Skip to content
Nando Peña

Operating model

How teams actually adopt AI

Enterprise AI adoption happens at three levels — individual, workflow, and organizational — and it fails at whichever level gets ignored. This is the model I use to diagnose where adoption is stuck and what to do about it.

LEVEL 01

Individual adoption

Does this person know what AI is for in their job?

Onboarding, training, confidence, habits, support paths, and practical usage.

Everything starts with one person deciding AI is worth their time. That decision is won or lost in small moments: whether onboarding produces a useful result in the first week, whether examples speak to their actual role, whether there's somewhere unembarrassing to ask questions, and whether early friction gets caught before it becomes quiet abandonment.

What real adoption looks like

  • Repeat usage on the same real task — habit, not novelty
  • Questions getting more specific over time
  • People describing AI as part of how they do a task, not a separate thing

The failure mode: licenses distributed, a launch training held, and six weeks later usage has settled to the handful of people who would have figured it out anyway.

LEVEL 02

Workflow adoption

Did the shape of the work actually change?

Process redesign, automation, decision support, knowledge access, handoffs, and daily work integration.

Individual skill doesn't automatically become team capability. Workflow adoption means the process itself changes: a step gets redesigned around AI assistance, a handoff carries better context, a recurring lookup becomes instant, a draft that took an hour takes ten minutes — every time, for everyone in the flow, not just for the enthusiast.

What real adoption looks like

  • A documented step in a real process now works differently
  • Teams describing the workflow change in their own words
  • Output quality less dependent on which person did the task

The failure mode: impressive individual usage that never crosses into the team's actual process — AI as a private productivity hack rather than an operating change.

LEVEL 03

Organizational adoption

Can leadership see it, govern it, and scale it?

Governance, rollout strategy, measurement, leadership reporting, scaling, operating rhythms, and sustainable change.

Workflow wins don't scale by themselves. Organizational adoption is the layer that makes change durable: governance that answers policy and data questions before they become incidents, measurement that tells leadership where behavior actually changed, rollout strategy that sequences teams deliberately, and operating rhythms that keep adoption alive after the novelty fades.

What real adoption looks like

  • Leaders asking for adoption data because a decision depends on it
  • Practices from one team deliberately scaled to others
  • Adoption work with a named owner and a recurring rhythm — not a launch project that ended

The failure mode: pockets of real value that stay pockets — no measurement to find them, no strategy to spread them, no governance to protect them.

Using the model

The levels are a diagnostic, not a sequence.

Organizations don't complete level one and then advance. All three levels run at once, and the practical question is always: which level is the constraint right now? Strong individual enthusiasm with no workflow integration calls for different work than strong workflows with no leadership visibility. The model tells you which kind of work this quarter needs.

To see the model applied, read From AI Access to Operating Capability — or the artifact gallery for the templates that operationalize each level.

Working on enterprise AI adoption?

I'm glad to talk about deployment, enablement, pilots, and what it takes to make AI stick inside real teams.