What AI changes when the work itself is the thing you sell.
Agencies and law firms are making their most expensive AI mistake right now. They're using it to get faster.
That sounds like a good thing. It's a trap.
Most of what gets written about AI adoption is written for companies where AI is an internal efficiency play. Customer service, HR onboarding, inbox triage: all of it internal work that happens between a company and its own staff. You take a cost center and you make it cheaper.
That is not your situation.
If you run an agency or a law firm, the work is not a cost center. The work is the product.
The thing you sell is judgment, delivered by people, billed by the hour or the retainer. When AI touches that, it doesn't just change your costs. It changes what you sell and how you charge for it.
So the generic advice doesn't fit. Here's the version that does.
Two kinds of productivity
There are two kinds of productivity. You're probably only using one.
The first is doing what you already do, faster. The second is doing things you otherwise wouldn't do at all.
The second kind is the one that keeps you out of the trap.
The faster trap
Plenty of firms are chasing the first kind. Same work, less time.
The associate drafts the client alert in 40 minutes instead of three hours. The account team builds the media list before lunch instead of over an afternoon.
Feels like a win. Looks like one on a Tuesday.
Then the renewal conversation happens.
Your client's CFO has read the same headlines you have. They know AI cut your drafting time. And they ask the obvious question.
"If AI does this in a third of the time, why am I paying the same retainer?"
If you run an agency, you've already heard a version of it. "If AI cuts twenty percent of the time it takes to manage our media, shouldn't our fees drop twenty percent too?"
If you run a firm, it's arriving through procurement and outside counsel guidelines. What are you using AI for, and is that efficiency coming back to the client as a discount?
This is the trap.
When the only thing AI changes is your speed, you've handed your client a stopwatch and taught them how to bill you with it. You've converted your own efficiency into a fee cut. You did the work to make yourself cheaper.
And faster, by itself, is a race you lose. There is always someone younger, hungrier, or offshore willing to be faster and cheaper than you.
That race has a bottom, and the bottom is a commodity. Your whole career has been about not being a commodity.
I'm not telling you to slow down. Speed is table stakes now. But speed cannot be your whole story, because your client can read the meter as well as you can.
And the damage runs deeper than fees.
The economics of both our industries run on a pyramid. A few senior people whose judgment clients pay for, sitting on top of a wide base of junior people who do the volume.
AI is strongest at exactly the work that base used to do: first-pass review, research, drafting, list-building.
If your only move is "do that work faster," you are dismantling the bottom of your own pyramid without building anything to replace it. You shrink the base, the margin pops for a quarter, then you wake up with no bench and a client who now expects everything in half the time for half the price.
Faster doesn't just commoditize the work. It commoditizes the firm.
The work you could never afford to do
Now the second kind. Doing things you otherwise wouldn't do at all.
Few firms are pricing this yet.
Think about all the work you already know would make you a better partner to your clients. The work you never do because the numbers never add up.
Take a law firm. The work you'd take on if the hours were free:
- A competitive intelligence brief on your client's three biggest rivals, refreshed every month instead of once at the pitch.
- The 50-state survey you'd never staff for a mid-size matter.
- A client alert that goes out the morning the ruling drops, not four days later when it's already old news.
- A standing memo tracking regulatory change in your client's industry that no associate has the hours to maintain.
Now an agency. Same idea, different output:
- Always-on monitoring of every journalist who has ever covered your client, with a personalized angle ready the day they publish.
- Earned-media analysis that ties coverage to your client's pipeline instead of impressions.
- Pitch research that used to take a week, now run on every prospect, every quarter.
None of that was possible before. You knew it was valuable. You just couldn't bill enough hours to staff it at a price your client could afford.
AI makes it affordable for the first time.
And this part is new. That work is not on the meter. The client has never paid for it before, so there is no old price to discount against.
You're not discounting work you already do. You're selling value the client could never buy before.
That is the conversation that protects your retainer. The line isn't "we got faster." It's "we now do things for you that no firm our size could do a year ago."
One of those answers invites a discount. The other justifies a raise.
AI is a tool, not a strategy
A lot of firms have an AI tool. Almost none have a vision.
A tool is "we bought Copilot." A vision is naming the part of your client work that was impossible to deliver, then showing exactly how you deliver it now.
The difference shows up in your pitches.
You can see it in competitive pitches. The AI audit is now a live criterion. Clients have moved from "do you use AI?" to "show me the specific workflows and the people who run them."
The agencies losing those pitches walk in with a tool. The ones winning walk in with work the client didn't know they could get.
The same pressure is hitting law firms through RFPs and outside counsel guidelines. Clients no longer ask whether you use AI. They ask what new thing your use of it produces for them.
If your answer to that is "we're faster," you're back in the trap.
How to do this without boiling the ocean
Now the practical part.
You don't choose between the two kinds of productivity. You sequence them.
First, fix the boring, frequent stuff:
- Meeting prep and call summaries.
- First-pass document review.
- Research briefs and inbox triage.
If the task happens every week, has a known output, and a human can check it in five minutes, it's a candidate. This buys back time and builds your team's fluency. Call it the floor.
Then take the time you just freed and spend it on the impossible-before work.
The standing brief. The always-on monitoring. The project your best people would build if they had a week they don't have.
Two moves, in order: clear the existing pain points for immediate value, then place strategic bets on the bigger work.
The mistake is stopping at step one. You automate the drafting, feel productive, then never reinvest the freed time into anything new.
The savings just sit there until the client asks for them back. Which the client eventually does.
And one more thing, because I owe you honesty here.
The first workflow you try will probably disappoint you.
We put AI on a workflow last year that we were certain would be a slam dunk. The first month, it took us 150 percent of the time it used to take. We nearly killed it.
We didn't. We fixed the instructions, fed it the right context, built a few checks. Now it does in an hour what used to eat a day.
But that first month was humbling. And if anyone says their first AI workflow worked on day one, they're selling you a story.
This is craft. It rewards the firms willing to be bad at it for a few weeks.
This was never about replacing your people
One last point, because it's the one I care about most.
None of this works without your experts. The change I'm describing only happens because a senior person decides which new work is the priority, points the tools at it, and judges whether the output is good enough to put your name on.
AI didn't shrink the value of that judgment. It made it the whole job.
The associate who knows which precedent controls. The account lead who knows which journalist will roll their eyes at a pitch and which one is waiting for it. The partner who can look at a draft and feel, in their gut, that something is off.
That is the work clients always paid the premium for. AI just cleared away the hours of grunt work that used to bury it.
The firms that pull ahead over the next two years won't be the ones with the most tools. They'll be the ones who freed their best people for the work only their best people can do, then sold the new work that became possible.
Where this leaves you
AI is going to make your firm faster whether you plan for it or not. That part is happening to you.
The only real choice you have is what you do with the speed.
You can hand it to your clients as a discount. Or you can reinvest it into work they could never buy before and become the partner they can't afford to lose.
Faster makes you cheaper.
Better makes you necessary.
Choose the second one, because your retainer depends on it.
What's the first piece of "impossible-before" work your firm would take on if the hours were free? I'd love to hear it.