AI ROI Fails When Management and Leadership Misalign

Welcome to the show. Daniel, here is the sentence I cannot stop thinking about: on Monday morning, the CEO says, "AI is transforming the business," and by Tuesday afternoon, a manager is still sitting there checking every single line the tool produced because if it goes out wrong, their name is on it. That Tuesday afternoon image is the whole thing for me. "Every single line" is not transformation.

That's a second job. That's a manager doing quality control on a machine they were told would save time. Exactly. And I think we keep misdiagnosing this as a tooling problem, as though the answer lives in the next model, the next vendor, the next rollout plan.

But the fault line is not technical. It is organizational. AI ROI fails when leadership and management are not actually describing the same company. [questioning tone] Say that more plainly.

Because "not describing the same company" sounds elegant - and I know you enjoy an elegant sentence - but what does it mean in the room? Fair. It means the executive team thinks the organization has clear priorities, permission to experiment, and a path from pilot to scale. Meanwhile the managers living inside the work experience something much messier: extra review, vague rules, unclear risk, and one more initiative layered on top of nineteen others.

One group says, "We are moving." The other says, "We are absorbing impact." And those are not small wording differences. "Moving" versus "absorbing impact" - that's the split.

I coach a lot of new VPs, and this is where they get trapped. Upward, they hear the language of transformation. Downward, they hear, "I don't know who's allowed to approve this, I don't know what good looks like, and I don't have another four hours to babysit a chatbot." Yes.

And once that mismatch hardens, the organization begins to lie to itself in very polished language. Dashboards show activity. Town halls show enthusiasm. Pilots exist.

Demos happen. But the lived reality has not shifted in the places where work is actually decided, reviewed, and shipped. Which is why this is less about artificial intelligence than ordinary management. Not glamorous management, either.

Decision rights. Workflow design. Incentives. Escalation paths.

The boring plumbing. If the plumbing is bad, the smartest model in the world still leaks. Let me make it concrete. Tuesday, 2:17 p.

m. A customer operations manager gets told her team should use AI to draft responses. Sounds efficient. But nobody has said which messages are safe to automate, who owns errors, or whether response time or accuracy matters more.

So what does she do? She checks every draft line by line. The tool did not remove work. It changed the shape of work and increased her risk.

And there is such an important distinction there: changed the shape of work, increased the risk. Because from the executive floor, that same scene may be reported as adoption. "Fifteen agents are now using AI-assisted drafting." But usage is not impact.

Activity is not relief. Right. Or take the pilot with no decision rights. A team spends six weeks testing AI for internal reporting.

They get decent results - maybe not miraculous, but decent. Then they hit the wall: legal hasn't weighed in, IT hasn't approved data access, no one knows who funds the next phase, and the VP sponsoring it sort of assumed somebody else would carry it. So the pilot becomes a museum exhibit. People point at it.

Nobody lives in it. A museum exhibit is painfully accurate. The pilot is displayed as evidence of innovation, rather than used as a mechanism of change. And here's where I may disagree with you a little.

I don't think this is only a leadership clarity issue. Sometimes leaders are perfectly clear. The bigger problem is operational overload. Managers already have full calendars, staffing gaps, service targets, and quarter-end pressure.

Even a clear AI strategy can land like a sandbag. I think that's true, but only in part. Overload matters, certainly. Yet overload without clarity is chaos, and overload with clarity can at least be sequenced.

Where I push back is this: many leaders call something clear because they announced it clearly. Those are not the same thing. [laughs softly] That is annoyingly well put. "Announced it clearly" versus "made it clear."

Fine. I'll give you that. Because if I tell two hundred managers, "Use AI to improve productivity," I have not given clarity. I have transferred ambiguity downward.

They now have to decide what counts as acceptable use, what quality threshold matters, where to take exceptions, and whether experimenting will be rewarded or punished if something goes sideways. And there is a third problem - incentives. If a manager is measured on error reduction, compliance, and throughput, they are going to behave very differently from an executive who is measured on strategic narrative and future value. The manager hears "adopt AI" and thinks, "Wonderful, and when this breaks, I get blamed first."

Yes. That is the Tuesday-afternoon truth executives often miss. The manager is not resisting transformation in some abstract, cultural sense. The manager is making a perfectly rational calculation about workload, accountability, and personal exposure.

This is where corporate language becomes actively unhelpful. We hear phrases like "accelerating our AI journey," "embedding intelligent workflows," "unlocking enterprise value." And perhaps all of that is directionally fine. But in plainer language, many teams are dealing with confusion, burden, and drift.

"Embedding intelligent workflows" is often just Karen from finance copying text from one window into another and then checking whether the numbers got scrambled. Precisely. And I do not say that to be dismissive. I say it because reality deserves better nouns.

If we use polished language for messy conditions, we lose the ability to manage what is actually happening. Grab that phrase - "better nouns." Because that's what strategy decks often fail to provide. A deck can say alignment.

A town hall can say transformation. But neither one automatically changes who approves exceptions on a Thursday, who gets extra headcount when review work spikes, or which workflow is genuinely different now than it was ninety days ago. And that assumption - that alignment follows announcement - is one of the great management delusions. People hear the strategy.

They do not necessarily hear their role in it, their protection within it, or the trade-offs it requires. So they nod in the meeting and improvise in private. "Nod in the meeting and improvise in private." I'm stealing that.

Because that's exactly what happens. And when enough people do that, leadership reads surface compliance as momentum. Which brings us to the stronger point of view here: AI transformation is a management discipline problem far more than it is a model problem. Of course the models matter.

Of course capability matters. But most organizations are not failing because the model is insufficiently dazzling. They are failing because managers were handed ambiguity, extra risk, and no operating system for change. Yes.

If your managers cannot explain when to trust the tool, when to override it, who decides, and what success looks like, then you're not in transformation. You're in theater. So what should a VP actually do? First, go learn the manager experience directly.

Not from a steering committee summary - from the managers. Ask three brutally simple questions. Where does AI save you time? Where does it create rework?

Where are you making judgment calls with no policy behind you? And listen for concentration, not anecdotes. If seven managers in different functions all describe the same friction point - review burden, approval ambiguity, data access, whatever it may be - that is not local noise. That is organizational signal.

Second, map one actual workflow. One. Not "sales" or "operations" in the abstract. Pick something embarrassingly specific: drafting renewal emails, summarizing support tickets, preparing weekly forecast notes.

Then compare the official story to the lived sequence. Where does the work slow down? Where does someone duplicate effort? Where is the human judgment still doing all the heavy lifting?

I would add a standard that I think is quite important: if a VP cannot describe the manager experience in concrete terms, that VP is narrating transformation rather than managing it. In other words, if you cannot tell me what has become easier, what has become riskier, and which decisions remain muddy on the ground, then you do not yet know enough to claim progress. That's sharp - and true. Third, separate pilots from adoption with adult honesty.

A pilot means you learned something in a bounded space. Adoption means a team can use it repeatedly with known rules, known ownership, and acceptable risk. Those are different planets. And if you are a leader who wants to pressure-test that approach with peers, this is exactly the kind of work inside the AI Ready executive cohort.

It is not about collecting prettier AI talking points. It is about translating ambition into management practice. We will put the sign-up landing page naturally in the show notes at aibreakthrough.com/ai-ready.

And I like that resource because it forces the right conversation. Not "which tool are you excited about," but "where is the friction concentrated, who owns the decisions, and what would your managers say if you weren't in the room?" That's the useful level. So perhaps the cleanest test is also the most uncomfortable one.

If you walked past the strategy deck, past the town hall, past the dashboard - and you asked your managers, privately, on a random Tuesday afternoon, "Is AI genuinely helping you do the work, or is it simply creating more work in a more modern wrapper?" what would they say? And would their answer sound anything like yours? Because the AI fault line runs straight through the org chart.

Not between believers and skeptics. Between people declaring change and people carrying it. Leadership has to bridge that deliberately. [questioning tone] If your managers answered with total honesty tomorrow at 2:17 p.

m. - not in a survey, not in a town hall, just the truth - would they describe AI as leverage... or as another layer of supervision?