I could not clone my mother, so I built her as an AI Coworker in Slack and Teams

*The origin story
*
I ran a marketing agency for ten years. Performance, content, brand, lead gen, the whole spread. At the peak I was juggling fifteen clients at once, walking into meetings with no time to prep, holding it together on instinct.

The client work was never the problem. The problem was everything around the work. Reporting deadlines. Contracts. Recruiting. Finance. Checking the accounts. The endless admin that has nothing to do with the client and everything to do with the day.

When I decided to build again, I knew I could not stand doing that part twice. So I had a coffee with my mother and told her the truth: I will only do this if you run the back office.

She said yes.

Her name is Patricia. People call her Pati. She is sixty-five, and she is the most quietly capable person I know. For years she took care of whatever needed doing, usually before anyone asked. Decades of judgment, the calm of having seen it all, and the instinct to fix things before they became problems. The team did not really run without her.

Most teams have someone like her. The ones that do not wish they did.

I could not give everyone my mother. So we built one anyone can have, and named it after her: Patricia.

What Patricia is

Patricia is an AI coworker for startups and small teams that run on Slack (Microsoft Teams is rolling out next). Not a chatbot you visit in a separate tab. A teammate that joins your workspace like a new hire, and anyone on the team can @mention.

Five things make her different from "ChatGPT but in Slack":

1. Finished work, not suggestions. She hands back the post, the report, the outreach list, ready to ship. Not a bullet list of ideas for you to execute.
2. A brain that remembers. She holds your product, your voice, your goals. You never re-explain your company in every thread.
3. An approval gate on everything that leaves the building. Nothing publishes, sends, or spends without an explicit yes, and every action lands in an audit log.
4. Proactive by default. She watches the channels you invite her to and raises the work you would forget, before you ask.
5. She works on a schedule. Monday recaps, Friday reports, month-end invoice chasing, on their own.

Today she handles Google Ads and Meta ad management, content and social, email and lifecycle, reporting from connected tools like GA4 and Search Console, research and lead lists, sales follow-ups, and the finance chores nobody loves, like chasing invoices.

How we got here

We did not start with a product. We started with a pattern we kept seeing in small teams and agencies: one person on the team is AI-savvy and drowning in prompts, and everyone else has not adopted AI at all. The tools are there, the adoption is not, because every AI tool is another tab, another login, another thing to learn.

The bet we made: the interface already exists. It is the chat tool the team lives in all day. If the AI shows up there, as a coworker instead of a product, adoption stops being a rollout project. You @mention her like you would a colleague, and that is the entire learning curve.

Getting there meant solving the unglamorous parts:

• Trust before capability. An AI that can spend your ad budget or email your customers needs a hard approval gate, not a disclaimer. We built the gate first and the capabilities behind it.
• Memory as a first-class feature. A teammate who forgets everything between conversations is not a teammate. Patricia keeps a company brain so context compounds instead of resetting.
• Real integrations, not demos. Deep, native connections to the tools where the work actually happens: Google Ads, Meta, GA4, Search Console, HubSpot, Stripe, Mercury, and a long tail of others.

For the builders here: the backend is Python and FastAPI on Postgres with pgvector, agent tasks run in isolated E2B sandboxes, and integrations run through Composio plus native adapters where depth matters. Happy to go deeper on the architecture in a follow-up post if people are interested.

Why we are building it

Because small teams are built on talent and starved for time. The founder doing invoices at 11pm is not doing the thing they started the company for. The work around the work eats the work.

Big companies solve this with headcount. Small teams cannot, so the admin lands on whoever cares the most, and it quietly burns them out. We think that layer of work should just be handled, by something with judgment, memory, and the good sense to ask before acting.

That is what Pati did for us every single day. That is what we are building for everyone else.

Everybody deserves a Patricia.