How it works
From napkin to
running in weeks.
No discovery decks. No six-month timelines. One call and we know enough to build. Here's exactly what happens.
How it works
No discovery decks. No six-month timelines. One call and we know enough to build. Here's exactly what happens.
Phase 01
Week 1 — ~1 hour of your time
One call. We ask about your operations — what takes time, what gets missed, what you've tried already. You don't need to have it figured out. We find the leverage points.
Discovery call
45–60 minutes. We map your workflows, find what's manual, and identify what an agent could own. You talk, we listen and ask dumb questions on purpose.
Scope doc
We write up exactly what the agent will do, what tools it needs access to, and what success looks like. You approve it or push back. Nothing gets built until we agree.
Output
A clear scope document. What the agent does, what it doesn't, how we'll know it's working.
Phase 02
Weeks 1–2 — minimal input needed
We build. You don't need to be involved much here. We'll ask a few questions, get access to the tools it needs, and send you progress updates. The goal is a working agent in under two weeks.
Agent architecture
We decide how the agent is structured — what it reads, what it writes, how it makes decisions. Custom for your use case, not a template with your logo on it.
Integration setup
We connect the agent to your tools — your inbox, CRM, review platforms, internal systems. We handle the plumbing. You hand over credentials once.
Voice and tone training
If the agent writes or responds on your behalf, we train it on your voice. Feed it examples of your writing. It learns to sound like you, not a support bot.
Output
A working agent in a test environment. You see it do the job before it goes live.
Phase 03
Week 2–3 — a few hours of review
Before anything goes live, you see it work. Real inputs, real outputs, real edge cases. We run it against your actual data in a safe environment and fix what's off.
Staged testing
You throw real scenarios at it. We watch what happens. The goal is to find the three weird edge cases before your customers do.
Your sign-off
Nothing deploys without you saying yes. You approve the behavior, the outputs, the escalation logic. It's your business — you stay in control.
Output
A tested, approved agent ready for production. And confidence it won't embarrass you on day one.
Phase 04
Week 3 onwards — you mostly just get briefed
The agent goes live. We monitor it, handle issues, and improve it over time. You get a brief — daily or weekly, your call — on what it did. You step in when you want to, not because you have to.
Production deployment
Live. Running. Handling the real thing. We stay close for the first week to catch anything unexpected.
Ongoing monitoring
We watch performance, catch failures, and make improvements. On retainer, we run regular cycles to expand what the agent can do.
Regular briefs to you
What it did, what it flagged, where it struggled. You stay informed without being in the weeds.
Output
An agent running your ops. You doing other things.
How much of my time does this take?
About 2–3 hours total during the build. One scoping call, a review of the scope doc, and a sign-off before launch. After that, you mostly just read the briefs.
What if the agent makes a mistake?
It will, at some point. That's why we build escalation rules — anything it's uncertain about goes to you first. We also monitor and fix issues as they come up. The goal is to earn autonomy over time, not assume it on day one.
Do I need technical knowledge?
No. You describe what you need. We handle the rest. If you can explain what you're tired of doing, we can probably automate it.
What tools does it work with?
Most things with an API or email access. Gmail, Slack, Notion, HubSpot, Airtable, Google Business, Stripe, Shopify — we've connected agents to most of the common ones. If you use something niche, ask.
What happens if I want to change it later?
You can. On retainer, changes are part of the deal. On one-time builds, we can scope an update. Agents aren't static — they get better over time if you let them.
Is this really 3 weeks?
For a focused, single-use-case build — yes. More complex builds take longer. We'll tell you honestly during scoping. We'd rather under-promise and ship early than the other way around.
One call is all it takes to find out if we can fix it.