AI automation for companies from audit to running

I can, but first we check that you need one. A website chatbot makes sense when the same questions keep coming in (availability, shipping, order status) and you have documentation it can draw on, otherwise it will just hallucinate. Built on your own materials it prices like a regular pilot, from €2,400 including integration, testing on real questions and staff training. Often, though, the audit shows that automating email brings more savings than a widget on the website, and I'll tell you that upfront.

The honest answer: nobody knows upfront, and anyone quoting you a percentage before the audit is guessing. That's why the pilot starts by measuring what the process costs in time and errors today, and measures again after deployment using the same method. The numbers you'll see come from your operation, not from a deck. If they come out badly, you'll hear it from me first.

Often more than for large ones, a small team feels every saved hour immediately and has no ten rounds of approvals. For companies up to roughly 10 people a full audit usually isn't worth paying for: we go through it on the opening call, and if it's clear where to start we go straight to the pilot and you save the €1,000 audit fee.

A completely legitimate requirement, and a routine one. Depending on sensitivity I choose between a business tier on a cloud model (where your inputs aren't used for training) and a model running locally on your server, where data never leaves the company. For personal data that comes with a processing agreement and written limits on which agendas never reach AI at all.

I work with the goal of taking routine away from people, not taking people away from the company, at this size the work is typically retyping, sorting and searching documents, not somebody's job. I deliberately don't automate HR decisions (candidate screening, employee evaluation or scoring): the EU AI Act classifies that as high-risk use with separate obligations, and for an ordinary company it isn't worth it.

It depends on the task. n8n or Make for wiring tools together, OpenAI, Anthropic or Google models for language work, local models (Ollama) where data must not leave, and custom code when off-the-shelf tools fall short. I take no commission from any vendor, so I choose on fit and running cost, and the proposal always says why.

It goes into the audit and the proposal so it doesn't surprise you on an invoice. For a typical smaller process, licences and model usage run in the hundreds to low thousands of crowns a month; for high-volume agendas we price it exactly from the volume measured in the audit. You pay providers directly on your own accounts, it doesn't go through me with a markup.

Since 2 February 2025 the EU AI Act requires companies to ensure sufficient AI literacy among people who use AI at work, and from 2 August 2026 transparency rules apply to AI-generated content. For ordinary office automation that mainly means training and written internal rules, both are part of the delivery. I'm not a lawyer; for sensitive agendas I recommend having the setup reviewed legally.

Training teaches your team to work with AI themselves. This is done-for-you: I deploy the process and hand it over working. Most companies need both, which is why training the affected people is part of every deployment. If you'd rather start gently, AI training for companies first and implementation afterwards is a sensible order.

Then you belong in custom app development, a booking system, a customer portal, orders wired into accounting. Here I handle AI in operational routine on tools you already have. The opening call tells the two apart quickly, and I'll point you to whichever comes out cheaper, not to whichever is the bigger job for me.