Meeting with the board of a large company. The CEO says: "We want to deploy AI. What tools do you recommend?" I ask: "In which process?" A moment of silence. "Well, generally. We want to be more efficient." I push further: "Are you prepared to change how people work?" The answer I hear more often than any other: "No, our processes are established. We want AI to make them work better."
This is a statement that should trigger a red flag. And it almost never does.
Ambition Without Readiness
Boards in companies — from publicly listed firms to the mid-market — increasingly want AI. The pressure is real: investors ask, competitors communicate, industry reports sound alarms. Not wanting AI in 2025 is a reputational risk.
The problem isn't the ambition. The problem is the assumption behind it: AI will improve the company, but the company doesn't have to change. Just add a tool and results will improve. Like a new software program — you install it and it works.
Except AI isn't a software program. AI is a change in how people work. And changing how people work requires changing processes, roles, responsibilities, and — often — how decisions are made.
Why "Add AI to the Existing Process" Doesn't Work
Imagine a claims handling process in a manufacturing company. A claim arrives by email, an employee transcribes the data into the system, classifies it, assigns it to a department, waits for a response, manually prepares a letter. The entire process takes seven days.
The board wants to "add AI" to speed things up. So we add AI to classification — the model reads the email and suggests a category. Savings: five minutes per claim. The entire process still takes six days and fifty-five minutes, because the rest — transcription, assignment, waiting, manual letter — remained unchanged.
AI made no difference, because nobody changed the process. For AI to actually reduce claims handling time, you need to redesign the entire flow: automatic email parsing, automatic classification and routing, response generation for human review, CRM integration. That's not "adding AI" — it's changing the process, with AI as one element.
What Has to Change for AI to Make Sense
Roles and responsibilities. Someone must own the AI-enabled process. Not "someone in IT," but a business person who understands the process and has the mandate to change it. Without an owner, AI belongs to nobody — and what belongs to nobody gets used by nobody.
Decision-making. AI generates recommendations, predictions, classifications. Someone must decide: do we trust these recommendations? To what extent? Who verifies? What happens when the model is wrong? These questions require new rules, not just new tools.
Information flow. AI needs data. Not "data somewhere in the company," but data that is accessible, current, and in a format the model can process. In many companies, data sits in silos, spreadsheets, and email threads. Before AI can function, you need to change how information flows through the organization.
People's capabilities. This isn't about prompt training. It's about people understanding what AI does, what to expect from it, and when not to trust it. That requires investment in people — not one-time, but continuous.
"Don't Change Processes" Actually Means "Don't Deploy AI"
It sounds harsh, but it's true. A company that wants AI without changing processes actually wants a demo. It wants a presentation for the board. It wants a checkbox: "deploying AI." It doesn't want transformation — it wants the appearance of transformation.
And that's fine, if it's a conscious decision. It's worse when the board genuinely believes AI will work without changes, spends budget on tools and pilots, and then is disappointed by the results. "AI doesn't work" — I hear then. It works. But not in an environment that denies it the conditions to function.
A Board That's Serious About AI
From my experience, boards that actually deploy AI — not just declare it, but deploy it — share common traits:
They accept that change is part of the package. They're not buying an "AI tool." They're buying a change in how people work, with AI as one element. They know that means changing processes, roles, and expectations.
They start with one process, not "AI across the company." One process, one team, measurable outcomes. Then the next. Then the next. Not a grand transformation program with a hundred-slide PowerPoint — but a concrete change in a concrete place.
They give it time. They don't expect results after a month. They know that changing how people work takes one, two, three quarters. And they're prepared to support, measure, and adjust throughout that time.
They engage personally. They don't delegate AI to the "innovation team" or to IT. They participate in reviews, ask about results, remove blockers. People in the organization look at the board — if the board takes AI seriously, they will too.
Change Is Hard. Not Changing Is More Expensive.
I understand why boards prefer not to touch processes. Change is risky, costly, unpredictable. It's easier to buy a tool and hope.
But not changing while simultaneously investing in AI is burning budget. Licenses, pilots, trainings that lead nowhere, because the organization refuses to change the foundation AI is supposed to stand on.
I work with boards that want to understand what AI truly requires from their organization — not just technologically, but operationally and culturally. I lead strategic AI workshops for boards that don't end with a presentation but with a plan for change that someone takes responsibility for.
If you want AI that truly changes how your company operates — start with the question of what you're ready to change. That's where every serious conversation begins.