When no product fits the job, we build the agent or the app for it.
We build it in your systems, one of your people approves what goes out, and you own the code, the rules and the data.
- Quote 2417 · Sample Ltddraft ready
- Order 4711 · 3 itemsmatches the quote
- Supplier invoice · Marchmatches the order
- Order confirmation 0932delivery date differs
Your back office approves each one
What you have in hand at the end
Agents for the job, a small app for the people who work with them, and the connections to the systems you already run.
If your software can already do the job, we tell you.
A custom build only makes sense where no product does the job. So we start with the trial run or the AI assessment, and we look at what your ERP, your Microsoft 365 or a standard product can already do.
If one of them does the job, or your data can't carry it, we tell you and stop there. We build only what is missing.
Firms with 10 to 250 staff start with the trial run. Larger firms start with the AI assessment.
How the agents and the app work together
The agents do the work between your systems. The app is where your people see the result and decide.
- ERP
- Mailbox
- Files
- CRM
Your agents
Work by your rules
skillspluginsn8n
Approved in your appNothing goes out without a yes
Set up before the first run
- AccessOnly the folders and systems you allow.
- GuardrailsRules the agents work within.
- LogEvery step written down.
Secret files stay on a local model that runs under your firm's name.
Where your agents and apps run
Unless you choose otherwise, your agents run in your own environment and we operate them.
The job: drafting quotes
Your Microsoft 365
In Outlook, Teams and Copilot, where your team already works.
For public and internal files
Your own server
Custom agents, apps and n8n workflows on a server you control.
For confidential files
Private AI
A local model for the files that must not leave the house.
For files that must stay secret
What happens after go-live
You own what we build
We build in your systems and with your people from the first day, so at the end there is nothing on our side to hand over.
If you stop, the agents, the rules and the data stay with you, and so do the app and its code.
We look after it every month
Every month we make the small changes that come up, such as a new price list, a changed rule or an update in a connected system. We check the agents against fixed examples and send you a one-page report.
The monthly fee starts only once the agents pass acceptance on your own past cases.
How we build it
- 01
Check first
The trial run or the AI assessment shows whether the job needs a custom build.
- 02
Build in your systems
We build the agents, the app and the connections with the people who do the job today.
- 03
Acceptance on past cases
The agents work on cases you've already finished, and your team tests the app.
- 04
Go-live
Your team starts working with it, and we stay on.
Who builds it
Your own server runs on AI Admin Panel, software we build ourselves. We publish other tools under the MIT licence, such as an MCP server for Bexio.
On a project, an engineer and a business operator work inside your team, next to the people who handle the cases every day.
Questions about custom agents and apps
Agent, workflow or app: which do we need?
An agent suits work between several systems where each case is a little different, such as reading a request and drafting the quote. A workflow suits fixed steps that run the same way every time, and we build those in n8n. An app gives your people a screen of their own, for example to approve drafts. Many jobs use two of them, and we tell you which after the trial run or the AI assessment.
Who owns what you build, and can we stop?
You own it: the code, the rules and the data. You agree each step on its own and can stop after any of them, and what we have built stays with you.
Who looks after it after go-live?
We do, every month. Small changes, such as a new price list or an update in a connected system, are part of the monthly service. A larger change, such as a new job, is a step of its own that you agree with us first.
Where do our data and the AI model run?
In your Microsoft 365 or on your own server, depending on how sensitive the files are. You choose the model for each job: a local model, your own key or a cloud model.
How is the price decided?
From the AI assessment on, each step has a fixed price and a fixed scope, and we bill no hours. The price depends on the job and on how many systems the agents connect to. We name it in the first call.
Which job is no product built for?
In a 30-minute first call we look at the job with you and tell you whether a product already does it or whether we should build it.
