AI agents for startups
You have a job that keeps eating someone's week: triaging support, chasing invoices, summarizing calls, keeping a CRM honest. You suspect an agent could do it. We build that agent, hand you the keys, and leave.
- Checked the books. 3 invoices are more than 14 days late.
- Drafted 3 reminders in your usual tone.
- Logged why: due dates passed, no reply to the first notice.
- Waiting for your OK before sending anything.
Agents we build
Things the two of us can actually deliver, end to end.
-
Inbox and queue triage
Reads an inbox or a queue and drafts the reply, the ticket, or the record.
-
Call and meeting notes
Turns calls, meetings and notes into summaries and next steps.
-
CRM and spreadsheet upkeep
Keeps a CRM or a spreadsheet current from what happens in email and chat.
-
Watchers
Watches a source (a site, a feed, a folder) and acts when something changes.
-
Inside your tools
Sits inside a tool your team already uses, instead of adding a new one.
-
Something else?
Tell us the job. If it shouldn't be an agent, we'll say so on the first call.
Ask us →
We don't build chatbots for your website, and we don't build anything that pretends to be a person.
How a project runs
- Week 0
A call, then a one-page scope
We find the one job worth automating first. If it isn't worth it, we'll tell you on the call.
- Weeks 1 to 4
A working version
Running on your data, with a human approving every action until you trust it.
- Handover
It's yours
You get the code, the prompts, a runbook in plain English, and a recorded walkthrough. Nothing is locked to us.
- Optional
Upkeep
A monthly retainer if you want us to keep it tuned. Most teams don't need it after the first quarter.
Fixed price, scoped up front
We'll put a number in writing before you commit to anything.
A call, then a written scope with a fixed price, before you commit to anything.
- The one job worth automating first
- The tools it will live in
- A straight answer if it shouldn't be an agent
Fixed price per agent, scoped up front.
- Working version in two to four weeks
- Human approval on every action until you trust it
- Code, prompts, runbook and a recorded walkthrough
Optional. Most teams don't need it after the first quarter.
- We keep the agent tuned
- Same two people who built it
- Only if you want it
Questions founders ask us
What does a first agent cost?
Most first agents land between $500 and $2,000, fixed and scoped up front. We'll put a number in writing before you commit to anything.
How long until it works?
A working version in two to four weeks, running on your data, with a person approving every action until you trust it.
Do we own what you build?
Yes. You get the code, the prompts, a runbook in plain English and a recorded walkthrough. Nothing is locked to us.
Will the agent act on its own?
Not until you say so. It asks before it acts, logs what it did and why, and never sends a message in your name that you haven't seen a sample of.
Do you build chatbots?
No. We don't build chatbots for your website, and we don't build anything that pretends to be a person.
Where are you, and when do you reply?
One of us is in Wisconsin (UTC−6) and one in Nepal (UTC+5:45). We reply within one working day, whichever of us is awake.
Have a job that keeps eating someone's week?
Email us two lines: what job you want off someone's plate, and what tools it lives in. We reply within one working day.