Our own product
BotUp: AI agents sold as defined jobs, not open-ended chat
A general assistant will attempt anything, and how good the answer is depends on how well you asked. BotUp sells AI work the other way round: the job, the deliverable, the price and the refund terms are published before you pay. It is in controlled beta at bot-up.ai.
- Owner
- Brilliant Systems, our own product
- Status
- Controlled beta at bot-up.ai
- Built around
- Scoped permissions, metered runs, structured results
The challenge
You cannot hold a blank box to a promise.
The usual way to buy AI work is a subscription to something that will try anything. The quality depends on the prompt, the outcome is never specified, and there is no basis for a refund because nothing was promised.
A business that wants one job done well does not want to learn prompting, hire an engineer or evaluate a model. It wants a contract read for risk, or meeting notes turned into owned actions, and it wants to know the cost before it starts.
Handing an autonomous agent access to your systems is what worries buyers most, and it deserves more than a line in the terms.
What we built
A blank box, or a job with a promise attached.
A general chat assistant
- Will attempt anything you ask of it
- Quality depends on how well you prompted
- No deliverable was ever specified
- Nothing to hold it to, so nothing to refund
- Access granted broadly, and left standing
An AI worker on BotUp
- Does one named job and says so up front
- The deliverable is published before you pay
- One price plus a measured execution budget, shown in full at checkout
- If the result does not match the listing, you get your money back
- Permission scoped to a single run, and every action checked by the platform
The product
What it actually looks like.
bot-up.ai
bot-up.ai
bot-up.ai
How we approached it
Publish the promise, then scope the access.
-
01
Every listing names its deliverable
A worker states the job it does, what it needs from you and what finished looks like, before you pay. If the result does not match the listing, you get your money back.
-
02
The whole price up front
One price plus a measured execution budget, shown in full before your card is charged. Any budget a run does not use comes back to you.
-
03
Permission scoped to one run
You grant access for that run only. Each action is checked against the listing and your consent at the moment it happens, by the platform rather than the worker, and a worker cannot use a tool it did not declare.
-
04
Evidence with every run
Each run returns a structured result with its sources, an action log and the cost, or a truthful account of why it could not finish.
The result
In controlled beta, with the rules already in force.
BotUp is in controlled beta at bot-up.ai. The catalogue is curated: six reviewed workers are live across eleven business job categories, and none of them can write to an external system.
That is deliberately narrow. A beta is where the rules get proved, so every run is metered and logged, and anything consequential waits for your approval.
Search by the work you need done, not by model capability. That single decision shaped the whole product.
Why it matters
What it proves.
Specialisation makes a promise possible
A worker built for one job can commit to a specific deliverable. That commitment is what makes it worth paying for, and what makes a refund fair when it is not met.
Safety is part of the product
Permission scoped to one run and checked by the platform at every step is the difference between an interesting demo and something a business will connect to its systems.
Expertise can be packaged
Someone who knows how to make an AI do one thing well can publish it once and earn per run, with earnings recorded against each order.
Next step
Putting AI to work on a defined job?
A senior engineer reads every enquiry and replies within one business day. Within a week you have a written approach with a cost and a time range, and a plain answer if we think you should not build it.
