The AI Productivity Lie: Executives Gaslight Their Own Companies

Here is a number that should make every board member uncomfortable: over 80% of companies report no measurable productivity gains from their AI investments. Not "disappointing gains." Not "slower than expected." No gains. Meanwhile, only one third of executives surveyed actually use AI themselves, and those who do clock an average of 90 minutes per week. That is less time than most of them spend in Monday standups. This is the state of enterprise AI in May 2026, and I think it is the single biggest misallocation of capital since the dotcom bubble.

I run HEIMLANDR.IO from Jönköping, Sweden. We build custom AI solutions and AI agents for companies that actually want results, not slide decks. So I have a front-row seat to the gap between what companies say about AI and what they actually do with it. That gap has become a canyon. Let me walk you through why, what it means, and what serious builders should do about it.

The Corporate AI Gaslighting Loop

A survey of 6,000 executives published this week confirms what many of us have suspected: the C-suite is running a con on itself. Here is how the loop works.

Step one: CEO announces AI transformation at the annual offsite. Consultants are hired. Budget is allocated. Press releases are issued. LinkedIn posts are written.

Step two: Middle management scrambles to "implement AI" without clear direction. They buy licenses. They run pilots. They spin up a Center of Excellence that produces dashboards about dashboards.

Step three: The CEO and CFO themselves barely touch the tools. They do not prompt. They do not build workflows. They do not even read the output. They sign off on quarterly reports that say "AI initiative on track" because nobody wants to tell the emperor about the wardrobe situation.

Step four: Harvard Business Review reports that managers and executives now fundamentally disagree on AI's value. Managers see the gap between the promise and the reality. Executives see their own press coverage. The disconnect costs real money. Real momentum. Real talent.

This is not an AI problem. This is a leadership problem wearing an AI costume.

90 Minutes a Week Is Not Adoption. It Is Tourism.

Let me put this in context. 90 minutes a week is roughly the time it takes to make two espressos and scroll through your inbox on a slow Tuesday morning. If your most expensive strategic initiative gets less executive attention than your coffee ritual, you do not have an AI strategy. You have an AI hobby.

I talk to CEOs and CTOs across Europe every week. The ones getting real value from AI automation in their business share a common trait: they use the tools themselves. Not in a demo. Not in a boardroom. At their desks. In their terminals. On real problems.

The executives who can describe what Claude Code does because they have used it to refactor a codebase, or who have built an internal AI agent to handle their own email triage, or who have personally tested an MVP built with rapid prototyping. Those are the ones whose companies show up in the 20% that actually see productivity gains.

The rest are tourists. They visited AI. They took a photo. They went home.

Sweden and the Nordics: Two Tracks, One Fork in the Road

Here in Sweden, I see a split forming that mirrors the global dysfunction but with a distinctly Nordic flavor.

On one track, you have the enterprise theater. EY is showcasing agentic AI through Microsoft Copilot as an "employee empowerment" play. The Big Four consulting firms are running the same playbook here as everywhere else. Deploy Copilot. Run workshops. Produce a report. Invoice. Repeat. Swedish corporates dutifully participate because nobody gets fired for buying Microsoft.

On the other track, something more interesting is happening. FMV (Försvarets materielverk, Sweden's defence procurement agency) has been assigned to protect sensitive technology. This is a national security move that signals the Swedish government understands AI is not just an enterprise productivity question. It is a sovereignty question. The Nordics have always been good at quiet, serious infrastructure work while the rest of the world makes noise. This is that pattern repeating.

But here is what worries me. The Swedish tech ecosystem, particularly the venture and scale-up layer, is caught in the same gaslighting loop as everyone else. I see Breakit headlines about AI-first companies that, when you dig into the product, have a ChatGPT wrapper and a hope. The EU AI Act adds regulatory overhead that large companies can absorb and small companies cannot. And Swedish universities are producing excellent ML researchers who leave for San Francisco because Stockholm cannot match the offers and Jönköping does not yet have the critical mass to retain them.

Sweden is getting the defense and infrastructure side right. Sweden is getting the enterprise adoption side wrong for the same reasons everyone else is. Leaders are not leading from the front.

Why the Consulting-Industrial Complex Makes This Worse

Deloitte's Tech Trends 2026 and KPMG's Global Tech Report both frame AI as the defining executive priority of the year. Of course they do. Their revenue depends on it being the priority. I am not saying they are wrong that AI matters. I am saying the way they frame it guarantees the gaslighting continues.

When a consulting firm tells you AI is your top priority and then sells you a 12-month transformation roadmap, they are selling you a process, not an outcome. The 80% of companies with no productivity gains all have roadmaps. They all have governance frameworks. They all have Centers of Excellence.

What they do not have is a CEO who can open a terminal and run a prompt.

I am not saying every CEO needs to be an engineer. I am saying every CEO needs to have put hands on the actual tools enough to have an informed opinion. You would not lead a manufacturing company without ever visiting the factory floor. Why are you leading an "AI-first" company without ever writing a prompt that matters?

What AI Agent Development Actually Looks Like When It Works

At HEIMLANDR, we build AI agents for companies that want them to do actual work. Not demonstrate potential. Not win innovation awards. Work.

Here is what works: small, specific, measurable. An AI agent that processes incoming RFPs and produces first-draft responses, saving a sales team 15 hours a week. An automation pipeline that monitors regulatory changes across EU jurisdictions and flags what matters to a compliance team. A custom AI solution that reads customer support tickets, categorizes them, drafts responses, and routes edge cases to humans.

None of that is glamorous. None of it makes a good keynote. All of it produces ROI you can measure in weeks, not years.

The companies in the 20% that see real gains are doing this kind of work. They skipped the roadmap and went straight to "what costs us time and what can an agent do about it." They treated AI development like any other engineering project: scope it, build it, ship it, measure it, iterate.

The companies in the 80% are still on slide 47 of the strategy deck.

Where This Goes: 2027-2030 and the AGI Shadow

Let me be direct about the trajectory, because this is where the stakes get uncomfortable.

In the next 2-3 years, the gap between companies that actually adopted AI and companies that performed AI adoption will become an existential competitive divide. Not a disadvantage. A death sentence for some. When your competitor's cost structure is 40% lower because they have actual working AI automation throughout their business, your premium pricing and brand loyalty will not save you.

The path toward AGI, or whatever we end up calling systems that can reason across domains and execute autonomously, makes this more urgent, not less. Every month that passes, the tools get more capable. The companies that have been building real AI muscle memory will be able to adopt the next generation of tools in weeks. The companies that have been performing will need another 12-month roadmap from Deloitte.

EU regulation, including the AI Act, is designed for the world of 2024. It is already struggling with the realities of 2026. Agentic AI, where systems take actions rather than just generate text, sits in a regulatory gray zone that nobody in Brussels has adequately addressed. Swedish companies operating as an AI development company in Europe need to be building with compliance awareness but not waiting for regulatory clarity that may not arrive before the market moves past them.

The companies that survive the next five years will be the ones where leadership actually understands the technology. Not from a briefing. From use.

What to Look At

If you are a CEO or CTO reading this and feeling the gap between your AI announcements and your AI reality, here is where I would start:

1. AutoGPT

AutoGPT (184K stars on GitHub) remains one of the most accessible ways to understand what autonomous AI agents can actually do. Install it. Give it a real task from your business. Watch what happens. Form your own opinion instead of reading someone else's.

2. Langflow

Langflow (148K stars) lets you visually build and deploy AI-powered agent workflows. If your team is trying to figure out what AI agent development looks like in practice, this is a serious tool. Not a toy. Not a demo. A tool your engineers can use on Monday morning.

3. Claude Code

Claude Code (120K stars) is Anthropic's agentic coding tool that lives in your terminal. If you are a technical leader, spend a week with it on a real project. If you are a non-technical executive, sit next to an engineer who is using it and watch what happens to their velocity. That direct observation will teach you more about AI's actual value than any consulting report.

4. System Prompts Collection

This repo (137K stars) exposes the system prompts of major AI tools: Cursor, Devin, Windsurf, Replit, and more. Understanding how the best tools are engineered at the prompt level gives you a real education in what matters versus what is marketing. Read these before you sign another enterprise AI contract.

The Uncomfortable Bottom Line

The real AI crisis is not that the technology does not work. It works. I see it work every day in the things we build at HEIMLANDR. The crisis is that the people signing the checks and making the speeches and setting the strategy are not doing the work themselves. They are funding something they do not understand from experience, championing something they barely touch, and then wondering why the results are not there.

90 minutes a week. That is what billions of dollars in AI investment gets from the people who authorized it.

If you are a leader reading this, here is my challenge: this week, cancel one meeting. Use that hour to actually work with an AI tool on a real problem you care about. Not a demo. Not a sandbox. Your real work. Your real data. Your real decisions. Do this every week for a month. Then look at your AI strategy again and tell me if it still makes sense.

I bet it will not. And that is the beginning of actual transformation. Not the kind you announce. The kind you do.

Fredrik Brunnberg is the CEO of HEIMLANDR.IO, building AI and software solutions from Jönköping, Sweden. This is the daily HEIMLANDR briefing. If you found this valuable, share it with someone who builds things.