning seriously even at small scale. Before signing, ask for usage limits, advance notice of any change to per-minute or per-conversation rates, and a clean exit. Firms that do this end up in the 81%. The ones that skip it end up in the 42% who cannot say whether it worked — and that is a far more expensive place to be than simply waiting.

Working in law firms & legal services? See how TalkRev works for Law Firms & Legal Services.
By type of business: Law firms.

One short email per new post

No ads. Unsubscribe anytime.

Start saving hours every
week with TalkRev

Reduce repetitive questions, capture more bookings, and deliver a faster customer experience — without adding headcount.