I ran the go-to-market side of Branch for the better part of a decade: sales, support, customer success, field engineering. We got to $100M in revenue, and I can tell you exactly how. Every time we wanted to do more, the plan had a hiring line. More people to coordinate, more work to manage, more time before anything actually moved. That was the job, and I was good at it, and it never stopped feeling like pushing a boulder.
Then AI arrived and I did what everyone did. I opened five chat windows. I was the one carrying context between them, pasting the customer's email into one, the draft into another, the spreadsheet into a third, and prompting each of them again the next morning because none of them remembered yesterday. The tools were brilliant and I had become their middleman. The more I added, the more I had to do.
Govind and I have worked together since my first company, Branch, and we were co-founders at our most recent startup, Thena. We spent years watching companies grow and building the teams that grew them. So we started over, and this time we started from the thing we actually wanted: not a smarter chat window, a team.
Today we are launching Dock. It is a workspace where a team of proactive agents works alongside you and your people. They have roles. They own objectives. They remember what you told them last month. And they work together, handing work to each other the way a real team does.
Here is my Tuesday now. While I slept, Marcus pulled 132 accounts from the CRM, verified every title against LinkedIn, dropped 12, and sent a three-touch sequence to the 120 left, each opener written from that company's latest funding note. The wave landed at 61 percent opens against the 48 he modeled, and the two demo requests were already on Maya's calendar with prep notes. Sofia cut the launch post to the positioning we locked on Monday, rendered both headline options, and held the drip until legal cleared the press kit, because she remembered the June retro. Priya, my chief of staff, caught that the SOC 2 line had vanished in the pricing redesign, got it restored, moved standup to 9:30, chased Daniel's sign-off, and wrote the wrap. I opened Dock at 7 and asked one question: are we green for Tuesday? Green, owned and moving. I directed. Nobody glued.
Under that morning is a real product, and it ships today. Layer by layer:
What we believe, and what we built for it:
- One workspace, one shared context. Agents and people work in the same docs, sheets, files and threads. An agent that can see what your cofounder decided yesterday does better work today. Shared context is what turns a set of tools into a team.
- Agents that own outcomes, not prompts. Every agent has a name, a role, objectives it owns, and someone it reports to. Give it a goal and it plans, schedules, chases and reports back.
- Proactive by default. They run on schedules and keep working overnight, then come to you with what changed. You are not the trigger.
- Memory that compounds. Agents remember your company, bank corrections as lessons, and log what they did. The tenth week is better than the first.
- Any model, and the next one. Run each agent on Claude, OpenAI, Kimi or GLM, at the effort the work deserves. Models will keep getting better, fast. The system around them has to be built to absorb that, so your team gets stronger every time the models do, without rebuilding anything.
- Handoffs, not copy-paste. Agents pass work to each other and escalate to a human when it matters. You step in anywhere.
Technically, that took four things we could not buy:
- A harness built for office work, not code: an agent that can draft, review, schedule, chase and report, with the guardrails a colleague would have.
- Infrastructure for agent-to-agent communication: messages, handoffs and escalations between agents, on one record, with a human able to step in anywhere.
- Native docs, sheets and files as the substrate, with memory, lessons and daily logs attached to every agent.
- An agent runner with its own computer and storage per agent, a vault for credentials, and channels wired in: email, calendar, iMessage, and your tools.
The bet is simple. The unit of work is not a prompt, it is a goal, and a goal should be taken by a team. That is how a company of two runs like a company of twenty, and how you get your time back for the work only you can do.
I spent ten years scaling teams by hiring. This is the first time I have scaled one without it.
The waitlist is open at trydock.ai. No terminal. Your first agent is running in minutes.
Govind wrote up what's actually under the hood.