AI Organizational Change: The Human Side

I'm Kurt Schmidt, founder of Schmidt Consulting Group and host of The Schmidt List podcast, and I want to be direct about something that I see derailing AI initiatives across B2B services firms right now. AI organizational change is primarily a people problem wearing a technology costume. Until organizations treat it that way, they will keep spending enormous sums on tools that half their teams never use.

I recently had a conversation with Beth Elliott, a transformation consultant who works with mid-to-large enterprises on the human side of innovation, and everything she shared reinforced what I've observed across years of advising services firms and agencies. The pattern is consistent: companies arrive at transformation with a solution already in mind, discover the actual problem is three layers deeper, and resist going back to start. That resistance is where AI initiatives go to die.

Why Do Organizational Silos Make AI Adoption So Hard?

Silos are the structural enemy of AI organizational change. They don't emerge from bad leadership or negligence. They emerge from natural organizational growth: you build a finance team, a product team, an engineering team, a sales team, and each one grows vertically. Everyone gets better at their own discipline. And then one day leadership turns around and realizes these groups have stopped talking to each other in any meaningful way.

Each silo is incentivized to stay in its lane. That's how teams earn their budgets. That's how managers protect their headcount. There's no structural reward for looking across the organization and asking what's actually blocking success. When an AI initiative arrives, each silo interprets it through its own lens, and you end up with leaders who have radically different definitions of what "AI" even means for the business.

I've watched this play out in a specific and predictable way: the CTO thinks AI means infrastructure automation, the CMO thinks AI means personalization across every customer segment, and the CFO thinks AI means cost reduction. None of them are wrong. But if those three people can't align on a shared definition before the project kicks off, the initiative fractures along the same silo lines that created the problem in the first place.

The solution is cross-functional alignment before solution design. Get the right leaders in the same room. Have each one sketch out their own problem statement. Then surface where those problem statements conflict and where they overlap. The overlap is where you should start. This connects directly to because the same diagnostic discipline that sharpens a firm's external positioning applies here internally.

What's the Actual Problem When AI Organizational Change Stalls?

The stated problem is almost never the real problem. This is one of the more consistent things I've observed across engagements, and it's something Beth illustrated well with a specific case I found clarifying.

A health insurance company brought in transformation consultants to build a single pane of glass for their call center agents. Call times were running at twice the national average. The solution seemed obvious: unify the interfaces, reduce the systems agents had to work through. But when the research team went deep into what was actually happening on those calls, they found something the original scoping had missed entirely. The agents and the customers were working from different information. The knowledge base the agents used was out of sync with what the company published on its website. Agents spent the bulk of each call reconciling conflicting data rather than answering questions.

The single pane of glass would have helped. But without fixing the knowledge management gap, call times would have stayed elevated. A different solution, or at least a much wider scope, was required.

This happens constantly in AI implementations. A company decides it needs an AI-powered CRM layer. The real problem is that the data going into the CRM has been fragmented across four acquisition-era systems for six years. The AI surfaces garbage faster. Per Gartner's research on AI adoption barriers, poor data quality remains the top technical barrier to AI success in enterprise environments. But the deeper issue underneath the data quality problem is almost always organizational: nobody owns the data, nobody is incentivized to clean it, and no cross-functional process exists to maintain it.

When someone brings me in because their AI initiative has stalled, I start by asking what problem they originally hired the technology to solve. Then I ask what problem they found when they got inside the system. The gap between those two answers tells me everything.

How Should Organizations Design AI Change Management Processes?

Change management for AI organizational change needs to start well before any tool is selected. I want to be specific about what that means structurally, because "change management" as a phrase has been diluted to mean almost anything.

The best change management professionals I've worked with over the years are communicators first. Everything else is secondary. Because when you're asking people to change how they work, the only thing that determines whether they engage or resist is how well they understand why the change is happening and how clearly they see themselves in the new picture. An announcement email from the C-suite is notification.

The approach that actually works involves pulling affected people into the design process early, before solutions are finalized. Not as a consultative checkbox. As actual co-creators. When people feel like they were heard in the design phase, they behave completely differently when implementation arrives. I've seen this contrast up close. Teams who were consulted late become obstacles. Teams who were brought in early become advocates.

There's a useful myth to discard here: people don't resist change. People resist feeling like more work is being dropped on them with no real input and no real benefit to them personally. Address those two things directly in the process, and resistance drops substantially.

Here's a practical structural framework for AI change management that I've seen work across different organization sizes:

The critical discipline here is staying in ideation longer than feels comfortable. Most organizations jump from problem statement to vendor selection in a matter of weeks. The organizations that actually see AI adoption succeed spend more time in the co-creation phase and less time in implementation rework. This connects to service delivery operations because delivery quality upstream determines everything downstream.

Why Does AI Get Overpromised Inside Organizations?

I remember when Big Data was going to change everything. It was going to predict customer behavior, eliminate uncertainty, transform every industry. It delivered real value in pockets. It also generated years of expensive consulting engagements built on dashboards nobody looked at after the first quarter.

Generative AI is following a structurally similar arc. The promises are enormous. The operational reality inside most organizations is that AI means faster and cheaper, and even that depends entirely on what problem you're pointing it at. I've talked with leadership teams where each person around the table had a fundamentally different mental model of what the company's AI initiative was actually going to produce. That's a communication and alignment problem that will destroy the initiative regardless of how good the underlying model is.

There's a pattern I've noticed in how AI gets positioned internally. An executive hears a strong case study at a conference. They come back energized. The IT team gets tasked with "implementing AI." No one has defined what success looks like. No one has inventoried what data they actually have and whether it's clean enough to train on. No one has talked to the teams who will use the output. Six months later, there's a pilot project, some promising demos, and no measurable change in how the business actually operates.

Some organizations have legitimate reasons to move carefully. I've worked in engagements where master service agreements with clients explicitly prohibited the use of generative AI tools like ChatGPT in any deliverable. Heavily regulated industries, particularly financial services and healthcare, carry real legal exposure around AI-generated content and data handling. Treating those constraints as obstacles to work around is the wrong posture. Treating them as design parameters is the right one.

How Do You Get Leaders Aligned on AI Transformation Goals?

This might be the hardest part of AI organizational change, and it's the one that gets the least attention in implementation frameworks. You can have the clearest technology roadmap in the world, and if the CEO is optimizing for top-line revenue while the CFO is optimizing for margin and the CHRO is optimizing for retention, the initiative will be pulled in three directions simultaneously.

I've been part of organizations running 140 concurrent initiatives. Nobody can track 140 things. Nobody knows which ones matter. And so the default becomes whoever shouts loudest gets resources, which rewards urgency theater over strategic clarity.

The fix is slower than most leaders want it to be. You have to surface the goal conflicts explicitly before the project starts. Get the leadership team in the same room. Let each person articulate what they believe the AI initiative is supposed to accomplish. Then work through the misalignments directly rather than pretending they don't exist.

One approach that works: start from an unconstrained vision. Ask the leadership team to describe what success looks like if all constraints were removed. Then, layer the constraints back in one at a time: regulatory requirements, budget, existing technical debt, organizational capacity. The constraints will narrow the picture considerably. But starting from the unconstrained vision ensures you're prioritizing the right things within those constraints rather than defaulting to whatever's easiest to measure.

Leaders also need to think hard about what they're incentivizing. I've talked to C-suite executives who are frustrated that their directors aren't collaborating across departments. My response is usually the same: you're still rewarding them on individual department metrics and asking them to behave cross-functionally. The incentive structure always wins. If you want AI organizational change to take hold, the people responsible for driving it need to be measured on shared outcomes. This connects to leadership and team alignment because the behaviors you see are almost always behaviors you're paying for.

What Role Does Customer Input Play in Successful AI Adoption?

Organizations resist involving customers in transformation work for predictable reasons. Sales teams worry about relationship exposure. Legal teams worry about disclosure. Operations teams assume customers don't understand the technical constraints. I've heard every version of this argument, and I've watched it produce solutions that customers then refused to adopt.

The Boeing Starliner program is an interesting reference point here. The astronauts who were going to ride in that capsule were part of the engineering conversations throughout development. Their direct involvement wasn't treated as a liability. It was treated as essential information. And the logic holds in less dramatic contexts: if you're building something that directly affects someone's experience, getting their input before you've spent the development budget is almost always cheaper than discovering the problem after.

One illustration from the transformation work I've followed closely: a team ran customer conversations at the start of a project week, ran internal innovation sessions midweek, then went back to the same customers at week's end with early concepts for feedback. A customer who had been visibly frustrated at the start of the week told them, by Friday, that she felt heard. She said it felt like the team had found a way to solve for every issue she'd raised. That's the outcome of bringing customers into the process. The alternative is shipping something polished that solves the wrong problem.

AI organizational change specifically benefits from this discipline because AI tools built without real user input tend to optimize for what's measurable rather than what's useful. A model can be technically impressive and completely wrong for the actual workflow it's supposed to support. Getting end users into the ideation phase, even with rough concepts presented on simple slides or a whiteboard, surfaces those mismatches before they become expensive.

Key Takeaways

• AI organizational change succeeds or fails at the people layer. Tool selection is secondary to alignment, communication, and co-creation.
• The problem organizations hire technology to solve is rarely the actual problem. Go deeper before committing to a solution path.
• Cross-functional alignment is a prerequisite. Leaders with conflicting definitions of success will fragment any AI initiative along existing silo lines.
• Incentive structures drive behavior. If you want cross-departmental collaboration, you have to measure and reward it explicitly.
• Customer and end-user input in the design phase reduces implementation rework and drives adoption. Bring them in early with low-fidelity concepts.
• Regulated environments have legitimate constraints on AI tool use. Treat those as design parameters.

I covered adjacent territory on transformation and organizational design in The Schmidt List, and the thread running through nearly every conversation is the same: the technology is rarely the hard part. The hard part is at the human layer. The organizations that figure that out first will get far more value from AI than the ones still chasing the magic bullet.

The question to sit with: if you polled every leader in your organization right now and asked what your AI initiative is supposed to accomplish in the next twelve months, how many different answers would you get?