AGI Is Years Away. So Why Can't Your Company Use AI Right?

Here is a question I ask every founder who corners me at a conference to talk about AGI: does your customer service team still copy-paste answers from a spreadsheet? Nine times out of ten, yes. So why are we talking about superintelligence when you haven't automated the boring stuff yet?

This week, Demis Hassabis says AGI remains years away despite the breakthroughs at DeepMind. Jensen Huang says it already arrived. Two of the most credible people in the entire industry, standing in the same news cycle, contradicting each other completely. If they can't agree, no CTO should be building a roadmap around a fixed AGI date. That date is fiction. Plan around fiction and you will get fictional results.

Meanwhile, Sequoia is out with its "Tale of Two AIs" framing and the Millennium Project just dropped a warning about the total absence of coordinated global AGI regulation. Big week for AI discourse. And Sweden? Zero national headlines on AI policy. Zero strategy documents. Zero seat at whatever table is deciding how this gets governed. That silence is not Nordic calm. That is absence. And absence is a business risk, not a personality trait.

The Real Problem Isn't AGI. It's That You're Still Bad at AI.

I run HEIMLANDR.IO out of Jönköping, a city most Silicon Valley people couldn't find on a map. We build AI agents and custom AI solutions for companies who are, frankly, embarrassingly behind. Not because they lack ambition. Because they got mesmerized by the AGI debate and forgot to fix the fifteen small, ugly, immediately solvable problems sitting in their own operations.

Let's be blunt about what I see walking into Swedish and Nordic companies in 2026:

• Customer support still runs on templated replies a human copies by hand
• Sales teams manually chase leads that a basic automation workflow could qualify and route in seconds
• Finance teams reconcile invoices in Excel while their competitors run it through an agent overnight
• Nobody has connected their internal tools, so every "AI initiative" is actually three people's part-time hobby project

None of this requires AGI. None of this requires a model that reasons like a human. It requires someone in the building deciding to actually ship. That's it. The AGI debate is a luxury distraction. It lets executives feel like they're thinking strategically about the future while avoiding the unglamorous work of automating what's already automatable today.

AI Automation Business: What Actually Works Right Now

An AI automation business in 2026 doesn't need frontier models. It needs discipline. The tools that matter this week are almost boringly practical: n8n for connecting your existing systems without rebuilding them, and agent frameworks that let you wrap a language model around a specific, narrow, well-defined task instead of asking it to be a genius.

The open source world has quietly figured this out faster than most enterprise software vendors. Look at what's trending on GitHub this week: openclaw pulling in serious traction as a personal AI assistant that works across platforms, or superpowers, an agentic skills framework built around a methodology that actually works instead of a demo that looks good on stage. These are not AGI. They are practical scaffolding for getting real work done with the models we already have. That's the whole point.

Sweden vs. The World: We Are Not At The Table

Here's the uncomfortable comparison. In the US, Nvidia, Google DeepMind, OpenAI and a handful of frontier labs are the ones effectively writing AI policy in real time, because Washington is reacting to them, not the other way around. In Brussels, the EU AI Act is imperfect but at least it exists as a framework companies have to plan against. In China, state and corporate AI strategy are fused so tightly there's barely a line between them.

And in Sweden? Silence. Not one national headline this week on AI policy or strategy while Sequoia and the Millennium Project are publishing frameworks that will shape how governments regulate frontier AI for the next decade. Swedish companies are not sitting this out because we're careful and considered. We're sitting it out because nobody sent us an invitation, and nobody in Stockholm is fighting for one.

I've said this before and I'll keep saying it: Swedish tech policy has a habit of confusing caution with strategy. Dagens Industri covers Swedish startups raising money and pivoting, but where's the coverage of what happens when Google, Nvidia and DeepMind finish writing the AGI governance rules without a single Nordic voice in the room? Those rules will apply to us whether we showed up or not. That's the part nobody in Swedish boardrooms wants to admit. Absence from the table doesn't protect you from the outcome. It just means you have zero say in it.

Compare this to how fast Nordic companies moved on GDPR compliance once it became mandatory, versus how slowly they're moving on preparing for AI governance that's still being drafted. We are historically excellent at complying once the rules exist. We are historically terrible at shaping the rules before they exist. That has to change if we want Swedish AI development companies to compete in Europe rather than just follow whatever Brussels eventually mandates.

What This Means If You Run An AI Development Company in Europe

If you're building an AI development company in Europe, you're playing a different game than a Bay Area startup with unlimited compute and light regulatory friction. Your advantage isn't speed to AGI. It's trust, data locality, and the ability to ship something that actually works within EU compliance boundaries without six months of legal review. Lean into that. Stop trying to win a race you were never invited to run.

Where This Actually Goes: 2026 to 2031

Let me give you my honest read, not a hedge. Over the next two to five years, three things happen regardless of when or if AGI arrives:

First, the gap between companies that automated their boring processes early and those still debating AGI philosophy widens dramatically. This isn't linear. It's compounding. Every quarter a competitor runs on agentic workflows while you run on manual processes, they pull further ahead on cost structure and speed. By 2028, that gap becomes existential for the laggards, not just inconvenient.

Second, regulation arrives in fragments, not in one clean AGI treaty. Expect sector-specific rules (finance, healthcare, defense) tightening well before any general AGI framework exists, because governments regulate what scares them concretely, not abstract superintelligence. The EU will likely move first on liability and auditability of agentic systems making autonomous decisions, well ahead of any AGI-specific legislation. Sweden needs people in those working groups now, not after the text is finalized.

Third, the definition of AGI itself becomes almost irrelevant to business strategy. Hassabis and Huang disagree because they're measuring different things: Huang means "systems that outperform humans on economically valuable tasks," Hassabis means something closer to general, flexible, human-level reasoning across any domain. Businesses don't need the philosophical version. They need systems reliable enough to run unsupervised on a specific workflow. That bar is already cleared for dozens of use cases today. Waiting for consensus on the philosophical question while ignoring the practical one is how you lose a decade.

The AGI conversation is important. I'm not saying ignore it entirely. Someone at your company should be tracking it, especially if you're in a regulated industry where the rules will eventually cascade down. But it should not be the reason your invoicing is still manual in 2026.

What To Look At Right Now

If you want to move past the AGI distraction and actually fix something this quarter, here's where I'd point you:

• n8n: fair-code workflow automation with native AI, self-hostable if data residency matters to you (it should, in the EU)
• Firecrawl: if your agents need real, structured access to the web instead of stale training data, this is the practical infrastructure layer
• Ollama: run models like Kimi-K2.6, GLM-5.2 or DeepSeek locally, which matters if you're a Swedish company worried about sending customer data to a US-hosted API
• superpowers: a genuinely useful methodology for agentic development instead of another framework promising magic

None of these will bring you AGI. All of them will make your team measurably faster this month. That's the trade I'd take every time.

What To Actually Do About It

Stop debating AGI in your leadership meetings unless you're prepared to also debate why your support team still handles tier-one tickets by hand. Pick three processes this month that are repetitive, rule-based, and currently done by a human clicking through five systems. Automate those with an agent. Not a chatbot demo. A working agent that does the task end to end.

If you don't have the internal capacity, this is exactly the gap we close at HEIMLANDR. We build rapid MVPs so you can test whether an AI workflow actually works before committing budget, and full fullstack builds when you know it does. AI agent development cost is not what it was two years ago. It's dropped enough that "we can't afford it" is no longer a real objection for most mid-sized companies. It's a comfort excuse.

And if you're in Stockholm, Jönköping, or anywhere else in the Nordics with actual influence over policy conversations, start showing up to the ones happening about AI governance before the framework gets finalized without you. Waiting for Brussels to hand you a finished rulebook is not a strategy. It's an abdication.

AGI will arrive whenever it arrives, on nobody's predicted schedule, least of all the people currently arguing about it on stage. Your invoicing problem will not fix itself while you wait.

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.