AI Change Management: Why AI Projects Stall on People

Frank Shines

November 10, 2026

AI change management: leaders and frontline staff adopting an AI workflow together

Quick answer: AI change management is the people side of an AI rollout: readiness checks, a sponsor who shows up, redesigned work, equipped managers and adoption metrics. It matters because most AI projects stall on people and process, not on the model. BCG puts roughly 70% of AI implementation challenges in people and process, 20% in technology and 10% in algorithms.

Every executive team I talk to has a version of the same story. The pilot looked great. The demo got applause. Six months later, usage is flat, the business case is quietly being rewritten, and someone is asking whether the model was the wrong choice.

It usually wasn’t. The evidence points somewhere less comfortable: AI projects stall because organizations treat them as technology installs instead of changes in how people work.

Why do AI projects stall on people, not technology?

AI use is now nearly universal. McKinsey’s 2026 State of AI survey found that nearly nine in ten organizations use AI regularly in at least one business function. But only 37% of respondents attribute any EBIT impact to AI, essentially unchanged from the year before, and only about 6% qualify as AI high performers.

Gartner reports that by the end of 2025, at least 50% of generative AI projects had been abandoned after proof of concept. Gartner lists poor change management among its critical failure points, noting that without it, even technically excellent GenAI tools see minimal adoption.

BCG’s Where’s the Value in AI? research put the balance plainly: about 70% of AI implementation challenges stem from people and process issues, 20% from technology, and only 10% from the algorithms themselves. Yet the algorithms often get most of the attention.

If 70% of the risk sits with people and process, 70% of the plan should too. We see the same pattern when we rescue a stalled pilot: the model is rarely the first thing that needs fixing.

How do you test change readiness before you build?

Most teams check data readiness. Few check people readiness. Before a single workflow changes, you should be able to answer four questions:

  • Who is affected, and how much? Map every role that touches the new process, from executives to the night shift.
  • What are they losing? Familiar screens, informal workarounds, sometimes a sense of expertise. Name it out loud.
  • What is already on their plate? A team absorbing a new ERP module, a reorganization and an AI rollout in the same quarter is not resistant. It is saturated.
  • Do they believe the reason for the change? BCG’s 2026 AI at Work survey found that only a third of frontline employees say leadership communicates clearly about AI.

That last gap is where adoption quietly dies. People do not resist change they understand and trust. They resist change that feels like it is being done to them.

What does an effective AI sponsor actually do?

Prosci has studied change management for more than 25 years, and active and visible executive sponsorship has been the top contributor to success in its research since 1998. In Prosci’s data, projects with extremely effective sponsors were 79% likely to meet objectives, compared with 27% for projects with extremely ineffective sponsors.

McKinsey’s 2026 survey points the same way: AI high performers are about twice as likely as others to say their senior leaders demonstrate commitment to AI initiatives.

A real sponsor does more than approve the budget. They explain why the change matters in their own words. They show up at the floor meeting, not just the steering committee. They make the hard calls when two departments’ priorities collide. And they keep showing up after go-live, when the novelty fades and the real work begins.

If your AI sponsor cannot name the three business metrics the project is supposed to move, you do not have a sponsor yet. You have a signature.

Why redesign the work before you train for it?

McKinsey found that nearly three quarters of AI high performers report fundamentally redesigning workflows because of AI, compared with about one quarter of other respondents.

Deloitte’s 2026 State of AI in the Enterprise report found something similar from the other direction: 60% of companies have redesigned jobs around AI capabilities to at least a moderate degree, but only 15% report a significant shift in decision-making authority. People get better information but are not empowered to act on it.

Dropping a tool into an unchanged process gives you an expensive add-on. Redesigning the process, including who decides what, gives you a different result. That means the training plan cannot be written until the new workflow is, and training is not a one-hour webinar the week before launch. It is role-specific practice on the real workflow, before and after go-live. Our guide to training for ability, not attendance covers how to build it.

How do you equip the managers in the middle?

Frontline adoption runs through frontline managers. Prosci lists engaging and supporting people managers among its seven best practices, alongside engaging frontline employees directly.

That matters because the middle of the organization is often where AI strain shows up. In McKinsey’s 2026 survey, 47% of mid-level managers and individual contributors said they had experienced at least one AI-related negative effect, compared with 31% of executives and senior managers.

Give managers what they need before their teams ask: a clear explanation of why the change is happening, answers to the five questions their people will ask, early access to the tools, and a simple way to raise problems that actually get fixed.

What should you measure instead of training attendance?

Ninety-five percent of staff completed training tells you almost nothing. What matters is whether people are using the new workflow correctly, whether quality and cycle times are improving, and whether people are falling back to old workarounds.

Prosci’s research shows how much this discipline pays off. Among participants with excellent change management, 88% met or exceeded project objectives. Among those with poor change management, only 13% did. That is roughly a sevenfold difference.

What should you do on Monday?

  • List every AI initiative in flight and name the executive sponsor for each. Mark any where the sponsor has not spoken directly to affected teams in the last 30 days.
  • Pick one initiative and map the roles it changes, including frontline and shift-based staff.
  • Ask five frontline employees: “Why are we doing this?” Compare their answers with the business case.
  • Check what else those same teams are absorbing this quarter.
  • Replace one training completion metric with one adoption or outcome metric.

How does Analytics AIML approach AI change management?

AI does not fail in the server room. It fails in the gap between what a new system makes possible and what people are ready, willing and able to do with it. We call closing that gap People-First AI, and it runs through our whole method: Secure. Solve. Scale. Secure the data on a governed lakehouse, solve one problem the business owns, then scale it with the people who will use it every day.

If an AI or ERP initiative is stalling, or you want to keep one from stalling, start with the free AI Flight Recorder. In 30 minutes it scores where you stand on secure, solve and scale and shows the few moves most likely to help. Our process improvement services cover the workflow redesign behind it.

Frequently Asked Questions (FAQs)

What is AI change management?

AI change management is the structured work of helping people adopt an AI-enabled process. It covers readiness checks, executive sponsorship, workflow redesign, manager enablement, role-based training and adoption measurement, so the tool is actually used after go-live.

Why do most AI projects fail to deliver value?

Most stall on people and process rather than technology. BCG attributes about 70% of AI implementation challenges to people and process, 20% to technology and 10% to algorithms, and Gartner names poor change management among its critical GenAI failure points.

How much does an active executive sponsor matter?

A great deal. In Prosci’s research, projects with extremely effective sponsors were 79% likely to meet objectives, compared with 27% for projects with extremely ineffective sponsors.

How do you measure AI adoption?

Measure correct use of the new workflow, quality and cycle-time changes against a baseline, and how often people fall back to old workarounds. Training completion rates alone say little about whether the change worked.

Analytics AIML is an AI performance firm. We rebuild the three foundations that decide whether an AI investment ships, scales, and shows up on the P&L: a sharper problem, a governed data foundation, and demand that survives the zero-click age.

— Rise above the flood

Build a content engine that gets cited.

AIMGrowth is the discipline for the AI-answer economy. We ship it in 90 days, fixed scope.