Change Capacity

Change capacity is what most leaders skip past when greenlighting an AI initiative. They focus on the strategy and the tools, then assume the org will absorb whatever comes next. It usually doesn't.

AI is unusual because it doesn't move at the pace of a normal transformation. The tooling shifts every few weeks. New capabilities appear. Old ones break. The change isn't a one-time rollout, it's a constant low-grade churn that every team has to absorb on top of their existing work.

Most organizations don't have the change muscle for that. They have change processes designed for slower, bigger, less frequent transitions. AI adoption needs something different: faster feedback loops, more direct paths to surface concerns, and an honest acknowledgment that people need time and space to learn, not just access.

There's also a quieter dynamic at play. People are afraid. Of being replaced, of accidentally doing something they shouldn't, of falling behind, of admitting they don't know how to use the tools they've been given. If that fear isn't named and held, even great change management won't land.

It rarely looks like "we don't do change management." Usually it looks like one of these:

• Leadership procures a stack of AI tools (ChatGPT, Claude, Perplexity, Cursor, Copilot) with no clear mandate or experimental scope. People are told to "play with it" but nobody defines what success looks like or how decisions will get made.
• Leadership doesn't actually know what tools teams are already using or what they're doing with them. The bottom-up adoption is happening, but it's invisible to the people approving the strategy.
• MCP integrations and security reviews take far longer than anyone planned for, and the rollout stalls in approvals.
• People feel like they're being asked to learn AI on top of their full workload, with no time built in for the learning itself.
• The first person to raise a concern about an AI rollout gets quietly sidelined, which sends a signal to everyone else.

If any of these are familiar, your score here is lower than it looks.

Acknowledge the experimentation problem, then close it.

"Spray and pray" works for a short window, but only if you've defined what you're testing, what success looks like, and when a decision will be made. If teams are experimenting with multiple AI tools, set a clear timeframe, a clear feedback loop, and a clear point where you'll commit. Otherwise everyone picks the tool they're personally most comfortable with and you get a fragmented stack with no shared rationale.

Find out what people are already using.

Before deciding on the future state, do a real audit of current AI use across the org. Not what was procured, what is actually being used and how. This usually surfaces both risks (people using personal accounts on sensitive data) and signals (specific workflows where AI is already producing value).

Build in the time for learning.

If you want people to use AI well, they need protected time to learn. Not optional lunch-and-learns. Actual hours in their week. If leadership won't approve the time, they shouldn't expect the adoption.

Name the fear out loud.

People are worried about being replaced, about exposing sensitive data by accident, about falling behind. Leaders who don't address these openly leave their teams to interpret silence as confirmation. Be specific: how is AI going to affect roles? What are the guardrails for sensitive data? What's the support for people who feel behind?

• Leadership can describe what AI tools the org is actually using right now, not just what's been procured.
• There's a clear experimentation window with a defined endpoint and decision criteria.
• People have time on their calendars for AI learning that isn't taken from their existing workload.
• The first person who raises a concern about an AI rollout gets thanked, not sidelined.
• Conversations about AI's impact on roles are happening directly and openly, not just whispered.

• Treating procurement as adoption. Buying the tools is the easy part. Getting people to use them well, consistently, and in alignment with the broader system is where the real work is.
• Underestimating the approvals timeline. MCP integrations and security reviews take time. If you've planned a six-week rollout, you probably need six months. Plan for the slowdown.
• Mandating tools without mandate clarity. Telling people to "use AI more" without defining what good use looks like creates pressure without direction. People will either burn out trying or quietly opt out.
• Confusing enthusiasm with capacity. Early adopters and AI enthusiasts will absorb change easily. They are not representative of the org. Build your change plan for the people who are skeptical, busy, or quietly worried, not for the cheerleaders.

Change Capacity is the dimension that determines whether everything else actually lands. You can have Strategic Clarity, well-mapped Operating Model Fit, clean Data & Infrastructure, and thoughtful Education & Enablement, and still fail here. If people don't have the time, safety, or psychological permission to change how they work, the rest of the system stays theoretical.

Conversely, strong change capacity won't compensate for weakness in the other dimensions, but it does buy you the time and trust to address them.