Why we now talk about AI in processes, not AI in products

For most of our history, companies came to us with an idea for a product. A mobile app, a web platform, a connected device and the software around it, Bluetooth and UWB localisation, asset tracking. We designed and built those products, and we still know how to do it well.

This year the questions we hear have changed. Fewer people ask us to build something new. More ask how AI will change the way their company works day to day: how calls and enquiries are answered, how offers are prepared, how documents are processed, how new enquiries are followed up. At the end of August our management board drew the conclusion: from now on, the first thing we talk about is AI in processes, not AI in products.

Why the question changed

Everyone knows ChatGPT. Few companies have managed to put AI to work in processes that already run, so that it brings value week after week and not only in a demo. The models are good enough and widely available. The bottleneck is no longer the model; it is the process around it and the integration with the systems a company already uses.

We saw the same thing in our own work. When we moved from AI assistants to agents, the real gain came not from a better model but from putting the work in order: clear tasks, defined checks, agreed responsibilities. The same is true in any company. AI added on the side becomes one more tool nobody opens. AI built into the way work flows takes routine off people.

What this looked like in practice this summer

The change did not start with a strategy document. It started with projects:

• A voicebot that calls every enquiry back. We connected online forms, the CRM and an AI voice agent that calls each new enquiry back while the person is still interested, including outside office hours. Its privacy notice was prepared with the client before regular operation.
• A multilingual first line for customer support. For a connected-vehicle service in Europe, a voice agent answers basic questions in the caller's language and opens a ticket in the CRM when a person needs to follow up. It has been on the phone lines since early summer.
• Offers from supplier catalogues. Last week we sat down, over coffee in their office, with a trading company whose salespeople put together proposals by hand from several suppliers' catalogues and price lists. We agreed to start with an assistant that drafts the proposal and leaves the final word to the salesperson.

None of these is a new product. Each is a step in a process the company already runs, now done faster and more consistently. In August we also started calling companies we have known for years, not with a list of technologies, but with one question: which part of your week would you most like to give away?

Where AI pays off in a company

Most of the time there is no need to build AI from scratch. It is enough to connect the systems, SaaS tools and AI components a company already has in a smart way. What we most often look at:

• Voicebots as the first line: callbacks, incoming requests and first classification, handled around the clock.
• B2B quoting: offers prepared from the product catalogue and price lists in minutes, not days.
• Reading documents: data from invoices, orders and forms extracted and passed to the right system.
• Customer service: an assistant connected to the ERP that answers with real order and stock data.
• Knowledge assistants: answers based on company documents, procedures and product knowledge.

Existing systems, data and equipment stay where they are. We build on them.

How we work: from process review to maintenance

We start with the process, not with a model or a vendor. Our AI process automation work has four steps:

1. We study your processes. Our analyst sits with the people who do the work and maps the processes that take the most time. It is a short engagement with a fixed scope, and you decide what happens next. You get a map of your key processes, a list of places where AI pays back fastest, an estimate of the effect in hours and money, and a recommended first step.
2. We show where you gain. From that list we choose a short set of automations with the quickest return. It is better to start with one process that clearly works than with five that half work.
3. We design and implement the integrations. We connect the systems, SaaS tools and AI components you already have, and deploy the solution into daily work.
4. We maintain and measure. Every automation has a target effect agreed up front, and we check it after launch. Monitoring, alerts and a pool of support hours keep the solution running as the business changes.

Responsible from the first workshop

Almost every automation touches data about clients or employees. We design with GDPR and the EU AI Act in mind from the first workshop, not after launch: personal data processed only where it is needed, clear information for people who talk to a bot, human oversight over decisions that matter, and documentation that supports the client's own compliance. The callback voicebot above waited for its legal review before it went into regular use, and that was the right order.

What has not changed

We have been building and integrating systems for well over a decade. Apps, backends, devices, integrations with ERP and CRM systems: that experience is exactly what AI in processes needs, because most of the work is making systems talk to each other reliably. Our IoT background helps too. Where machines already produce data, AI can spot failures before they stop production.

AI is another tool in a well-organised workshop. What changed is where we point it: less at new products, more at the everyday work of the companies we work with.

If you would like to see where AI could save time in your company, read more about AI process automation or book a conversation with us.