AI for small businesses. Built to ship.
Small business is who we work with: 15 to 250 people, no in-house AI team, allergic to slideware. Working systems in the tools your team already uses, handed over so you don't need us forever.
What we hear most
Pulled from real conversations with small-business operators. The phrases buyers actually use.
We've been asked what's our AI plan and don't have a confident answer.
We start with a 2–3 week Advisory pass: a prioritized list of use cases, a recommended first system, and a buy/build/wait call on each.
We tried ChatGPT, but it didn't really stick.
Public LLM in a separate tab is a starting point, not a system. We build into the tools your team already uses, with the data they actually need.
I just need it to work.
Working AI systems in production. Source code, evals, runbook, kill switch. Handover so your team can run it.
I don't want a six-month strategy doc.
We don't sell those. Strategy and implementation are the same engagement, the same team, the same room.
I just want something in our Slack that actually works.
That's Workplace Agent: an AI employee in your Slack, configured channel by channel. Coming soon; tell us you're interested. The bigger build comes later if it makes sense.
We can't afford a Big-4 firm, and wouldn't trust them anyway.
Small enough to move fast, deep enough to do it well. Same person from kickoff to handover. No account team, no analyst layer.
Where most engagements start
Two starting points cover most small-business engagements. Pick the one that fits where you are.
Services
When the workflow is named and you want a working system built into your stack. Strategy, software, and enablement in one engagement. 8–12 weeks, fixed-fee.
See the Services processWorkplace Agent
When you want something useful in your Slack. An AI employee configured channel by channel, with its own purpose, persona, and workflows. Coming soon.
See Workplace AgentAlready running Claude Desktop or another MCP-compatible AI client? MCP Manager is the third route: a hosted MCP server with managed OAuth across the SaaS tools your team already uses.
What an engagement looks like at this size
For a 15–250-person company, a typical first engagement runs 8–12 weeks, fixed-fee. One contract, one team, the same people from discovery to handover. No analyst layer between you and the work.
Every engagement uses some blend of four capability areas. Most use two or three. Some use all four. Pick the one closest to where you are. We'll scope the rest.
What this looks like in practice
Three representative engagements pulled from the playbooks. Same engagement model, different workflows.
Problem
45-person services company with a 12-person support team needed AI-assisted ticket triage and response drafting.
What we built
8-week build pulling from Zendesk, KB, and product docs. Drafts replies for human review, logs every model call.
Outcome
Agents review drafts instead of writing from scratch. Handle time drops; quality stays flat or improves.
Problem
60-person services co ran month-end close in a flurry of Slack DMs. Close slipped 2 days every quarter; the Controller spent 3 of those chasing status.
What we built
8-week build into QuickBooks + bank feed. A #month-end-close channel posts daily status with outstanding items, owners, due dates.
Outcome
Close lands on time. The Controller gets a third of close week back. The tracking spreadsheet stopped getting opened.
Problem
25-person services company had an AI summarization tool, a CRM, and a support tool. None of which talked to each other.
What we built
4-week integration. Call summaries land in the CRM. Tickets get auto-tagged. One timeline across three systems.
Outcome
Same people, same work. They stopped copying and pasting between five tabs.
Tell us what you're working on.
One sentence is fine. What's the workflow, what's stuck, what would good look like?