AI in integrations, legacy risks and lessons from the public sector — what we heard in Jyväskylä

Public sector IT and digitalization professionals gathered in Jyväskylä, Finland, last week for another Automation Forum. The morning program covered topics like integrations, automation and, of course, AI. But what made it memorable was the quality of the conversations.

Here are three things that kept coming up.

1. The single-person dependency problem is more common than anyone admits

Niklas Baarman, Account Executive at Frends, opened the event with a story that landed hard: an organization had 27 years of integrations with no documentation, apart from what lay in their creator's own head. It's an extreme case. But the room recognized it immediately: the one-person-holds-all-the-knowledge example.

Though common, situations like this offer high organizational risk. When all integration knowledge lives in one person's head, you're one resignation — or one sick leave — away from a serious problem. And in the public sector, where resources are already stretched, that risk is rarely acknowledged until it becomes a crisis.

A modern integration platform doesn't just make integrations faster to build. It makes them visible, documented and shareable by design. When integrations are built visually using a standard like BPMN, a new team member can look at a process and understand what it does, without needing a 27-year briefing.

That's not a small thing.

2. AI belongs in integrations, but only where it earns its place

Solution architect Miika Länsi-Seppänen's session was the most interactive of the morning, and the most honest. The message wasn't "use AI everywhere." It was: use AI where data is unstructured and human judgment is currently the bottleneck.

The clearest example: email triage. When a single inbox receives everything from urgent system outages to thank-you notes, someone has to read every message and decide what to do with it. That's a task where AI genuinely helps. It is far from impressive, but it does what AI should do: add value. After all, the alternative is a person having to spend their day working on something that a model could do in seconds.

The same logic applies to invoice processing. Suppliers send invoices in their own formats. No two look the same. Extracting the right fields, matching them to the right customer and routing them to the right system are exactly the kind of unstructured, variable work where AI adds real value inside an integration flow.

The room pushed back in the right ways. What about prompt injection? What about token costs spiralling out of control? What about organizations that adopt AI not because it solves a problem, but because leadership decided it was time?

All fair. All discussed. The point is AI works best when it's introduced into a specific, well-defined step in a process, instead of being dropped in as a general-purpose fix. And when it's part of an integration platform, you get something important: transparency. You can see why the model made the decision it made. That matters, especially in the public sector.

3. Real results from a real customer

Suvi Hänninen, Information Technology Manager at the City of Kemi, shared what a cloud migration actually looks like when it's done gradually and deliberately.

Not a big bang. Not a crisis-driven scramble. A structured shift, supported by a central integration platform, that has made new integrations 30% faster to onboard and kept costs predictable from year to year.

A few details that resonated with the room:

• In a stable environment, maintaining the Frends platform takes roughly 0.5 FTE per month. That's a number people can take back to their own organizations and actually use in a budget conversation.

• The pricing has been transparent and consistent: what was quoted is what was charged. In a world of hidden costs and scope creep, that's worth saying out loud.

• Data stays within the EU, and the platform is GDPR-compliant by design. For public sector organizations navigating data sovereignty questions, that's not a nice-to-have.

The City of Kemi started with Frends in 2019. The relationship has scaled as the environment grew. That's the model: start small, build visibility, scale when you're ready.

The conversation continued over lunch

That is usually the sign of a good morning.

If any of this resonates with your own situation — whether you're dealing with legacy systems, thinking about AI in your integration flows or just trying to understand what a modern integration environment looks like in practice — the Frends team is happy to continue the discussion.

Reach out to Niklas!