How law firms are billing clients for AI

How we know this

Sidebar puts one question a week to legal management professionals at firms of 10 to 200 attorneys. Members are verified by title, employer, and firm size before they are admitted, and every reply is private. This page draws on every Sidebar cycle that has touched this question, and it is updated as new replies come in. We publish patterns across the group, never individual firms, and only once at least five members have replied to that question. Full methodology at gosidebar.ai/methodology.

Efficiency kills the pass-through argument

AI makes work faster, and clients can tell. Billing a client the full cost of a tool that cut your attorney's research time in half is difficult to defend. Firms trying usage-based or matter-level pass-through are running into this directly, not as a hypothetical ethics question but as a client objection on an actual invoice. The problem is not the ethics rules or the billing system. It is that the client can do the same math you can.

Vendor pricing is pushing back from the other direction

While firms work out what to charge clients, vendors are quietly working out what to charge firms. Several members describe watching AI get bundled into products they already license, at a markup that shows up at renewal rather than in the original sales conversation. The response is not absorption. It is shorter contract terms, more selective renewals, and outright cancellations when the number does not clear a budget ceiling that was set before AI pricing entered the picture. Firms treating vendor cost as a fixed input are missing that it is moving too, usually in the direction that makes the client-billing question more urgent, not less. This is the piece of the billing conversation that gets the least attention internally, because it happens in a procurement negotiation rather than a partnership meeting. A firm can spend months debating how to bill a client for AI work and never notice that the underlying cost of the AI itself moved twice in the same period, which means the internal policy the firm eventually settles on may already be pricing against a number that no longer holds.

Client willingness, not ethics rules, sets the ceiling

Ask a firm why it absorbs AI costs rather than passing them through, and the answer is rarely a bar rule or a billing-system limitation. It is that the client will push back. Rate increases are already landing annually on most engagements, and there is little appetite left to add a distinct technology line on top of them. That constraint is commercial, not regulatory, which means it moves with the market rather than with a rule change. A firm waiting for bar guidance to settle the question is waiting on the wrong body. The client is the one who decides what sticks. That is a harder constraint to plan around than a rule, because a rule is published once and a client's tolerance shifts with every renewal conversation, every competitor's pricing move, and every year the base hourly rate climbs on its own. Firms treating client willingness as a fixed boundary are underestimating how much room actually exists to move it with the right framing, and firms treating it as infinitely elastic are the ones getting the pushback on an invoice.

License fees stay internal; attorney time does not

The model gaining ground is not a line item for subscriptions or tokens. It is attorney time: hours spent prompting, reviewing output, catching errors, and signing off. That framing works because it maps to something the engagement letter already covers, professional judgment applied to a task. License fees do not map to anything the client recognizes as a deliverable. The telling gap: 45% of firms that have added AI billing language to their engagement letters explicitly address how AI-assisted work gets charged, meaning most firms have named the tool without explaining how it shows up on the bill. That gap matters more than it looks, because the firms that skip it are not avoiding the question. They are deferring it to the first invoice a client actually reads closely, which is a worse moment to explain a new billing category than a paragraph the client signed off on before the matter opened.

Nobody has a finished answer

No firm has a settled policy. Vendor pricing models are still shifting, bar guidance is thin in most jurisdictions, and the efficiency math changes every time a firm renews a subscription. Firms defining the logic now, what counts as AI-assisted work, how it is categorized, how it appears on an invoice, will have a defensible answer when a client asks. The ones waiting for a finished external standard, one that fits every firm and every matter type, may be waiting for something that never quite arrives.

The broader billing shift is real, but it is not being driven by AI yet

A 2025 survey of nearly 4,900 U.S. law firms found that 72% now offer some form of alternative fee arrangement, rising to 90% among firms with more than 50 lawyers. That looks like confirmation of a market-wide move away from pure hourly billing. But the same survey found that 58% of firms report AI has had no effect on their billing practices at all, and where it has had an effect, the more common outcome is quiet efficiency gains inside the existing hourly model rather than a new AI-specific charge. Read together, the two findings say something more specific than a general claim that AI is changing legal billing. The alternative-fee shift is happening for reasons mostly unrelated to AI, and AI billing itself is still an open question for the large majority of the market, not a solved one a firm can copy from a peer.

Source: Best Lawyers, Law Firms Embrace AFAs, But Clients Want More Flexibility

What to do with this

Draft a one-page internal position on what AI costs will and will not be billed to clients, even if it is provisional. Several members acknowledged they have no settled policy and are making ad hoc calls matter by matter, and a short written position, even one the firm expects to revise, forces the partnership to align on the time-versus-license distinction before a client raises it directly. Cover three things at minimum: whether license and subscription costs stay internal, how AI-assisted attorney time gets categorized and shown on an invoice, and who signs off when a matter does not fit the policy cleanly. The firms with an answer ready are not the ones with the most sophisticated billing software. They are the ones who wrote the position down before they needed it. Revisit the page every time a major vendor contract renews, since that is the moment the underlying cost structure is most likely to have shifted out from under the policy without anyone noticing. A position written once and never touched again drifts out of date at exactly the pace vendor pricing moves, which recently has been faster than most firms revisit their own internal billing rules.

Frequently asked questions

Do law firms charge clients for AI tools?

Mostly not directly. About 65% of mid-size firms absorb AI costs entirely rather than passing them through, largely because a tool that cuts research time in half is hard to bill at full cost without the client objecting on the invoice once they notice the difference.

How should a law firm bill for AI-assisted work?

The model gaining the most ground is billing for attorney time, prompting, reviewing output, and correcting it, rather than for the software subscription itself. Time maps to something the engagement letter already covers; a license fee does not map to a deliverable the client recognizes as work product.

Why do law firms absorb AI costs instead of passing them to clients?

Client resistance, not bar rules or billing-system limits. Rate increases are already landing annually, and clients have little appetite for an additional AI line item, especially once they can tell the tool made the work faster and cheaper to produce.

Is AI driving the shift toward alternative fee arrangements at law firms?

Not primarily. A recent survey found 72% of firms now offer alternative fee arrangements, but 58% report AI has had no effect on their billing practices at all. The AFA trend and the AI billing question are running mostly on separate tracks, at least so far in the data.