A few weeks ago, an agency CEO walked me through the restructure they were about to launch. He had a name for it; there’s always a name for it now. Pods, or a flat model, or the operating system from a book for everyone that a peer of his swears transformed their firm. And the striking thing was how much he’d already gotten right. He knew his current structure wasn’t working. He could feel it in the margin, in the way work moved, in how tired his senior people were. The instinct was correct. The sequence was backwards. The model had arrived unattached to the diagnosis.
I’ve sat in some version of that conversation for over twenty years, across a few hundred agencies and consultancies, and the pattern barely varies. Leaders reach for structure when things stop working, which is exactly the right moment to reach; then they reach for a structure that sounds best, which is exactly the wrong basis for choosing one.
Mintzberg answered this in 1981
Henry Mintzberg published an article in Harvard Business Review that January titled “Organization Design: Fashion or Fit?” The title is the whole argument. His research had shown that effective organizations cluster into a handful of coherent configurations, and that which configuration works for a given firm depends on its situation: its age and size, its technical system, its environment, where power actually sits. Structure that fits the situation performs. Structure imported from elsewhere, however well it performed elsewhere, doesn’t.
He also documented the failure mode. Fashion, he found, drives organizations to adopt structures that are inappropriate for them but current — the matrix in the seventies, divisionalization before that, each wave sweeping up firms whose situations called for something else entirely. The models weren’t wrong everywhere. But they were wrong for the adopters who chose them because they were circulating. And if you look at all the bones that litter the trail named “Agile organization” you can see how hard it is to make it work, when it is ultra-fashionable…and you don’t know how to make it fit.
When Mintzberg needed an exemplar for his most fluid configuration, the adhocracy — the shape built for novel problems and shifting project teams — he pointed at the advertising agency. Your firm is in his book. Which means his theory wasn’t written about companies unlike yours; it was partly written about yours. My book, UNMANAGED, used Mintzberg’s work on adhocracies, and it extended into details that he never learned about how agencies work. I labeled that subtype of the adhocracy a NOCO, a naturally occurring chaotic organization.
Huge already ran this experiment, in public
You don’t have to reach back to the matrix era for an example. In 2021, Interpublic installed Mat Baxter as CEO of Huge, the Brooklyn-born digital agency; he was the fifth CEO in four years, inheriting slow growth, client churn, and thin margins. Baxter brought in The Business Model Company, a consultancy with a confident thesis about what’s wrong with agency economics, and together they rebuilt the firm around fixed-price productized services: a catalog of offerings in the back, client-facing teams in the front. It was, in effect, a front-back delivery model; the shape that works in banking and enterprise product businesses, where the back office manufactures standardized, separable capability the front can sell.
The transformation had everything a restructure could want: a decisive CEO, an outside architect, holding-company backing, and even an embedded chronicler — Michael Farmer, who documented the whole effort in Madison Avenue Makeover, published in mid-2023. By July of that same year, Huge was restructuring again and merging the client pods that had been central to the design. By early 2024, Baxter was out.
Fit: 1. Fashion: 0.
I’ve spoken with people who were inside that building, and the private account is rougher than the public one. But the public record alone makes the point. Effort, talent, and resources were all present. The flaw was an assumption imported with the model. Front-back structures depend on the back producing capability that survives being separated from the client. An agency’s back room doesn’t hold standardized product; it holds judgment-bearing craft that degrades the moment it’s cut off from client context. The precondition failed, so the structure failed, and no amount of execution quality could have saved it; the failure was upstream, in fit.
Smart, well-funded people got this wrong with a book documenting every step. That should recalibrate anyone’s confidence that they’d spot a misfit model from inside their own offsite.
Wrong models arrive sounding right
Two years ago a CEO told me his firm felt like a ship with barnacles on it. He was right; I’ve rarely heard the drag described better as a feeling. But you can’t solve for barnacles. The metaphor named his experience without naming a single thing he could change: not a role, not a handoff, not a decision path. Nearly every CEO who calls me has a name like that for what’s wrong — barnacles, molasses, herding cats, where-did-our-culture-go?, too many cooks — they’re always accurate, but not so useful except to inspire change.
Causes hide well. Structural misfit rarely kills a firm; it taxes one. The tax shows up as chronic drag: good people underperforming in roles that don’t fit the work, scoping that keeps missing, margin that bleeds a point at a time, client friction that everyone blames on the client. None of it arrives labeled structural, so it gets treated as a people problem, a process problem, a this-client problem. The fixes make noise, but fail while the cause sits untouched in the org chart.
If you’re a leader with an accurate feeling and no specific diagnosis, you are the ideal customer for a popular, confident model, because the model supplies what the diagnosis lacked: names, categories, a story about the pain. Borrowed specificity feels like insight when your own vocabulary has run out.
Or you may have people in your leadership team that propose a novel solution. A model like that can sound right because it flatters how the founder wants to see the company. Or it settles a long-running argument inside the leadership team by picking a side.
The most seductive part: because it promises relief from your current pain without requiring you to understand your current pain.
I think it is harder today… Certainly the workplace is more complex, but also because the situational factors Mintzberg called the drivers of structure are moving…all of them at once. AI changed the technical system of every knowledge firm, which changed the environment, which is shifting where power and value sit. The pressure to grab something has never been higher, and neither has the cost of grabbing the wrong thing.
Fit is diagnosed, not chosen
Your best chance of getting it right is to change the sequence, and it’s the same sequence whether the model under consideration is fashionable, bespoke, or mine. Before any model earns a hearing, the leadership team should be able to describe the firm’s actual situation with some precision: where the work really happens and how it has changed; what actually coordinates people today — the formal mechanisms or the three senior humans quietly holding it together; where judgment gets exercised, by whom, and whether the structure puts those people anywhere near the decisions. Most leadership teams, asked these questions cold, discover they’re describing the firm they had five years ago, or just aren’t sure these days because of how much has changed since they last really looked at something other than new business, delivery, and margin.
To be fair, it’s not easy: often, despite having some documentation (that nobody reads), there is no single way of operating. Each account runs its own way, shaped by whoever leads it; creative runs differently from media, media from production; what the org chart shows as one system is in practice a federation of local arrangements. Diagnosing a firm like that means describing the variation itself — which accounts run which way, and what the differences reveal — rather than averaging everything into a tidy picture that matches nothing real.
And the description has a second, harder half. The new structure and model should fit where you think you’re going, not where you’ve been. That means looking at which work AI is absorbing, which clients are changing what they buy, where judgment will need to sit two years from now. A structure fitted precisely to today is fitted to a situation that is already leaving. This is also the part a fashionable model most invites you to skip, because the model arrives with its own picture of tomorrow, and adopting the picture feels like having done the thinking.
Two ways to see through the fog
The good news is that getting past the fog costs almost nothing but patience. The CEO from the opening had the courage part handled; restructuring a firm you're responsible for takes plenty of it. Part of what he was missing could have been acquired in a few weeks, starting with two exercises any leadership team can run on its own. There are more; these two are where I'd begin.
First, describe your best-running account and your worst, side by side. Not the org chart version — how each actually operates: who decides what, how work moves, where it stalls, who quietly holds it together. In a firm that's really a federation of local arrangements, the difference between those two accounts is the most honest picture of your operating reality you can get. Any model worth adopting has to explain that difference. This takes an afternoon of honest conversation, and it's usually a shock.
Second, hunt workarounds. Ask your people what they do to get around the official way of working — the side channels, the shadow spreadsheets, the approvals everyone skips. Workarounds are where the real operating model lives; every one marks a spot where the current structure and the actual work disagree. A list of ten workarounds is a better diagnostic than most consulting decks.
Neither exercise tells you which model to adopt, and the two together are a start on the diagnosis, not the whole of it. What they do is convert the barnacles into specifics — named accounts, named stalls, named workarounds — and specifics are what fashion can't survive. A model that sounded right in the abstract now has to explain your list, and most can't. The one that can will feel less like a revelation and more like a confirmation; that's roughly how you know it fits.
And if no model on the table explains what you found, that's not a failure of the exercises. That's the exercises working — and a sign the rest of the diagnosis is worth doing.
If this one hit close to home, share it with another agency leader who's got an offsite coming up — you might save them from a very expensive crate.
The adhocracy-to-NOCO argument gets the full treatment in my book, UNMANAGED, if you want to go deeper on why agencies work the way they do.
And if you want pieces like this in your inbox when they publish, subscribe here — no spam, just the occasional article worth your time.
