In late 2025, BCG published Nordic AI research with a striking pair of numbers. Around 70 percent of Nordic enterprises have meaningful AI investment in motion. Only 4 percent report meaningful, measurable business value from that investment.
That 17-to-1 gap between activity and value is not a Nordic-only story. It is the global mid-market pattern, expressed in the Nordic operating context. The reasons are not what most of the consulting press suggests. They are structural. And they are fixable in months, not years.
What the data actually says
The relevant numbers from BCG and supporting research.
- 70 percent of Nordic enterprises report meaningful AI spend (BCG, 2025)
- 4 percent of organisations capture meaningful business value from AI initiatives (BCG, 2025)
- 13 percent of Nordic enterprises have a CAIO or equivalent senior AI function (BCG, 2025)
- 32 percent of Nordic AI activity is owned by the CIO function, often as a technology workstream rather than a commercial one (BCG, 2025)
- 74 percent of executives believe their AI governance is adequate. Only 3 of 9 facets of AI governance are rated as strong by their own teams (EY, 2025)
- DKK 800M Danish government investment commitment in AI through 2030
The activity is real. The capability investment is real. The value capture is the bottleneck.
Where the value leaks out
Across BRAGI client engagements and the wider BCG / McKinsey / Cisco / NIST research base, the same five structural causes show up.
1. AI activity is not connected to commercial outcomes
Most AI workstreams are scoped against technology metrics (tools deployed, pilots launched, models trained) rather than commercial outcomes (revenue lifted, cost reduced, speed gained, risk avoided, capability built). The work is real. The accountability is misshapen. When a project is sponsored against "deploy GenAI assistant for sales team" instead of "lift qualified pipeline by 15 percent in two quarters", there is no anchor for honest assessment.
2. The pilot-to-production gap is structural
McKinsey's 2025 research showed organisations average 2.3 out of 4 on AI maturity scoring. Cisco's 2025 research showed only 12 percent of organisations are mature in AI deployment. Most pilots stall not because the technology fails, but because the operating model around the pilot was never built. There is no production handover discipline, no measurement framework, no escalation path when the pilot produces signal that is hard to act on.
3. AI ownership sits with the wrong function
When AI activity is owned by the CIO or a Head of Data, it tends to be treated as a technology programme. Vendor selection, infrastructure, model performance. Commercial outcomes drift to the periphery. The 13 percent of Nordic companies with a CAIO function consistently outperform the 32 percent where AI is owned by IT, because the role is specifically built to connect AI to revenue, cost, speed, risk, and capability.
4. Vendor stacks are accumulating exposure faster than governance can keep up
Most Nordic mid-market companies have added six to twelve new AI-enabled tools to their stack in the last 18 months. Most have not catalogued which of those tools touch which data, classify against the EU AI Act's risk categories, or have documented review cadence. The EU AI Act's main obligations land in August 2026. Most companies are operating blind on which of their existing tools will need remediation.
5. Capability has not transferred into the team
Most external AI advisory work in the region has been structured as "we will do it for you" rather than "we will do it with you, and then you will do it without us". When the engagement ends, the capability ends too. Janteloven and the broader Nordic enterprise culture also push against the kind of senior executive education that drives internal AI capability. That is a coordination problem, not an unsolvable one.
What changes when the operating model catches up
When the five causes are addressed, the value-capture rate moves. Across BRAGI engagements and the wider research base, the typical recovery looks like this on targeted workstreams.
| Outcome dimension | Typical range when operating model is in place | |---|---| | Revenue (qualified pipeline, conversion, retention) | +15-25 percent | | Cost (process overhead, automation lift) | -15-30 percent | | Speed (cycle time on targeted tasks) | 2-4x faster | | Risk (undetected risk events) | -30-50 percent | | Capability (organisation-wide AI competence) | Compounding over 12-24 months |
These ranges are indicative, not guarantees. They appear consistently when commercial sponsorship, programme management, vendor governance, and capability transfer all move at the same time. They almost never appear when only one of those is in motion.
Why the gap exists specifically in the Nordic context
Three Nordic-specific factors compound the global pattern.
Consensus-driven buying cycles. Nordic enterprise procurement typically involves 5-7 stakeholders and runs 6-9 months. AI tool selection inherits that cycle, which means by the time a tool is deployed the strategic context has shifted, and the operating model around it lags.
Janteloven and senior education. The cultural pressure against visible individual leadership in AI capability building means senior executives often defer the topic to specialists. The CAIO function exists in only 13 percent of Nordic enterprises because the role itself is culturally awkward to fill.
Regulatory readiness. Denmark was the first EU member state to implement the EU AI Act, with criminal penalties for breaches. The regulatory weight is real, and most Nordic companies are not yet operationally ready. 8 of 27 EU states are ready for EU AI Act enforcement, per recent McKinsey survey work, which means 19 are not, and the regulator is unlikely to be sympathetic to most of those when penalties start applying.
What this means in practice
For a Nordic mid-market company with AI activity already in motion, the practical first move is operating clarity. Three concrete steps.
- Score where AI activity currently sits across the five outcome dimensions. This is the BRAGI Baseline. It produces a number, not an essay. Numbers are easier to act on.
- Map vendor stack exposure against EU AI Act risk categories. Most companies have not done this. The exercise takes a day if the right people are in a room.
- Decide on a sponsor. AI activity that does not have a CEO, COO, CFO, or board-level sponsor will continue to drift to the periphery, no matter how much budget it has. The sponsor decision is the most under-discussed reason AI value capture stays at 4 percent.
The companies that act on this in the next two quarters will be the small minority that move from 4-percent value capture toward 25-percent value capture. The companies that wait will continue to spend the 70 percent without seeing the return.
How Bragi helps
BRAGI was built specifically for this gap. The four-week BRAGI Assessment produces a baseline (where the company actually is across the five dimensions), a vendor exposure map, and a prioritised opportunity recommendation. It is the standard entry point for mid-market companies that want operating clarity before committing to a broader programme.
For US companies with EU customer exposure, the same assessment surfaces EU AI Act risk through an AI Vendor Exposure Heatmap. For Nordic companies, it adds Danish AI Act and broader Nordic regulatory positioning to the picture.
Take the next step
If your company has AI activity but no clear operating picture, the highest-leverage first step is a scored baseline. The four-week BRAGI Assessment produces a baseline, an opportunity map, and a recommendation.
If you already know you want to talk, request a partner conversation directly.
Sources
- BCG Nordic AI research, late 2025 (4 percent meaningful value, 70 percent spend, 13 percent CAIO, 32 percent CIO-owned)
- McKinsey AI maturity research, 2025 (average 2.3 out of 4)
- Cisco 2025 AI deployment readiness (12 percent mature)
- EY 2025 governance gap (74 percent think covered, only 3 of 9 facets strong)
- Danish AI Act (Law No. 467) implementation
- EU AI Act enforcement timeline (main obligations 2 August 2026)
- McKinsey EU member state readiness, 2025 (8 of 27 ready)