Owner-operators and small teams
An hour every morning,
before the work starts.
Three or four of your actual daily jobs, written down and running on a schedule. One page shows what ran, what it produced, and what is waiting for your approval.
$1,800 to build · + $400/mo to run it
- doneinbox triaged, the three that matter pulled out✓
- donethe overnight numbers pulled and checked✓
- donetwo replies drafted, not sent✓
- refusedspend cap reached — stopped before the call—
- waitingthree drafts waiting for your click—
Illustrative. The steps are the real sequence; no figures are claimed.
What it is
The first hour of your day, back — without anything sending itself.
We take three or four of the jobs you do before the real work starts — harvesting, sorting, drafting, chasing — write each one down, and run them on a schedule. One page then shows what ran, what it produced, and what is waiting for your click. It drafts. You send. Nothing leaves without a person.
An hour a day is roughly 250 hours a year. That is the whole argument, and it is yours to check rather than ours to claim — we ask what you do every morning and how long it takes, and if the answer is twenty minutes we will say so on the call.
The frame is built and tested. Your workflows are not. We run our own version daily — ours, not a client’s. Yours is three weeks from signature.
The system
Your morning, while you are asleep.
Five things happen before you open anything. Two of them are the system refusing to act.
-
01
The inbox is triaged 06:30
Sorted against written rules you can read, not against a personality. The three that matter are pulled out; the rest stay where they are.
-
02
The overnight numbers are pulled 06:35
From the systems you already use. Every run is written to an append-only log, because a run that found nothing and a run that died must not look identical.
-
03
Two replies are drafted 06:50
Drafted, never sent. There is no send tool reachable from any routine. A person clicks send, every time, and that is what makes this safe to point at your inbox.
-
04
The spend cap refuses 07:00
It sums the month and exits before the request, not after the response. A refusal, not a warning — a cap that checks afterwards has already spent the money.
-
05
You open one page Whenever
What ran, what it produced, and what is waiting for your click. Not a terminal, and not a chat log.
Four conditions, and none of them are preferences
One, you hold your own Anthropic account and pay them directly — the terms require it, and it means the system is yours if you ever stop working with us. Running it on ours is prohibited outright and no price makes that version sellable. Two, your AI spend is yours, typically $20–80 a month, capped in code so it stops rather than overspends. Three, it drafts your email, it never sends it; a person clicks send. Four, it can write files, so what we are and are not responsible for goes in the quote in writing.
Delivery
Twenty-one days. Day one is watching, not asking.
Day 1
We watch you work
We do not ask you to describe the process. People describe the one they think they follow, not the one they actually follow.
Days 2–20
Built one job at a time
Each workflow tested before the next is started, then the routines, the logging, the dashboard, and the fences.
Day 21
Handover
Two training sessions and a one-page how-it-works. You must be able to run it without us.
We do not go live on an untested path. However tight the date, every path is tested end to end before anything reaches you.
Pricing
One build. One monthly. Your AI account.
Your Claude usage is billed to you by Anthropic and sits outside this price.
$1,800
to build and install
$400
per month to run it
Build fee due at signature; the monthly starts at go-live, not at signature. It covers monitoring, fixes when something upstream changes, documentation, and up to two hours of changes a month. Minimum three months. Your own AI spend, typically $20–80 a month, is paid to Anthropic directly.
Included
- Three to four workflows — your actual daily jobs, each written down and testable
- Two scheduled routines, connectors, and append-only logging
- A spend cap enforced in code, and a scoped tool allowlist
- Draft-only mail — no send tool is reachable from any routine
- A dashboard, two training sessions, and the full source. It is yours.
Not included
- Anything touching medical, legal or financial records
- Sending anything on your behalf without a human click
- Access to systems you do not own or administer
- Replacing a member of staff
- Any guarantee of hours saved or output produced
We guarantee the system runs, the fences hold, and every run is logged. Never a productivity figure.
If what you need falls outside a listed scope it is a separate quote, not a stretched package. Send the details to info@adapton.io.
Questions
The ones worth asking.
Why do I need my own AI account?
Because the terms require it, and because it means the system is yours. If you stop working with us tomorrow, it keeps running on your account.
What does the AI cost me?
Twenty to eighty dollars a month depending on volume, paid to Anthropic directly. We put a hard cap in the code — it stops rather than overspends.
Can it delete my files?
It can write files, so in principle yes. That is why destructive steps need your click, mail is draft-only with no send tool reachable, and it can only run the commands the job actually needs. All of that is in writing.
Will it replace my assistant?
No. It removes tasks, not people. If you are hoping for headcount savings, this is not it.
Next step
Fifteen minutes. Then you will know.
We ask three questions and do the arithmetic with your real numbers. If it is not worth building we will say so on the call.
Send us the detailsOr email info@adapton.io