I.
“Slop” was Merriam-Webster’s 2025 Word of the Year. By April, three-quarters of new web pages were AI-generated.
Everyone’s blaming the tools. They’re wrong.
The slop isn’t coming from the technology. It’s coming from organizations that distributed creative authority across committees and then handed everyone AI tools. Now they’re producing exponentially more mediocrity, faster.
In your organization, who has the authority to look at 100 AI-generated options and say “this one—and only this one—is worth making”?
If that answer is “a committee,” you’re producing slop. Even if the individual humans on that committee are talented. Even if your AI tools are best-in-class. Committees don’t have taste. Committees have process. And process plus infinite options equals infinite mediocrity.
The organizations that will matter in 2026 aren’t the ones with the best AI tools. They’re the ones with one person whose judgment directs the tools. Including knowing when to stop using them.
A decade ago, being a B+ creative could sustain a career. The gap between you and A+ talent was real but survivable. Not anymore. The B+ creative now competes with AI that’s faster and cheaper. The A+ creative uses that same AI to amplify judgment that can’t be systematized. The same bifurcation is happening to entire organizations.
II.
The cliff isn’t when quality drops. The cliff is when it can’t come back.
Crispin Porter + Bogusky was Agency of the Decade in 2009. Over 1,000 employees. The only agency in history to win Cannes Grand Prix in all five categories. By 2021: 40 employees, 2 clients. By 2024: they dropped “Bogusky” from the name entirely.
Everyone knows Alex Bogusky left in 2010. The part that gets left out: they tried to bring him back.
In 2018, Bogusky returned as “Chief Creative Engineer” with full authority. Chuck Porter later admitted what happened: the client base had become “more conservative,” which made the agency “more conservative,” which “softened creative edge.” They closed offices, consolidated to Boulder, stopped entering awards. Bogusky left again eighteen months later.
Even the original curator couldn’t reconcentrate what had been distributed.
I was at CP+B for part of this story. Too junior to see the shift while inside it. But I remember the moment I noticed years later: the work stopped being undeniable. It became... reasonable. Not bad—competent, strategic, defensible. But the thing that made clients say “how did you get away with that?” was gone.
That’s the cliff. Not the decline. The point of no return.
For years I thought CP+B’s collapse was about losing one irreplaceable person. It wasn’t. It was about a sequence of structural decisions that made the organization incapable of producing the work that built its reputation—even when the irreplaceable person came back.
That sequence has a pattern. And it’s predictable.
The cliff isn’t when quality drops. The cliff is when it can’t come back.
III.
Three stages precede the cliff. Each is survivable on its own. Together, they’re fatal.
Stage 1: Authority Diffusion
Veto power gets distributed into “input.” The person who once could kill work now has “a strong voice at the table.” Decisions require consensus. Multiple stakeholders. Collaborative review.
What actually happens: Work gets approved based on who advocated longest, effort invested (”we can’t kill it after three weeks”), exhaustion (”let’s just ship something”), and compromise (”everyone gets a little of what they want”). Never based on whether it’s good enough.
Andrej Karpathy—who built Tesla’s Autopilot and helped found OpenAI—on why reinforcement learning is “terrible”: “You’re sucking supervision through a straw.”
When an AI model solves a problem correctly, the training process upweights every decision in that trajectory—including wrong turns that happened to reach the right answer. One binary signal (correct/incorrect) broadcasts across thousands of decisions. The model can’t distinguish the brilliant move from the lucky stumble.
Committees suck supervision through the same straw. Work gets approved; the organization implicitly validates every compromise that led there. The three weeks of meetings. The stakeholder who talked longest. The moment everyone agreed to just ship something. All of it reinforced as “how we make good work.”
The feedback signal is too compressed to carry quality information. It only carries survival information.
Still reversible. Reconcentrate authority in someone with taste and the work improves immediately.
Stage 2: Talent Calibration
A-players leave for places where their judgment matters. They don’t announce why—they just stop returning recruiters’ calls. B-players stay because committee structure protects them. Their work never gets killed, which feels like success. The organization’s taste baseline drops without anyone noticing.
New hires calibrate to the new standard. They’ve never seen undeniable work ship from this organization. They don’t know what’s missing.
Karpathy calls this “silent collapse”—what happens when AI systems train on their own outputs. The distribution of possible outputs narrows. Each sample looks reasonable. But ask ChatGPT for a joke and it knows maybe three.
Organizations collapse the same way. Train new hires on B+ work and the range of what the organization can produce narrows. Any individual project looks fine. The collapse isn’t obvious until you notice the agency can only tell three jokes.
The only fix is A-players from outside—exactly what a collapsed organization can no longer attract or recognize.
Harder to reverse. You’ve lost the people who could recognize excellence. Reconcentrating authority doesn’t help if the person you concentrate it in doesn’t have taste—and the people who could tell the difference are gone.
Stage 3: Client Portfolio Drift
Ambitious clients leave because the work no longer justifies the relationship. Conservative clients stay—and new conservative clients arrive, because reasonable work is what they wanted anyway. The revenue base shifts to people who actively prefer mediocrity.
This is the cliff.
Even if you reconcentrate authority. Even if you hire back the original curator. Your economic structure now depends on clients who don’t want what excellence produces. The organization can’t afford to make the work that would rebuild its reputation because the clients paying the bills would fire you for making it.
This is why Bogusky couldn’t revive CP+B. Not because he’d lost his taste. Because the client base had drifted to people who didn’t want what his taste produced. Porter said it plainly: “more conservative” clients made the agency “more conservative.”
The cliff isn’t creative. It’s economic.
The cliff isn’t creative. It’s economic.
IV.
Concord was Sony’s flagship bet on live-service gaming. $400 million, eight years in development. It launched in August 2024. Fourteen days later: shutdown, full refunds, studio closed.
The postmortem from one of its developers: “Design by committee. Failure of leadership to listen to the art team.”
Concord wasn’t Stage 1 or Stage 2. Concord was born at Stage 3. The economic structure—committee approval required at every gate—guaranteed mediocrity from inception. No cliff to fall off. They built the product at the bottom.
Same company, same year: Helldivers 2 launched to 400,000 concurrent players and became a cultural phenomenon. One studio had concentrated creative authority. The other had committees. $400 million bought Sony fourteen days of Concord. A fraction of that budget bought them a franchise.
V.
Four questions to locate yourself on the path to the cliff.
1. Who can kill work?
Not “who has input.” Not “who has a strong voice.” Who can walk into a room where fifteen people worked for three weeks, say “this isn’t good enough, start over,” and have that be final?
If the answer requires a meeting to determine, you’re in Stage 1.
2. When was the last time they did?
If you can’t name a specific recent instance, the authority exists on paper only. The organization has learned that veto power won’t be exercised. Work calibrates accordingly.
3. Who left in the past two years—and why?
Look at your departures. Not the ones who left for titles or money. The ones who left for “fit” or “the work.” They’re telling you something about Stage 2.
4. Which clients would fire you for making great work?
Not which clients want great work. Which clients would actively resist it—because it’s too risky, too unfamiliar, too hard to get approved internally?
If that’s more than a third of your revenue, you’re approaching Stage 3. Your economic structure is starting to require mediocrity.
VI.
What I know:
Concentrated curatorial authority produces better creative work than distributed committee structures. This was true before AI. AI widens the gap exponentially. The cliff—the point where even the original curator can’t revive what’s been distributed—is real, and it’s economic, not creative. Once your revenue depends on clients who prefer mediocrity, excellence becomes unaffordable.
What works:
Curatorial authority replicated, not distributed. Wieden+Kennedy has sustained quality across eight global offices for decades because they didn’t democratize the dictatorship. They cloned it. Regional leaders with real veto power. Autonomous units, each with its own taste authority. The model scales through replication, not through process.
But autocracy alone isn’t the answer. Plenty of taste dictators have driven agencies into the ground through ego, insularity, or failure to develop others. The formula requires all three elements: judgment, authority, and the productive anxiety that keeps the autocrat in the work rather than above it. Confident autocrats produce different failures than confident committees—but they still fail. The anxiety is the difference between a dictator who ships excellence and one who ships ego.
The holding companies are testing the opposite hypothesis right now. The Omnicom-IPG merger is killing FCB, DDB, and MullenLowe—three storied brands folded into BBDO and TBWA. Meanwhile, eight of eleven Adweek 2025 Agency of the Year winners were independents.
The industry is bifurcating. One path leads toward the cliff. The other leads toward something harder to build and harder to sell—but that actually works.
VII.
By the time you recognize Stage 3, your revenue structure has made excellence unaffordable. The clients who would pay for great work have already left. The clients who remain would fire you for making it. The talent who could produce it departed in Stage 2. You’re optimized for mediocrity, and the optimization is self-reinforcing.
The organizations making this choice now—before the cliff—will be the ones that matter in 2027.
I wrote the underlying thesis last month in “The Anxious Autocrat”—the formula of taste, authority, and productive self-doubt that makes creative organizations work at the highest level. If the pattern here feels familiar. If you’ve watched quality drift without being able to name why, or recognized your organization somewhere in the three stages. Read that next. It’s the diagnosis. This is what happens when organizations ignore it.
