I had not planned to spend the afternoon with Harry Truman.
It was the week before the United States turned 250 — flags out, fireworks already being tested somewhere down the block — and we happened to be in Kansas City for an entirely unrelated reason: the Dutch national team was playing Tunisia in the World Cup.
Close by, Independence turned out to be the hometown of Harry S. Truman. With a few hours to spare before kickoff, we wandered into the Truman Presidential Library, reading about one of the harder stretches in 250 years of U.S. history.
It is hard to spend much time with the period from 1945 to 1953 and still believe our own moment has a monopoly on uncertainty.
When certainty runs out
The atomic bomb. The Cold War taking shape almost before the war itself had ended. NATO, which still had to be argued into existence. The Marshall Plan, with its enormous bet that rebuilding Europe was not charity but strategy. A domestic economy trying to find its footing after wartime production, with inflation and strikes close behind.
And underneath all of it was fear.
By the early 1950s, some of that fear had curdled into something uglier. Suspicion started to stand in for evidence. People were accused of disloyalty on thin grounds. Government employees, writers, professors, and actors lost work and reputations not because wrongdoing had been proven, but because the accusation itself had become powerful enough.
Walking through the exhibit, the present did not feel simple. But it did feel less exceptional.
The threats today are different. For most of us, they are less existential and less external. They do not arrive as armies, alliances, nuclear threats, or the rebuilding of a continent.
But that does not make them imaginary.
If you lost your job, if your role is changing, if your profession is being reshaped, if the economics of your work are suddenly being questioned, the uncertainty is not abstract. It is personal. It affects how people listen in meetings, how they interpret strategy, and how much trust they place in the next initiative.
That is the part I kept thinking about on the way home.
How did people make decisions when so much was unsettled? How did they know whether a bet as large as the Marshall Plan was working? How did they keep moving with so much uncertainty?
Clearly, none of it came from following an existing script.
There was no playbook for a nuclear monopoly. No precedent for rebuilding a continent. No clean model for turning a wartime economy back into a peacetime one. The decisions were too large, the evidence too incomplete, and the consequences too uncertain.
That feels familiar in a smaller but still consequential way.
What counts as enough evidence to make a decision?
A firm deciding what to do with GenAI is not making Truman-era decisions. But it is making decisions before the evidence is complete.
A workflow is being redesigned before anyone knows exactly what the second-order effects will be. A staffing model is being questioned before the technology has fully proven what it can and cannot replace. A client expectation is shifting before the firm has settled how quality, price, speed, and judgment should now relate to each other.
In that kind of environment, the hardest question is not whether to use AI.
The harder question is what will count as enough evidence to make a decision.
What evidence is strong enough to change a workflow? What evidence is strong enough to change a role? Who is accountable if the work becomes faster but not better? Who is accountable if margin improves but employee satisfaction suffers? Who decides whether a successful pilot is ready to become a firmwide operating model?
These are not technical questions alone. They are questions of governance, accountability, and judgment.
A firm may decide that AI changes a workflow. It may decide that AI changes staffing assumptions. It may decide that AI changes pricing, leverage, or service delivery. But those are decisions. They require evidence. They require accountability. And they require a way to revisit the decision when reality turns out to be more complicated than the pilot.
A constitution, not a playbook
So what kept the Truman years — for all their turbulence, and despite their worst instincts — from completely flying apart?
Not a playbook. Nobody had one.
What held was something underneath the playbook: a constitution. A fixed set of rules for how decisions get made, challenged, reviewed, and corrected, even when nobody agrees on what the decision should be.
The constitution did not prevent every bad decision. It did not stop fear from doing damage. But it created mechanisms that could eventually check excess, expose mistakes, and force correction.
That is the lesson worth bringing into a GenAI strategy.
A firm will not get a universal AI playbook for this. Nobody has deployed GenAI in your exact firm, with your lawyers, your clients, your economics, your risk tolerance, and your culture before. The work is too specific for borrowed certainty.
But a firm can build its own version of a constitution.
It shouldn't be a long document. Not another governance artifact that sits untouched after the launch meeting. Not an all-encompassing GenAI strategy either. Instead, it should be a practical set of decision rules that leaders agree to before the pressure arrives.
What evidence is strong enough to change a workflow?
What evidence is strong enough to change a role?
When do we redesign the work before we reduce the people?
Who is accountable for the outcome — the lawyer using the tool, the practice leader, IT, innovation, management?
How will we know, three months from now, whether the decision worked?
What would make us reverse course?
Those questions matter because GenAI decisions rarely arrive as clean technical choices. They show up inside pricing conversations, staffing conversations, client service conversations, associate development conversations, and partner expectations. By the time the decision is visible, the assumptions behind it may already have hardened.
Deciding well before certainty arrives
That does not mean firms should move slowly. Truman’s era was not defined by patience. It was defined by consequential decisions made under pressure.
But pressure is exactly why the decision rules matter.
A playbook tells people what to do when the situation is already understood.
A constitution tells people how to decide when it is not.
That may be the more useful model for GenAI. Not a script. Not a promise. Not a strategy. Not a claim that this particular technology will solve everything or destroy everything. Just a steadier way to make decisions when the evidence is still forming and the consequences are already real.
The country did not get through 250 years of turbulence because it always had a map. It endured because it had a way to argue, test, challenge, and correct decisions when the map ran out.
That is a better thing to build before your next AI initiative than certainty ever was.
Rise Above AI Chaos: A Business Fable About Leading Organizations Through the AI Revolution