Almost every company now "uses AI." Very few run on it.
Anthropic published a five-level model for rolling out Claude across a company. I adapted it so it works for any AI tool your team uses:
Level 0, Ask. People chat with AI for answers and drafts.
Level 1, Assist. AI works from your own files and tools to build something new.
Level 2, Repeat. Your best workflows become saved skills that run the same way every time.
Level 3, Run. Those workflows run on schedules and triggers, without anyone remembering to start them.
Level 4, Operate. Each team runs its own set of governed agents, with an owner and a log.
The jump that pays is usually from 1 to 3: pick two or three workflows, make them repeatable, then let them run.
Anthropic's original, for reference: Deploying Claude across your organization: How Anthropic uses Claude Cowork
What level is your team at right now?
Chase Parks writes about putting AI to work at WorkSpec AI.