Anthropic's framework for adopting Claude, taught in their Driving Enterprise Adoption of Claude course, moves through three phases: Activation, Acceleration, and Expansion. It's a sound structure, and one you may already know. There's a phase worth placing in front of it though, and adding it is usually what turns a rollout that works on paper into one that holds up in practice.
The phase is Discovery, and it centers on you and how your work gets done today. With it in front, the four phases give you a clear picture of what a thoughtful rollout looks like. Whether you're designing a rollout yourself or watching one unfold, you can use that picture to build, or to read what's happening and where you stand.
Phase 1: Discovery
Discovery is everything that happens before anyone turns anything on. It's where the organization works out what it's trying to solve, how the work gets done today, and what any solution will need to meet.
Two tracks are worth running here.
The first is the baseline requirements: what you name before you can evaluate anything. Plenty of these are technical, security and data handling especially, and the business requirements belong here too. You already know most of these from running the place. Discovery is where you write them down.
The second is workflow mapping: drawing out the day-to-day processes you already run and finding where Claude fits naturally, rather than picking the tool first and reshaping the work around it. That order is worth protecting: a tool that asks you to change how you work to suit it rarely sticks, and you're the one who knows how the work flows.
Three questions are worth answering before Discovery closes:
- Does this problem need AI at all?
- What is the real gap between how things work now and how they would work with a good solution in place?
- Do you have the infrastructure to put this in place and keep it running well?
For leaders: This is where you set everything downstream on solid ground. When Discovery has happened, the rest of the framework rests on what you learned rather than on what you assumed. For contributors: Your knowledge of how the work really gets done is the most valuable input here. Offer it, and if no one has asked, it's worth raising as a discussion point.
Phase 2: Activation
Activation is the setup phase, and it has two parts.
The technical foundation is the visible one: standing up Claude based on what Discovery surfaced. Permission structures, SSO, seat assignments, connecting Claude to the tools you already use through connectors, and putting guardrails in place. This is the groundwork that makes everything after it possible.
Organizational readiness is the other part, and it deserves equal weight. This is preparing your people: early versions of playbooks (a shared Claude Project is a natural home for them, so the guidance lives where the work happens), internal resources, honest conversations about what AI's role is here, and setting expectations before anyone is asked to meet them. In plain terms, the change management.
For leaders: The move that pays off is treating the human side as carefully as the technical one. When both happen together, people get the tool and know what to do with it, and adoption follows on its own. For contributors: You can usually feel this work happening: clear communication, resources that arrive early, and the sense that someone thought about the people before flipping the switch. When you feel it, trust it; the ground under you is ready for the next phase.
Phase 3: Acceleration
Acceleration is where teams build habits, and it's the phase that decides how far the rest goes. This is where Claude becomes part of how the work gets done.
For leaders: Your job here is to find what's already working and help the rest of the organization do the same. Notice where adoption is strong, understand why, and make room for those people to share it. The playbooks and guidelines from Activation are working drafts at this point, and this is where they get tested and made better by real use.
For contributors: This is the phase to invest in. The foundation is in place, and what you build on top of it now tends to compound. The habits you form here are the ones that carry, and you'll enjoy watching adoption take hold and improve your day-to-day. That holds for any well-executed technology adoption, AI included.
Phase 4: Expansion
Expansion is where what's working in one corner of the organization reaches the rest. New teams, new use cases, deeper integration with the workflows you mapped back in Discovery.
Done well, this phase feels like colleagues sharing what they've been doing and other people wanting in, rather than a second rollout. That's what the first three phases, done with care, set up.
A good signal that you're here: the demand for access comes from the teams themselves rather than from whoever holds the budget.
Putting it in order
Plenty of rollouts start at Activation. Discovery gets folded into a few slide bullets, or set aside because there's pressure to move quickly. The good news is that none of these phases takes months or asks for heavy process; they only need to happen in order.
That order is where the framework earns its keep. When something in a rollout feels off, look at the phase that came before it; that's almost always where the answer sits, and it's usually a quick fix once you find it.
Whether you're designing a rollout or living inside one, you now have the map. You can see which phase you're in, what good looks like at each step, and where to look when something needs attention. That's enough to move with confidence.