Moving off the billable hour, the way I actually did it

Why the billable hour breaks once AI is actually working

I didn't set out to abandon the billable hour on principle. It broke on its own, mechanically, once I had systems doing real drafting and research work in a fraction of the time it used to take me. A first draft of a standard filing that used to take me two and a half hours started taking forty minutes of my time, most of it review rather than composition. Billed at my hourly rate, that's a massive drop in revenue for the same delivered value, and clients who paid attention noticed the invoices shrinking for work that was, if anything, more consistent than before.

That's the real tension nobody tells you about when they pitch AI to lawyers: the billable hour actively punishes you for getting faster. Every hour of time you save is an hour of revenue you no longer bill, unless you change how you charge. I went through about four months of watching my own revenue compress before I forced myself to redesign pricing around the new reality instead of pretending the old model still fit.

The other problem with sticking to hourly billing once AI is in the loop is more subtle: clients start asking, reasonably, why they're paying full hourly rates for work that took you visibly less time. Once a client has seen one invoice that feels disconnected from effort, trust erodes fast. You either hide the efficiency gain, which is its own ethical problem, or you change the pricing structure to reflect it honestly.

Measure your actual delivery time before you price anything

1. 1. Log real completion time for a month, AI-adjusted

   For every matter type, track how long it actually takes you now, start to finish, including review and revision time. Don't estimate. I was consistently wrong estimating my own new speed by thirty to forty percent in either direction until I logged it for real.

2. 2. Separate variable matters from standard ones

   Standard petitions, standard agreements, standard filings: these are where flat fees work cleanly because the time variance is low. Genuinely novel or contested matters still carry more uncertainty and need a different structure.

3. 3. Price in a buffer for the outliers

   Every matter type has a tail: the client who needs six rounds of revisions, the filing that hits an unexpected procedural snag. Build fifteen to twenty percent buffer into your flat fee based on your logged variance, not on a guess.

4. 4. Re-check your numbers every quarter

   As your systems improve, your true delivery time keeps dropping. If you don't revisit pricing, you'll eventually be dramatically underpricing your own new efficiency, which is its own kind of leaving money on the table.

The three pricing structures I actually use now

I settled on three structures across my own work and the practices I've built systems for, and I use different ones for different matter types rather than picking one and forcing everything into it.

• Flat fee per matter type, for anything standard and repeatable: most immigration filings, most standard contract work, most routine estate documents. Priced off logged delivery time plus buffer, not off a guess.
• Value-based fee tied to the client's actual stakes, for higher-consequence work where the outcome matters far more than the hours spent: a business-critical contract negotiation, a dispute with real exposure. Here the price reflects what's at risk for the client, not what it cost me to produce the work.
• Retained capacity pricing for ongoing relationships, particularly with small business clients who want a lawyer they can call without a meter running: a fixed monthly fee for a defined scope of ongoing support, refreshed and renegotiated periodically as the relationship's shape becomes clearer.

Plan for a messy transition year, not a clean cutover

I did not flip a switch and move every client to flat fees overnight, and I'd tell anyone else not to either. Existing engagement letters, existing client expectations, and your own confidence in your new numbers all take time to catch up. For about a year I ran a mixed book: new matters and renewing clients moved to flat or value pricing, while a shrinking set of legacy hourly engagements ran out their natural course.

The hardest part of that year wasn't the client conversations, which mostly went better than I expected. It was my own nerve. There's a real psychological adjustment in quoting a flat fee for a matter and then delivering it in a fraction of the time you used to spend, because the old mental math (more hours worked equals more earned) doesn't apply anymore, and it takes deliberate effort to stop feeling like you're leaving money on the table when you finish early.

How to actually have the pricing conversation with clients

The framing that worked for me was never about AI at all. Clients don't care that a machine helped draft something; they care whether the price is fair and the outcome is good. I talk about predictability: you'll know the full cost before we start, and it won't creep upward because a filing took an extra round of revisions or because I had a slow week. That's a genuinely better deal for almost every client, and it's the honest version of what changed.

The clients who pushed back hardest on flat pricing were, almost without exception, the ones who had been burned by open-ended hourly bills before and assumed any change was a trick. The fix there was transparency about how the fee was calculated, showing the logic (this matter type typically takes this range of hours, here's the flat fee that covers it with room to spare) rather than asking them to just trust a number.