I founded and ran Good2bSocial for the better part of two decades. We built it into one of the leading digital marketing agencies for law firms in the country. Like every agency owner, I kept one eye on the competition: the outfit down the street, the national firm with the bigger team, the newcomer undercutting us on price.
Then, sometime in the last year, the picture changed. I realized AI was about to take over a large share of the work my team did every day. The real competition was sitting on the client's side of the table. It was the same client who could now do more than half of what we used to bill for, in-house, with a decent model and a free afternoon.
Agency founders who built real businesses are working through this too, watching the economics shift under the bread-and-butter work that paid the rent and kept the lights on between marquee wins. The mood is calm and clear-eyed, which is somehow harder to witness than panic would be. They run the numbers late at night and do not love what those numbers say back.
For most of my career, the deepest moat an agency had was production capacity.
You could make more, faster, and better than a client could on their own. You had a seasoned team and a process refined over years of repetition. Your team could turn out forty localized versions of a campaign by Thursday, at a volume and speed no in-house group could match. That raw-output advantage built the whole business.
That advantage is closing fast.
Big brands are now producing in-house the creative work that used to flow straight to agencies. Kimberly-Clark, Target, and Catalyst Brands have each pulled production onto their own teams: image and video, copywriting, localization, the optimization passes. Edelman is reorganizing its entire structure around AI. This is production at scale, happening out in the open. Brands are re-cutting the value chain in public, while the rest of us watch.
The same thing is happening in professional services, just further from the headlines. Law firms, accounting practices, and consultancies are heading the same way, building inside what they used to hand to outside specialists. Nearly half of legal professionals now use AI every day for marketing work, from drafting social posts to building business plans. The Big Four have turned it into a budget line: EY committed $1.4 billion to AI aimed partly at market insights and client engagement, while PwC and Deloitte built internal platforms that already handle content creation. Gartner now puts in-house marketing labor at a bigger slice of the professional-services budget than outside agencies. No managing partner is announcing that they cancelled their agency. They are funding the inside team instead.
It happened to one of our own clients. A midsize law firm spent a few months running their PR work through AI workflows we had licensed to them, then let their PR agency go. The firm decided it no longer needed what the agency was doing, because the work it used to bill for had become work the firm could do itself. There had been no fee fight and no botched campaign. I felt for that agency. A few years ago, it could have been mine.
The hardest part is the way it happened. The moat drained on its own, with no rival in sight, because the same tools that once lived only inside your agency now sit on your client's laptop.
That is a difficult thing to absorb. You did nothing wrong. The water simply found a new way down the hill. The reflex is to skip straight to the fix. It is alright to grieve the old model a little. You built a real business that worked for a long time, so watching its economics change is hard in a way that no pep talk fixes.
When a client starts pulling work in-house, they begin with the easy wins rather than the hard ones. The first tasks to go are the most repetitive and scalable: the endless campaign variations, the localized versions, the rough first drafts, the resize-this-for-nine-placements grind.
A junior marketer drops a brief into a model on a Friday afternoon and has nine usable variations before the coffee finishes brewing. They forward them straight to the VP, skipping the agency email and the usual round of revisions. The VP says nothing about it but files the moment away. Three months later, when the budget conversation comes around, that Friday afternoon is the reason your scope gets trimmed.
Look at where that work sits on your profit-and-loss statement. It is the profitable middle of the business. The output was never glamorous, but it paid the bills reliably, and it covered salaries in the slow months between big strategic wins.
That middle is exactly what AI hands back to your client first. The price of that production work collapsed for them the same way it collapsed for everyone else; talent had nothing to do with it.
The advice usually turns glib right around here. "Move up the value chain" sounds effortless from a stage. It is far harder when the profitable middle is what makes payroll on the fifteenth of the month. So I will not pretend the repricing ahead will be painless, because it will not be. Pretending the middle is coming back, though, costs you more, because it burns the year you could have spent preparing for it.
There is room to breathe in what comes next.
One category of work stays out of a client's reach no matter how strong their model becomes. It is the strategic judgment to look at a perfectly competent campaign and say "this is wrong, here is why," before real money goes behind it.
A model hands your client a confident answer every single time. The piece they lack is the practitioner who has run this kind of campaign across forty different brands and knows that answer is often the one that sinks the launch without warning.
That recognition lives in the people you have spent years training. AI makes a strong strategist noticeably faster, though it does nothing to create the instinct in someone who never had it. Your client owns the speed but lacks the seasoned expert who should sit on top of it and anchor your whole business from here.
The trap is to treat resistance as the answer. Holding the line against AI accomplishes nothing, because the clients reclaiming production run on the very same models you do. The agencies that come through this are the ones who climb toward judgment faster than their clients can climb toward it.
You already feel this one in your gut.
Commoditized execution makes hourly billing a melting iceberg.
Hourly production billing rests on one assumption, that the hours are scarce and valuable. The moment your client can generate the same forty variations in an afternoon, those hours stop being scarce. You can keep putting them on an invoice, but you are charging for work the buyer can now do at home. They know it, too. Every quarter, that conversation gets a little harder to win.
The firms moving early are repricing around outcomes and judgment rather than effort. They charge for the strategic call, the brand system, the decision that costs real money if you get it wrong. They package fifteen years of hard-won expertise instead of renting out hours a sharp intern with Claude can now approximate.
That shift feels risky. It changes how you scope a project, how you staff it, and how you talk about your value on a sales call. The alternative is worse, because competing on price against your own client's internal team is a contest no agency wants to win.
The practical moves are not complicated, even if none of them are quick.
Every one of these asks you to rebuild a habit the old model rewarded you for. They all point the same direction, up toward the work a model cannot do for your client.
This is the work my cofounder Sally Slater and I do all day now at InnovAItion Partners, on both sides of the table. On the agency side, we help firms turn their senior people's judgment into AI workflows they own, so the production layer gets faster without getting generic. On the law firm and professional services side, we sit with the in-house marketing, communications, and business development teams and help them build that same muscle from the other direction. Both rooms want the same thing: better work the client can feel. That is the only version of this change I would chase.
Your client becoming your competitor is the end of one version of the agency, the one that won on sheer volume and billed by the hour. That approach was always going to age out. AI only moved the timeline forward.
What remains is the reason you got into this work in the first place: the judgment, the taste, and the kind of relationship where a client trusts you enough to hear "do not do that" and thank you for the honesty afterward. That kind of work grows scarcer every year. Scarcity is what clients pay the most for.
You win this by climbing to ground the model cannot reach, then bringing your client up the cliff alongside you.
It is a harder business to run. It is also a better one. The founders willing to make that climb are the ones I would bet on.