If AI makes our work faster, should we charge clients less?

Why does this question make partners uncomfortable?

It usually surfaces the same way. Somebody in the management meeting says the quiet part out loud: if a report that took most of a day now takes an hour, are we really going to keep billing for the day? The room goes still, because everyone can see two answers and neither feels safe. Say yes and you have just cut your own revenue. Say no and you sound like a firm charging for work it no longer does.

The discomfort is a sign the question has been framed badly. It assumes the fee is a function of elapsed time, and that the client bought hours. That was never quite true, even when the invoice was expressed in hours. The client bought a judgement they could rely on, delivered by people whose names mean something, with the firm carrying the risk if it is wrong. Time was the measuring stick, not the product.

Which hours actually disappear?

Before touching a rate card, be precise about what has got faster. The first hours to go are almost never the advisory ones. They are the write-up after the meeting, the reformatting of a report into the house template, the hunt for last year's version, the chasing of a colleague for a missing figure, the assembly of a bid pack from six previous bid packs.

Ask which of those were ever on an invoice. Some was written off at the bill review. Some was never recorded at all, because nobody logs twenty minutes spent finding a document. Some sat in a line item the client queried every quarter. If the time you have saved lives mostly in that category, there is nothing to pass on. You have recovered margin on work the client was already getting free, which is the point of fixing reporting or bid production in the first place.

You cannot have this argument on instinct. It needs a real split between billable and unbillable hours on the specific deliverable, which is why reliable time capture matters more once efficiency enters the picture, not less.

What happens if you cut the fee anyway?

Suppose you decide to be generous and reduce the price of the faster deliverable. Three things follow, and they are worth thinking through before you offer anything.

First, the reduction is permanent and the saving may not be. Rates come down easily and go back up with difficulty. If your efficiency gain turns out to be narrower than you thought, or the review burden grows because the work is more complex than the average matter, you are holding a lower price against the old cost base.

Second, you have taught the client that your fee tracks your internal effort. That conversation will now repeat every year, about every improvement you make, and it caps your incentive to improve anything. A firm that funds better systems out of its own margin has a reason to keep funding them.

Third, you have said something about your value that you probably do not believe. If the report was worth its fee because of the thinking in it, the thinking has not changed. If it was worth its fee because it took a long time, you had a pricing problem before AI arrived.

When is a reduction the right call?

There are real cases. Commoditised, high-volume work priced against competitors is one: routine searches, standard filings, repetitive policy documents, high-turnover placements. On that work, price is the basis of competition and efficiency is how you defend the relationship. Passing some of it on is a commercial decision, not a moral one, and it should be made deliberately for a named client or workstream.

The second case is where a client is genuinely paying for your inefficiency. If your fee assumed three people rekeying the same data and now it does not, and the client knows that, holding the old price will cost you trust worth more than the margin. Honesty here is cheaper than being found out later, which is the argument we make about telling clients you use AI.

The third is competitive tendering. If a panel review is coming and efficiency lets you bid a price others cannot match while keeping your margin, that is not a discount. That is winning.

What should change instead of the rate?

The more useful move is to change what you sell, not what you charge for the same thing. Three shifts are available.

• Price the output, not the elapsed time. A defined deliverable at a defined price puts efficiency where it belongs, inside your margin, and gives the client the certainty they wanted. This is the logic we apply to our own fixed fee against day rates position.
• Sell more of what is scarce. If senior time has been freed from write-ups, capacity for advisory work has gone up. That capacity is worth more than the fee reduction you were contemplating. The capacity calculator gives a rough sense of what has been released.
• Improve the offer at the same price. Faster turnaround, a monthly update instead of a quarterly one, a same-week response on queries. Clients notice service before they notice price, and none of this hands back margin.

How do you answer a client who asks directly?

Do not deflect, and do not make a speech about value. Answer with specifics of your own. Something like: the drafting and assembly on this report is quicker than it was, the analysis and review are not, and the review is where the risk sits. A named person still reads and signs off everything that goes to you.

Then offer something concrete: a faster turnaround, a wider scope, or a fixed price where there was an estimate before. A client who asked about AI savings mostly wants to know they are not being taken for a ride. Give them evidence you have thought about it and most will not press further.

Where should a firm start on this?

Start with one deliverable, not with the rate card. Pick the piece of work clients ask about most, break its cost into the parts that involve judgement and the parts that involve assembly, and see how much of each is actually billed. That split is your answer for that deliverable, and it will differ across a law firm's matter types or an accountancy practice's compliance and advisory lines.

Then get honest about whether the saving is real. A gain you cannot see in recorded time, bill realisation or turnaround is a story, not a saving, which is the test set out in measuring return on AI. Feeding that back into fee estimating is what turns efficiency into pricing you can defend.

If you do not yet know which process is costing you the unbillable hours, find that before you decide what to do with the savings. The audit takes about three minutes and tells you where the hours are going.