How we work
How do you start with AI in a small company? It depends on where you stand: training or a strategy session, the data foundation, the data pipelines, or a diagnosis with a report. Your IT, in-house or your IT partner, is at the table from day one. This page explains what changes over time, who stays accountable, and what you own.
Over time
Learning accumulates. Automation is earned.
Every correction makes the next proposal better. A step runs on its own only after it has proved reliable, and it can always go back to asking first.
Corrections go into the rules, the methods, and your company dictionary, a written list of what your words and numbers mean. They do not change the model itself, so what the knowledge worker learns stays with you. When the bookkeeper corrects a posting, the correction goes into the rules and the dictionary, so the same mistake is not proposed again. When the CFO rewrites a paragraph in the report commentary, the CFO's tone goes into the rules.
Repeated work shows which steps are reliable: the inputs are stable, the method is known, the exceptions are understood, and someone owns the result. Those steps can move from "asks first" to "acts alone". They are still monitored, they still have a way to handle exceptions, and they can move back to "asks first" if something changes.
How to decide when an AI agent may act alone, step by step
Responsibility and AI literacy
You stay accountable for what your knowledge workers do.
Handing work to AI does not hand the responsibility to the AI or its vendor. A machine cannot be held accountable; a person can. That person is the operator: someone in your company who checks the knowledge workers' work and signs it off.
You carry that responsibility with four things: access that you control, guardrails that hold, people who understand the system, and a logbook of what was done. These four carry it, not a person approving every single output; the law rarely requires that.
- Accountability
- Management remains responsible for satisfactory bookkeeping and controls, whether a person or a program made an entry. Each posting shows who or what made it.
- AI literacy
- Since February 2025, the EU AI Act requires businesses that use AI to take measures to support their staff's AI literacy. We help your people build it, so they carry the responsibility with confidence.
- Decisions about people
- When a decision is made solely by automation and significantly affects a person, GDPR requires meaningful human involvement.
Who is accountable when an AI agent makes an entry, and what the law asks of your books
Sources: EU AI Act Article 4, as amended by Regulation (EU) 2026/1744; GDPR Article 22; Danish Bookkeeping Act and order BEK 205/2024; Companies Act §§115 and 118. Not legal advice.
What you own
You own everything we build for you. It sits in your repository, the folder of code and settings that holds the pipelines, the checks, and the knowledge workers. You also own the builder agent, the AI agent that built all of it and maintains it. The code, the data, and the tools the agents use live in your own cloud project, and you control access.
We build from small parts. Each part is a cloud function or job that deploys unchanged into Google Cloud, Azure, or Fabric. We run each part in our own work first, so it is easier to set up for you.
Because the builder agent is yours, you can maintain everything without us, and you are not locked in. If you stay with us, it is because each new step builds on what we learned in the steps before.
The first step
The first step depends on you.
Some start with training or a strategy session. Some want the data foundation and build their own agents. Some want the data pipelines. Some want a diagnosis of where they stand and a report of where to go. The right first step answers one question: what can you do now that helps you today and prepares you for later?
Whatever you start with, your IT, in-house or your IT partner, is at the table from day one, and the first step leaves something behind: the first pages of the dictionary, the first rules, and the first entries in the logbook. The next step reuses all of it, so it is easier to set up and fits with the first.
- 1
Your IT at the table
Your IT, in-house or your IT partner, joins from day one, because it looks after access to your systems and the cloud project where the knowledge worker's tools and data live.
- 2
You choose the first step
Training, a strategy session, the data foundation, or a job: we help you pick the one that fits where you stand.
- 3
The first pages of the dictionary
We write down what things mean in your company, set the first rules, and agree who gets access to what. The next step reuses this groundwork.
- 4
The job runs, asking first
The knowledge worker proposes; your operator inspects and signs off inside the system where the work is done.
- 5
Corrections become rules
What the operator corrects goes into the rules and the dictionary.
- 6
You own what was built
You own the repository, the pipelines, the knowledge worker and the builder agent. The repository, the data and the tools stay in your own cloud. The next step starts from here.
We prefer an ongoing collaboration, because the learning builds up over time. Scoped work is a fair alternative, and we agree the terms together in conversation.
The ask
Where do you stand?
A conversation about where you stand, and what your first step could be.