Rewind is becoming an agentic company
Sally reviewed three NDAs at Rewind last week. She doesn’t have a desk.
Sally is our legal agent. Named after the lawyer in Cars. Competent, no-nonsense, gets things done. She reviews contracts, flags concerns, and summarizes what matters. She knows our standard terms, our deal history, and how we’ve handled unusual clauses before.
The tenth contract she reviews is better than the first because she’s learned from every one.
Most people hear “AI company” and picture a smarter version of what they already use. Better autocomplete. A meeting note-taker. A chatbot that answers HR questions. Useful, sure. But that’s not what I mean.
An agentic company builds agents: AI with a job, memory, context, and the ability to improve through repetition.
There’s a version of this that looks impressive but doesn’t really compound. You open a general-purpose tool, explain what you need, get an answer, and close the tab. Next time you open it, you’re starting over. No memory of you, your company, or the work you’ve done together.
An agent has a job. Sally isn’t a legal tool I occasionally use when a contract shows up. She’s a legal specialist with a track record. She accumulates what she’s learned. She gets better at our work specifically, not just at contracts in general.
Every morning before I’m awake, another agent pulls my calendar, scans my inbox, and sends me a brief. By the time I check my phone, it’s already there.
We have a competitive intelligence agent that monitors our competitors every week for product changes, pricing updates, and new messaging, then files a report in Confluence. Nobody has to remember to do it. It just happens.
I have a coaching agent that scores my leadership across six dimensions every day. A community agent that monitors forums for backup and data loss discussions. A writer agent that scouts articles and incidents relevant to our market.
These aren’t one-off prompts. They have jobs. They have tools. They have memory that builds over time. The longer they run, the better they get.
That’s what I mean by an agentic company. Not a company that uses AI here and there. A company that builds agents into how it operates.
Is any of this real, or is it mostly hype? It’s real. And it’s less complicated to set up than most people think.
The hard part isn’t getting a model to do something useful once. The hard part is giving it a real job, enough context to do it well, and a feedback loop so it improves.
I think a lot of companies will go this way. Rewind intends to get there early.
I’ll be documenting our transition here as we go. If you want to watch us build an agentic company, subscribe for email updates.