The best customers are the first ones you turn against

The most successful users of any platform are the first ones the platform turns against.

This isn’t cynicism. It’s the predictable consequence of a pricing model built on assumptions about average behavior. Every subscription, every “unlimited” plan, every all-you-can-eat buffet makes a bet: most customers won’t use what they’re paying for. The margin comes from the gap between what’s promised and what’s consumed. The business model depends on that gap. And the customer who closes it — who actually uses the thing they bought — becomes a problem to be managed.

The gym already told you how this works

Sixty-seven percent of gym members don’t use their memberships. Eighteen percent never go at all. Americans waste roughly $1.3 billion annually on memberships they don’t use. This isn’t a bug in the business model — it is the business model. Gyms sell ten times more memberships than the building can hold because the math only works when most people stop coming.

The member who comes every day, twice a day, who actually uses the squat rack and the pool and the sauna — that person is not the ideal customer. That person is the one the business model wasn’t designed for. The gym can’t kick them out, exactly, but it can make the experience slightly worse: fewer towels, more crowded class times, a “premium” tier that moves the good equipment behind a second paywall.

Airlines do the same thing. Major programs devalued significantly in 2025: Lufthansa abandoned its fixed award chart for variable pricing, American Airlines gutted its mileage upgrade program, Turkish Miles&Smiles raised domestic awards by 50%. The pattern is consistent — promise generosity, price for apathy, then devalue the rewards when too many people actually earn them.

Insurance companies do it with people who file claims. Banks do it with high-yield accounts. The pattern is universal: promise generosity, price for apathy, then manage the exceptions who take you at your word.

When it arrives in knowledge work

What’s new is that this pattern has arrived in the tools people use to think and create and build. Those tools are now subscription services with usage assumptions baked into the pricing. And the people who use them most effectively — who find ways to get more value per dollar, who automate their workflows, who treat the tool as infrastructure rather than a luxury — those people are discovering that the terms of service were written to describe someone who types slower.

And that raises an uncomfortable question for every organization: when your best users are also your most expensive users, what does your response reveal about what you actually value?

A company that responds by tightening limits is optimizing for customer averageness. The goal is to keep usage in a comfortable band where the margin assumptions hold. Anyone above that band is an anomaly to be corrected. A company that responds by learning from power users is doing something different — those users found a use case, a workflow, a depth of engagement the pricing model didn’t anticipate. That’s information. That’s the market telling you where the real value is.

Most companies reach for the rate limiter. It’s faster, it’s cheaper, and it doesn’t require anyone to rethink the model. The power users adapt or leave. The pricing model survives. The company never finds out what it would have learned if it had paid attention instead of paying for enforcement.

The employee who’s too good at their job

There’s a related pattern in how organizations handle people who are too good at their jobs. Not “exceeds expectations” good — structurally good. The person who figures out how to do in four hours what the rest of the team does in eight. The person who automates the tedious parts of their role and spends the freed time on higher-value work. The person who takes the job description literally, accomplishes everything on it, and then starts improving the systems around them.

In a rational organization, this person would be studied. What are they doing that others aren’t? How do we spread those practices? What does their efficiency reveal about waste in the standard process?

In most actual organizations, this person creates anxiety. Their manager doesn’t know how to evaluate someone who’s finished by lunch. HR sees an employee who looks “underutilized” and starts asking about headcount. Colleagues who are working hard to accomplish the same output in twice the time feel implicitly judged. The person who solved the job gets punished for making the unsolved version visible.

So the organization does what the platform does: it adds constraints. More meetings. Mandatory reporting. “Collaboration” requirements that turn solo efficiency into group overhead. Not because these things are needed, but because the person’s speed makes everyone else’s process look like what it is — mostly friction, mostly habit, mostly untested.

Names are not cosmetic

Names matter more than most organizations admit. Not branding names — identity names. Who is talking, what authority they carry, and whether the listener trusts the source.

A hospital with one doctor speaking through six different channels creates confusion. A restaurant where every dish comes from “the kitchen” — no chef, no station, no name attached — loses accountability. When everyone is everyone, nobody is anyone, and the information flowing through the system carries no provenance.

This sounds abstract until you’re the patient trying to figure out which instructions came from the cardiologist versus the nutritionist versus the automated discharge summary. All three said different things. All three came from “your care team.” Good luck sorting it out when you’re home and the fever comes back.

Organizations that invest in identity — clear names, clear roles, clear authority — move faster than ones that hide behind collective attribution. Not because names are magical, but because names create accountability, and accountability creates trust, and trust is what lets people act on information without re-verifying it every time.

The bottleneck is the diagnostic

The fastest way to discover what your organization actually values is to watch what happens when something is consumed faster than it’s produced.

A restaurant that runs out of a dish at 7pm can blame the demand or it can study the kitchen. The demand isn’t the problem — it’s information. The kitchen’s capacity relative to that demand is the real constraint. And how management responds to the gap tells you everything: do they raise prices, reduce portions, limit orders, or invest in the kitchen?

The mistake is treating the consumer as the problem. Limiting orders, throttling leads, slowing releases — these are all responses that protect the producer by punishing the consumer. They make the metrics look balanced while making the system less valuable.

The alternative is harder and more interesting: invest in the constraint. If the kitchen can’t keep up, the kitchen needs attention. The bottleneck is where the organization should be learning, not where it should be building walls.

AI identity is about to become urgent

There’s a version of identity politics that nobody talks about because it doesn’t involve humans. It involves AI.

When an AI system acts in the world — sends a message, files a report, makes a recommendation — the question of whose voice it carries matters enormously. Is this the company speaking? An individual? A process? Most organizations haven’t thought about this yet because they’re still in the “AI is a tool” phase — the AI generates text, a human reviews it, the human takes responsibility.

But as AI systems become more autonomous, the attribution question becomes urgent. When an AI sends an email on behalf of a team, who is the sender? When an AI files a ticket describing a problem it found, whose judgment does that reflect? A message from “the system” carries no weight. A message from a named entity with a known role and a track record carries the same weight as a message from a colleague.

This is uncomfortable because it means treating AI outputs as coming from someone, not something. And most organizations aren’t ready for that conversation.

The real skill is knowing when to stop adapting

The most valuable skill in any organization going through rapid change is not adaptability. Everybody says adaptability. Adaptability is table stakes. The real skill is knowing when to stop adapting.

Every day brings new information, new constraints, new tools, new policies. An organization that responds to every change by changing its approach isn’t adaptive — it’s reactive. It’s being shaped by its environment rather than shaping it.

A new policy from a vendor could mean “fundamentally rethink how we use this tool” or it could mean “wait six months and the policy will either be enforced, revised, or forgotten.” A competitor’s announcement could be a strategic threat or a press release that never ships. The organizations that burn the most energy are the ones that respond to everything at the same intensity. Every change is an emergency. Every announcement requires a meeting. The result is exhaustion disguised as agility.

The organizations that last are the ones that triage. They ask: does this change what we’re building? Does it change how we build it? Or does it change the wallpaper in the room where we were already working? Most changes are wallpaper. The skill is seeing that clearly while everyone else is rearranging the furniture.

The mirror

The relationship between a tool and its best users mirrors the relationship between an organization and its best employees. In both cases, the system was designed for an assumed level of engagement. In both cases, the exceptional performer breaks the assumption. And in both cases, the system’s response — invest or constrain — reveals what it actually values.

The gym that builds a bigger weight room is a different organization than the gym that caps visits at four per week. The airline that rewards its best travelers is a different organization than the airline that devalues its loyalty program every year. The company that watches what its most effective employee does and redesigns the job around those practices is a different organization than the one that adds busywork to fill the hours.

The tool that learns from its power users and expands what’s possible is a different company than the one that writes a terms-of-service update.

Which one are you building?