No Exit
The most dangerous platform in your organization is the one nobody is allowed to leave.
Enterprise Architecture knows how platforms get here better than anyone, partly because it helps put them there. Duplication gets spotted, consolidation gets funded, a roadmap appears, and the roadmap becomes a mandate. EA drives the standardization. Later, it's the same function called in to deal with the consequences when the mandate calcifies.
It never feels dramatic in the moment. A sensible recommendation becomes a rule. A rule becomes the only allowed path. But once a platform is mandatory, something subtle and essential disappears: the ability for teams to walk away.
In AI, where the ground moves monthly, that loss blinds the organization faster than it realizes. Your frontier becomes the platform team's backlog.
The Signals Organizations Depend On #
In 1970, economist Albert Hirschman published Exit, Voice, and Loyalty, a thin book with a long title and a sharp question: how do organizations find out they're getting worse? His answer: their members tell them.
- Exit. The blunt signal. Customers stop buying, people walk, the numbers move. Nobody has to be brave.
- Voice. The articulate signal. Someone says what's wrong to someone who can fix it. It carries more information and costs more to give.
- Loyalty. What keeps people using voice before they reach for exit. It buys the organization time to hear the complaint.
A healthy organization needs both exit and voice open, with loyalty buying time for voice to work. Now look at a mandated platform.
When Exit Is Removed, Data Goes Dark #
When a platform is required, teams lose the ability to choose another path, at least officially. Some still do, but they do it quietly, and the organization never sees the signal.
Then the mandate manufactures the opposite signal. Adoption climbs because adoption is required. Leadership watches usage trend up and reads endorsement. That's counterfeit loyalty: a metric that looks like satisfaction and measures compliance.
There's a second casualty. If no team is permitted to build on the emerging thing, nothing exists to compare the platform against. No baseline. The platform becomes its own benchmark, and a system that grades itself always passes.
Why Platform Teams Can't Hear the Truth #
With exit gone, voice is the only channel left. It routes straight to the people least able to hear it.
The platform team is closest to the problem, which should make them the best sensor. Three things get in the way.
They're too close. Familiarity smooths over the friction every new consumer hits on day one.
The baby is ugly, and it's theirs. Naming the platform's weaknesses means naming their own.
The costs are sunk. Even when they see it and admit it privately, years of investment make walking away look like waste.
Each one breaks a different step. Proximity breaks observation. Ego breaks reporting. Sunk cost breaks updating. I've written before about ego as a structural bug in the loop. Here you get all three bugs at once, in the team whose job is to notice.
Nokia lived this. When researchers interviewed 76 of its managers and engineers, they found top leadership afraid of competitors and middle management afraid of top leadership. Bad news got filtered on the way up. Some managers lied outright about launch dates and features. Leadership kept steering by a picture that was wrong before it arrived.
When Voice Fails, People Stop Trying #
Hirschman described three responses. Later researchers added a fourth, for when exit is blocked and voice doesn't work.
Neglect.
People stay, stop complaining, and disengage. They stop proposing alternatives because the last three proposals came back "use the platform." They stop flagging gaps because the gap goes on the roadmap and the roadmap doesn't move.
Psychology calls this learned helplessness. Psychologist Martin Seligman showed that when people are exposed to problems they can't influence, they later stop trying even when influence returns. Neuroscientist Steven Maier later clarified why: we learn agency from successful action. Take away the chance to act, and you take away the chance to learn you can.
It doesn't take much. In a 1970s stress study, people given a button that would stop the noise performed better, even though most never pressed it. The option was enough.
A mandated platform takes the button away. Teams never get the experience that would teach them they have agency.
And the organization loses the evidence that anything is wrong.
Shadow Systems: Exit That Doesn't Show Up #
Some teams skip neglect. They build what they need anyway, off to the side, and keep showing up in the platform's adoption numbers.
That's exit, but the kind that doesn't register. Hirschman's exit works because it's visible. This one is hidden by design. The organization pays for the fragmentation and still believes the platform is working.
These teams didn't give up on agency. They gave up on voice.
Hirschman noticed who leaves first: the most demanding customers. Here, that's your best engineers. Voice loses its sharpest critics before anyone thinks to ask them.
And their work is the exploration you refused to fund. The baseline you're missing already exists. Nobody will show it to you, because showing it means admitting they broke the mandate.
Mandates Protect Mediocrity #
Organizational theorist James March described the competency trap: you overinvest in the old thing because you're good at it. The mandated version is worse. Platforms that aren't especially good at their job, kept alive because the organization removed every mechanism that could have told it.
Every platform is mediocre at something. An organization that can't detect it is the failure.
Reopening Exit Restores Sight #
Platforms solve real problems. Keep them. Change what surrounds them.
Start with amnesty. Ask teams what they've built off-platform, no penalty. Your comparison data already exists.
Then reopen exit. When the platform can't keep up, let teams build elsewhere, and treat each defection as data about the platform. Netflix's paved road works this way: a supported default that teams can leave, as long as they own the cost.
Fund exploration as its own line item. Exploration is a sensor. In AI, where the field outruns any roadmap, that baseline is the only way to know whether your platform is a head start or a lag.
Measure the platform by what consumers would choose, not by what they're required to use.
The door was open the whole time. We just made it against policy to walk through.