What European IT leaders said when we brought the data to their door

AI ambition is everywhere. The budgets are there. The pilots are running. And yet, for most organizations, measurable results are still not showing up on the balance sheet.

That gap, between what organizations want from AI and what they are actually getting, is exactly what the State of Integration & AI 2026 report set out to investigate. Europe's first independent benchmark of AI adoption, integration maturity and automation readiness, the report surveyed 611 IT and business decision-makers across six countries and nine sectors. The work was conducted independently by Sapio Research.

The next step? Bringing it directly to the IT leaders who are living those numbers every day. Frends hosted local events in five European cities to put country-specific findings in front of the people they are most relevant to. 

What followed were five sharp, candid sessions where senior IT and business leaders took notes, shared their own experience and, in many cases, recognized themselves in the data. 

Denmark: Leading on deployment, racing against complexity

Denmark leads the survey on nearly every AI adoption metric: 44% of organizations have AI in production or widely deployed, 38% take an integration-first approach (nearly double the European average), and the mean AI project success rate of 32.6% is the highest in the study. It also carries the highest annual cost of manual work at €14.08 million per 1,000-employee organization.

The Copenhagen session, featuring AI Customer Advisor Line Dahl Holst and Principal Consultant Søren Helsted from Devoteam, centered on the tension that comes with leading: architectural maturity creates its own complexity. Integration build time in Denmark averages 99 hours — 25% above the European mean — because early digitalization left layers of technical debt that proper integrations have to work through.

Unlike most of Europe, where integration connectivity is the primary barrier to AI impact, Danish organizations cite AI transparency first (39%). The panel was direct about why: commercial LLMs are becoming less transparent over time, not more. Cost structures shift, outputs change, and there is no guarantee that what an agent costs today will be the same next quarter. 

The keynote from Ravish Gopal of Software Improvement Group closed with a governance argument. Keeping the lead requires boards to treat AI as a risk-adjusted investment and ask not just about expected outcomes, but about the actual state of the technical foundation beneath them.

With the high-risk provisions of the EU AI Act coming into effect in August 2026, Denmark's integration-first posture is also a head start on regulatory compliance that less mature markets will have to build from scratch.

"Denmark didn't get here by asking permission. It got here by taking the risk and being early on with those tough integration decisions and the actual willingness to move from a pilot to enterprise-wide introduction."
Ravish Gopal, Solutions & Advisory Director, Software Improvement Group

Finland: The strategy-action gap with good foundations

Finland is in a genuinely unusual position. It has the lowest platform fragmentation rate of any country in the study: just 3% of Finnish organizations run more than five integration platforms. The foundations are clean. The architecture is relatively tidy. By most measures, Finland should be leading the field.

And yet: only 2% of Finnish organizations have deployed AI broadly. The lowest of any country surveyed. Meanwhile, 45% are still stuck in piloting or proof-of-concept, the highest share in the study.

When the Helsinki session put those numbers to a panel that included the COO of Microsoft Finland, Antti Alila, the CIO of MuniFin, Juha Volotinen, and Frends Head of Product, Antti Toivanen, the conversation went straight to the core tension. Finland sees integration as strategic with 44% of Finnish organizations rated it a strategic enabler, the highest figure in the survey. But only 22% have actually built an integration-driven operating model in practice.

The panelists offered a frank diagnosis. Cost is Finland's primary AI blocker, but as one panelist put it, the better question might be: can we afford to stay off the train?

There was also an honest conversation about risk aversion. Finland has better foundations than most, but often does less with them. The panel challenged the room: if the architecture is already in better shape than much of Europe, the missing ingredient is courage, not capability.

The session closed on a practical note. HR onboarding and offboarding came up as Finland's third most common process bottleneck (24%, the highest in the survey) — a high-ROI starting point that does not require a full architectural overhaul to act on.

"Getting the first AI service into production is by far the heaviest lift. You have to build the entire support structure, governance model, monitoring — everything around it. After that first one, the path gets easier."
Antti Alila, COO, Microsoft Finland

Germany: Solving the documentation trap

Germany's headline number from the report is stark: German employees lose an average of 8.5 hours every week to manual tasks. That is more than one full working day per week, or the equivalent of 49 working days per year, gone to routine work that automation could handle.

For a typical 1,000-employee German firm, the annual cost of that manual burden sits at €11.43 million — above the European average. And only 22.2% of German AI projects deliver measurable P&L impact, compared to the European average of 26%.

The session, hosted online with speakers including Professor Dr. Moritz E. Behm from Fresenius University of Applied Sciences and Anna Kopp, CIO at Microsoft Germany, put those numbers in context. Germany's most distinctive finding is that 40% of organizations cite integration complexity as a top barrier to scaling AI — the single highest figure in the entire survey.

Report generation is Germany's top manual bottleneck at 35%, ten percentage points above the European average. The irony is not lost: the country with the strongest documentation culture is also the one where documentation is consuming the most human time.

The session drew a clear opportunity from that: automating reporting pipelines is a direct, visible path to ROI, and one that aligns naturally with Germany's existing governance instincts. The good news is that 69% of German leaders rate governance as critically or very important. That awareness, channelled into building integration as a governance layer, is the foundation from which AI can scale safely and, at the same time, meet the demands of the EU AI Act.

The closing message was practical: stop treating integration as operational maintenance cost. Start treating it as a strategic investment. After all, the architecture gap is what is holding AI back.

"For Germany to lead in the age of AI, we must first fix our digital infrastructure."
Thomas Vianden, Frends

Netherlands: Selling AI internally is harder than building it

On the surface, the Netherlands looks like it should be in great shape. Dutch workers spend the least time on manual tasks of any country in the study, just 6.6 hours per week. Automation of routine work is already mature.

But that efficiency masks a harder problem. Because the easy manual work has already been automated, what remains is business-critical. Only 21% of Dutch AI projects deliver measurable impact — the lowest success rate in the study. The Netherlands is the country that has done the most to clear the ground, but is struggling most to build on it.

The Amsterdam event — which also marked Frends' official Benelux launch — surfaced a finding that made for a sharp panel conversation: 38% of Dutch organizations report significant internal resistance to AI adoption, the highest of any country in the study. For many Dutch IT leaders, the hardest part of an AI initiative is selling it internally.

The panelists — Rob van der Feltz (Group Transformation Lead, Bergman Clinics), Chris Geertsma (CTO, Conclusion Integration) and Tom Desmet (Manager, Tech Harmony) — were direct about why. Governance ownership is unclear. When something goes wrong, who is responsible: the CIO, the business function owner, the AI vendor? The ambiguity slows decisions. Interestingly, Dutch organizations are fast to create dedicated AI leadership roles (Chief AI Officer, Head of AI), suggesting the governance conversation is moving, even if execution is lagging.

The other Dutch outlier: high cost of technology implementation was cited as a primary barrier at a rate significantly above the European average. Qualitative input from Dutch CIOs pointed to legacy systems — some 20 to 30 years old — where the cost of transition goes far beyond the system migration but includes knowledge transfer, reskilling and operational continuity.

Only 12% of Dutch organizations take an integration-first approach. That is the lowest rate in the study, and the direct structural reason why scaling AI keeps stalling. The session ended with a clear call: the architecture needs to come before the ambition, not after the pilot fails.

"When something goes wrong with AI, who owns it? Who is the governing power? It cannot be the CEO. The implementation of processes are across business functions."
Hugo Pereira, Frends

Norway: Saving time with one hand, losing it with the other

Norway produces the most memorable finding in the entire report, and the Oslo session leaned into it directly. Frends CTO Asmo Urpilainen named it in his technical deep-dive: the Norway Paradox.

Norwegian organizations save an average of 22.3 hours per worker per year through AI and workflow automation — the highest figure in the survey. They are fast to adopt, and they see the value. At the same time, Norway has the most fragmented IT landscape in the study: 21% of Norwegian organizations currently run five or more integration platforms simultaneously, double the European average.

The paradox: Norway is saving more time through automation than anyone else, while also spending more time managing the complexity that fragmentation creates. 59% of Norwegian employees still spend seven or more hours per week acting as human middleware, moving data between systems that do not talk to each other. The average annual cost of that for a 1,000-employee Norwegian firm is €12.59 million.

The session, hosted at Microsoft in Oslo with a keynote from Microsoft Norway National Technology Officer Christopher Frenning, was structured around that contradiction. The interactive Q&A with the room surfaced something the data also confirms: Norway's AI project success rate (30.4%) is above the European average. The foundation is strong. The problem is scale.

Asmo's technical session made the 79-hour figure visceral for the room: on average, implementing a single production-ready integration between two systems takes 79 hours. Not because of the code, but because of credentials, firewall configurations, security approvals, system mappings and testing. For an AI initiative that requires 20 system connections, that is 1,600 hours of work before a single prompt is written.

Norway also stood out in a less obvious way: it is the only country in the study where governance concerns outrank technical challenges as the primary barrier to AI. 45% of Norwegian IT leaders cited lack of integration governance as their top obstacle — not a fear of AI itself, but an awareness of what happens when data moves without control. Asmo's recommendation to the Oslo room: count your platforms. If you have five or more, pick one as your core and start consolidating. In the end, the infrastructure gap is more expensive than the AI ambition.

"A brain without a nervous system is just a jar sitting on a shelf. In 2026, our nervous system is the integration layer, and that's becoming paralyzed in this new world of AI."
Asmo Urpilainen, CTO, Frends

The pattern across all five markets

Every city had its own context. But the same structural story showed up everywhere.

AI ambition is high across Europe, and results do not match that ambition. In market after market, the root cause is the same: organizations are trying to scale AI on top of integration infrastructure that was not built to support it.

The 26% P&L impact rate is the single number that anchors every conversation in the roadshow. Three out of four AI projects are not reaching the balance sheet. The report shows why: 36% of organizations cite integration challenges as their top barrier, and only 23% take an integration-first approach. Those that do are consistently ahead on both ROI and adoption speed.

The roadshow made one thing clear that the report itself could only suggest: IT leaders across Europe know this. They understand the technical debt. What they are looking for is a practical path from knowing to doing.

Download the report

The State of Integration & AI 2026 is available here, including country-specific findings for Denmark, Finland, Germany, the Netherlands, Norway and Sweden.

Frequently Asked Questions

What is the State of Integration & AI 2026?

It is Europe's first benchmark of AI adoption, integration maturity and automation readiness among enterprise organizations. The research was conducted by Sapio Research and covers 611 IT and business decision-makers across Denmark, Finland, Germany, the Netherlands, Norway and Sweden, spanning nine industry sectors.

Who commissioned the report?

Frends, a European integration platform (iPaaS) company, commissioned the research. The fieldwork was conducted independently by Sapio Research to ensure objective results.

What is the main finding of the report?

The central finding is that AI ambition across Europe is high, but results are not following. Only 26% of AI projects deliver measurable P&L impact. The primary reason is a structural one: most organizations are attempting to scale AI without first building a solid integration foundation. 36% of respondents cite integration challenges as their top barrier to AI success.

What does the €10.7 million figure refer to?

It is the estimated mean annual cost of unnecessary manual work per organization with 1,000 employees, calculated from the survey data. It represents time lost to tasks that could be automated, like moving data between systems, generating reports, managing onboarding workflows, but are not.

What is an integration-first approach?

An integration-first organization builds connectivity between systems as a shared platform capability, not as a project-by-project response. Instead of wiring systems together for each individual initiative, these organizations create reusable integration assets that any new project — including AI projects — can draw on. The survey found that organizations taking this approach achieve significantly higher AI ROI and faster adoption.

Where can I read the full State of Integration & AI report?

The full report and country-specific findings are available at frends.com.