The Version That Runs

Your budget, calendar and promotion list can make the strategy hard to believe. They show where money goes, what gets time, and who gets rewarded. The plan can describe a different company.

If you notice that gap, it's tempting to reach for hypocrisy. Somebody must know. Somebody must be choosing the story over the work.

That reading has an appeal. It puts the problem in someone else's character. Better leaders would mean a better company, and you could keep doing your part until they arrived.

But it misses something harder to fix. People can believe the plan while protecting habits that prevent it from happening.

They can also benefit from those habits.

Sincerity and self-interest don't cancel each other out.

I call that shadow strategy: the pattern the company follows, whether or not anyone planned it.

Mintzberg and Waters distinguished intended strategy from realized strategy. Their work treated strategy as a pattern in what an organization actually did. A plan mattered, but so did the actions that followed, including patterns nobody had intended.

That distinction changes where you look. You can read the plan to learn what the company hopes to do. To learn what governs it, you have to follow decisions into the work.

A budget records choices. A calendar gives some work room and crowds other work out.

Promotions teach people which choices are safe to repeat. Defaults carry decisions forward without anyone having to make them again.

None of these tells the whole story alone. Together, over time, they give you something to check against the stated priorities.

The gap can sit in plain sight. At Ciba-Geigy, a Swiss chemical company, it took the form of a memo that people found useful but wouldn't pass on.

A memo that went nowhere

Edgar Schein began working with Ciba-Geigy in 1979. The company wanted to become more innovative.

Its head of management development had invited Schein to work with top executives. Its chief executive liked his finding that people could be creative in any kind of job.

Schein found useful practices inside the business. In one account, he described a memo about how the company was handling a difficult downsizing. He asked his contact, the head of management development, to distribute it to managers who could benefit.

On later visits, Schein kept finding managers who hadn't received it. When a manager requested the memo, it arrived almost immediately.

The information was available. The person responsible for passing it on had access to the chief executive. Yet it didn't move.

Schein also struggled to get support for meetings across units. People agreed that more communication could help innovation, but that agreement didn't produce much communication.

The pattern extended beyond his work. An internal colleague developed useful help for one unit, then watched other units seek help outside the company before discovering his solution.

Nobody had to reject innovation. They could support it in principle while leaving the route between units closed. Schein's problem was to understand why that route felt wrong to the people expected to use it.

The answer was ownership. A manager's job was private ground, and competence meant knowing what that ground required. Unsolicited advice suggested the manager didn't know enough already.

Schein reported that managers compared it to entering someone's home uninvited. Passing on a useful memo could therefore carry an insult.

His contact wasn't simply withholding information for an advantage. According to Schein, he was uncomfortable about not distributing it and couldn't explain why. They had to uncover the shared assumption before his behavior made sense.

That is the blind half of the gap. The rule guided action before the person following it could put it into words.

The rule also protected something valuable to its members: authority within their own area. A manager could sincerely favor innovation and still resist information that unsettled that authority.

I see both at work. Calling the whole thing unconscious would let the protected interests disappear. Calling it all calculated would miss how hard the rule was to see from inside.

The same boundary could defend a person's standing without requiring a deliberate decision to defend it.

A gap like that has a way of sustaining itself. Information stays local, so other units have less chance to learn its value. Their lack of shared work then makes separation feel reasonable.

Corporate marketing encountered that boundary when it proposed integrated programs across divisions. Managers pointed to differences between selling to farmers in fields and visiting doctors in hospitals.

Those differences were real. But they also gave the existing separation a sound explanation. Nobody needed to ask what the divisions might learn together if their differences settled the question first.

Schein also reported that bad news was handled locally where possible, limiting its movement upward. Useful information struggled to move sideways; trouble could stay below.

Even a workaround could leave the boundary intact. Schein found that managers accepted information directly from him as an outside expert. He could get the memo through without changing what colleagues were permitted to give each other.

The immediate task improved. The assumption could survive.

The explanation feels true

The blind half isn't peculiar to Ciba-Geigy.

In 1977, Nisbett and Wilson reported a study involving four identical pairs of nylon stockings. Passersby assessed their quality and explained their choices.

The pair on the far right was chosen over the far-left pair by almost four to one. Nobody spontaneously named position as a reason. When asked directly, virtually all denied that it had influenced them.

The researchers were careful about something the story can tempt us to overlook: they didn't know precisely why the position effect occurred.

The result didn't establish that every explanation people give is false. It showed that a confident explanation can fail to account for an influence visible in the record.

Argyris and Schon made a related distinction between the theory people say they follow and the theory their actions imply. People may or may not know the two are incompatible.

That uncertainty matters.

You can't infer dishonesty from the gap alone.

Schein encountered the reach of Ciba-Geigy's assumptions when he described headquarters culture to managers at its New Jersey subsidiary. They recognized themselves in his account. The pattern wasn't confined to the place where he'd first seen it.

And it didn't stay inside the company.

The customer meets the boundary

Ciba-Geigy acquired Airwick partly to develop its ability in consumer marketing. Airwick made air fresheners, including Carpet Fresh.

At an annual meeting of top management, Schein watched a senior board member and researcher react to television ads for Carpet Fresh. He recalled the scientist saying, "You know, Schein, those things are not even products."

Airwick made products and was profitable. The difficulty was what counted as a worthy product inside Ciba-Geigy.

Schein described a company whose identity rested on science and important work, including fighting disease and starvation. Air fresheners didn't fit that identity. Airwick was eventually sold despite being profitable.

I don't read that as proof that selling it was wrong. A company can reasonably choose which businesses it wants to own.

What matters is when the governing preference became clear. The acquisition was meant to bring the company closer to consumer marketing. Owning the business exposed an assumption that made that closeness uncomfortable.

My read is that the boundary around managerial authority and the boundary around acceptable products served a related need. Each protected an established understanding of who belonged and what counted.

Customers couldn't meet the ambition to learn consumer marketing. They could only meet the products and business decisions that survived those boundaries.

I've seen the distance between internal success and market results in my own work.

Early in my career, I ran the same SEO for two agency clients. Both got the same rankings.

One grew. The other couldn't trace a return and eventually doubted that the SEO had done anything.

The reports were right. They measured the channel we'd been hired to improve.

The client who grew was already doing the rest: outreach, email, brand advertising, and work supported by its reputation. The other was ranking for an audience that wasn't going to buy. Bad reviews stood between a good ranking and a buying decision.

At first, I was distraught. I'd done the work, and the report showed it. But the report couldn't tell me whether the business had given those visitors a reason to become customers.

We'd been hired to fix the channel. The problem extended into how the business ran.

That experience is part of what I mean by Market-Side Building. I take responsibility for what happens when the work meets the market, following the problem past the channel that produced it.

An internal account can be accurate and still stop too soon.

The customer meets the version that runs.

Faster plans, unchanged permissions

A model can now write a plausible plan in seconds. I expect that to make shadow strategy harder to spot unless checking changes with it.

My concern isn't that the plans will all be bad. A clear, sensible plan can still describe work that the budget, incentives and permissions won't support.

I expect the gap to widen when a company can revise its stated strategy faster than it changes those conditions. The language can keep improving while the customer keeps meeting the same business.

Microsoft's Copilot Autopilot announcement gives me a concrete question to follow. It describes an agent that works without waiting for prompts, follows threads, and returns to work while a person is elsewhere.

The announcement refers to permissions, audit and governance. But it doesn't name an approver or state what happens by default before an unattended action. Its Message Center notice leaves admin control details for a later update.

That isn't evidence that Autopilot has no controls. It means those pages don't answer the question. I wouldn't fill the silence with either reassurance or alarm.

Microsoft's Cowork documentation is more explicit, and it concerns a different capability. By default, Cowork asks for approval before an automated task sends email, posts a message, or changes a shared system. Users can pre-authorize actions when creating a task.

That difference is concrete. It tells me where a person must intervene and where prior permission can carry the work forward.

I expect those settings to matter more than language about control. I also expect customers to bear the cost when a company's account of oversight exceeds the oversight its settings provide.

I'd be wrong if companies used AI to check decisions as readily as they used it to describe them. Evidence would mean documented approval settings matching actual actions, mistakes feeding back into changed defaults, and better customer outcomes.

More convincing plans wouldn't prove me wrong.

A record connecting the plan to changed work would.

Words can change work

Still, a strategy document can do more than explain decisions already made. It can give people a shared direction and change what happens next.

Steve Yegge's account of Amazon offers a useful example. In his 2011 post, he recalled Jeff Bezos requiring teams to expose data and functions through service interfaces. No other route for communication between systems was allowed.

Yegge also recalled a threat of dismissal for noncompliance. This is his account, not an independently confirmed copy of the original mandate.

On his telling, the instruction mattered because it constrained implementation. Teams couldn't satisfy it simply by agreeing that interfaces were important. They had to build and use them.

That is a real limit on reading strategy backward from behavior. If only completed actions count, you can dismiss a serious change before it has had time to work.

Plans deserve room to become real. Emergencies also happen, and one spending cut doesn't establish a hidden priority. A promotion can have reasons you haven't seen.

The useful question is whether the company can connect its intention to changes in the work.

What has been funded? What route has closed? What can people now do that they couldn't do before?

Shadow strategy is worth investigating when the pattern persists through those questions. It becomes an excuse for cynicism if every answer merely confirms what you've already decided.

Some organizations build ways to make that comparison part of the work.

Toyota describes a production system in which workers can stop the line when they find a problem. The stop brings an abnormality into view and calls for help. Quality has a mechanism beyond agreement that quality matters.

At Bridgewater, Ray Dalio has described recording meetings so employees can hear discussions themselves. His principle of radical transparency includes exceptions; it isn't total access to everything.

Neither practice proves that its organization always lives by its stated values. A mechanism deserves checking too.

But both make a principle more concrete. People have something to use, and its use can leave evidence.

Alignment takes work because agreement doesn't settle what happens under pressure.

The test is whether the principle still has a route into action when following it becomes inconvenient.

Put the two accounts together

Start with a page, not another strategy exercise. Put one stated priority beside the recent decisions that should support it.

Use the budget, calendar and promotion list as records, not verdicts. Include the defaults that let work proceed without a fresh decision. Ask what happened when the priority became costly.

Then follow that work to the market. What did the customer actually receive? Where does the internal record stop?

Ciba-Geigy gives that check a concrete test. The stated aim was more innovation. Useful information existed, but the normal route between units didn't carry it.

Schein's response was specific. He sent information directly as an expert, or worked through senior authority when he needed wider action. He changed the route after learning why the expected route failed.

Airwick exposed another boundary. The company wanted consumer-marketing knowledge, but ownership revealed limits on the kind of business it wanted to be. That preference became available for scrutiny through a consequential decision.

You don't have to begin by deciding who is sincere. Begin with the mismatch and give people room to explain it. Check their explanation against the next decision.

I need the same discipline. I've handed over reports that stopped before the outcome. My confidence didn't make them complete.

The budget, calendar and promotion list aren't there to catch someone out. They're there to help you find what the company is prepared to change, and then check whether the customer feels it.

Everything else is narration.

The free letter brings one idea to check against the record, on Sunday mornings, most weeks.