Sweden's Supply Chain Panic Is a Software Problem in Disguise

Business Sweden published its report this week telling Swedish companies to regionalize supply chains and diversify suppliers. Every trade publication ran the geopolitics angle: tariffs, China risk, Trump unpredictability, war in Europe. All true. None of it is the actual story. The actual story is that regionalizing a supply chain used to require the scale of a Volvo or an Ericsson. Now a 40-person manufacturer in Gnosjö can do it with a laptop and the right AI agent stack. That's not geopolitics. That's an AI automation business shift, and almost nobody in the Swedish mid-market is naming it correctly.

I've sat in three conversations this month with CFOs who think their supplier diversification project and their software budget are two separate line items. They are the same project. If you can't see your supply chain in real time, adding more suppliers just adds more blind spots wearing a trench coat of "resilience."

The Geopolitics Costume

Here's what actually happened. For twenty years, regionalizing supply chains was expensive because complexity was expensive. Every extra supplier meant extra contracts, extra quality checks, extra logistics coordination, extra people staring at spreadsheets trying to figure out who's late and why. Big companies could absorb that cost because they had ERP departments and procurement teams the size of small towns. Mid-market Swedish industrial firms could not. So they consolidated. One supplier in China, one logistics partner, one point of failure. Efficient. Cheap. And, as 2020 through 2025 proved repeatedly, catastrophically fragile. Now the cost of managing that complexity has collapsed. Not because tariffs changed the math, but because custom AI solutions and AI agents can monitor forty suppliers as easily as they monitor four. An agent doesn't get tired reconciling shipment data across three ERPs in two languages. It doesn't need a coffee break to cross-check a customs document against a purchase order. That capability is maybe eighteen months old in any deployable, affordable form. The geopolitics gave companies a reason to move. The software gave them the ability.

Business Sweden's own numbers back this up quietly: the firms that report successful regionalization aren't the ones with the biggest supplier lists, they're the ones who upgraded their visibility systems first and expanded suppliers second. Read the Business Sweden report again with that lens and it stops being a geopolitics story and starts being an operations story.

Automating the Broken Thing Is Worse Than Doing Nothing

Here's where most Swedish mid-market firms are getting it wrong right now, today, this quarter. They're taking their existing broken process, the one built for a single supplier and a single logistics lane, and bolting AI tools onto it. A chatbot on top of a spreadsheet. An "AI-powered" dashboard that still requires someone to manually export data from SAP every Monday morning. This is automating dysfunction. You end up with a faster version of a process that was never designed to handle multi-supplier, multi-region complexity in the first place. The companies that will actually win this decade are rebuilding the process from the data layer up. Not "let's add ChatGPT to procurement." Instead: what does our supply chain actually look like as a graph of dependencies, risks, and lead times, and can an agent reason over that graph and act on it without a human in the loop for every decision? This is exactly the kind of problem the trending repo Graphify is built for, turning your codebase, your ERP configs, your supplier contracts and your logistics docs into a queryable knowledge graph instead of a pile of PDFs nobody reads. Most Swedish firms don't have a supply chain visibility problem. They have a "our supplier data lives in six formats and three languages and nobody built a graph of it" problem. That's a fixable engineering problem, not a geopolitical inevitability.

The Microsoft Trigger Nobody's Connecting to This

There's a second, quieter story breaking this week that CFOs haven't linked to their supply chain strategy yet: Microsoft killing volume discounts, reported by Computer Sweden. This sounds like a licensing footnote. It's not. It's a forcing function. When your legacy enterprise stack (the same stack that can't talk to your new suppliers' systems anyway) gets meaningfully more expensive, the ROI math on rebuilding versus patching flips overnight. I've watched three clients this year delay a proper systems rebuild because "the Microsoft stack works fine, why touch it." That excuse just got a lot weaker. If you're paying more for the old thing that can't do the new thing you need, you're funding your own obsolescence twice. Now is the moment to actually run the numbers on custom SaaS development versus renewing a bloated enterprise license that was never built for regional supplier complexity anyway.

Sweden vs the World: We Have the Infrastructure, We're Not Using It

Here's the part that should embarrass us a little. atNorth is opening a new metro data centre specifically to power Swedish digital transformation infrastructure. We have cheap, green, reliable power. We have a genuinely strong data centre buildout happening right now, arguably better positioned than most of continental Europe for compute-heavy AI workloads. Nordic hydro and wind power make our compute cheaper and cleaner than almost anywhere in the EU. And yet. Walk into a mid-size Swedish manufacturer in Jönköping, Gothenburg, or Norrköping and you'll find procurement running on the same fragmented spreadsheet-and-email workflow it ran on in 2015. Compare that to what's happening in Singapore or the UAE, where mid-market firms are deploying AI-driven logistics orchestration as standard practice because their governments treat it as economic infrastructure, not an IT nice-to-have. Compare it to the US, where venture money is throwing agent-based supply chain startups at every Fortune 500 procurement department, sloppy and overhyped in places, but moving fast. Sweden has the power infrastructure to win this. We have the engineering talent. What we don't have, yet, is enough companies treating supply chain visibility as a software architecture problem instead of a vendor negotiation problem. The EU's approach isn't helping. Regulatory energy is going into AI Act compliance paperwork and CE-marking frameworks for AI systems, important work, but there's no equivalent policy push helping mid-market firms actually rebuild their operational software. Compare that to how aggressively the EU pushed cloud adoption subsidies a decade ago. We're regulating the technology faster than we're helping companies use it.

Where This Goes: 2028 and the Agent-Run Supply Chain

Push this forward three to five years. The direction is not "more dashboards." It's autonomous agents making live sourcing decisions: reading tariff changes, reading a supplier's late shipment pattern, reading currency shifts, and re-routing an order to a backup supplier without a procurement manager clicking approve. We're not far from this. The agent tooling exploding on GitHub right now, projects like OpenHands for AI-driven development and ECC for agent harness performance, are the early scaffolding for exactly this kind of autonomous operational software. What's being built for coding agents today gets repurposed for procurement agents next year. It always does. As we get closer to genuinely general AI capability, the companies that already have their supply chain data structured, connected, and machine-readable will simply plug in more capable agents as they arrive. The companies still running on spreadsheets and quarterly Excel reconciliations will face a much harder, more expensive rebuild later, with less time to do it, under more competitive pressure. Regionalization without a rebuilt data and automation layer isn't a hedge. It's technical debt with extra shipping costs. The regulatory gap widens too. The EU AI Act will eventually touch autonomous procurement decision-making, and most Swedish firms deploying "AI-assisted sourcing" today have no governance framework for when that assistance becomes an actual autonomous decision. Nobody's thinking about audit trails for an agent that chose to switch suppliers. That's coming, and the firms who build clean, explainable systems now (not black boxes) will have a real advantage when compliance requirements catch up, likely around 2027 to 2028.

What to Actually Look At

If you're a CTO or founder trying to move on this instead of just reading about it:

• Graphify (github.com/Graphify-Labs/graphify) to turn your fragmented supplier docs, contracts, and legacy ERP schemas into an actual queryable graph before you let any agent touch decision-making.
• OpenHands (github.com/OpenHands/OpenHands) for teams who want to see where AI-driven development is heading, because the patterns being built for coding agents are the same patterns you'll want for operational agents.
• rtk (github.com/rtk-ai/rtk) if you're running agent workflows at any volume and your LLM token bill is starting to look like a second software license fee. It cuts consumption 60-90% on common commands. Worth knowing before you scale anything.
• Your own Microsoft renewal quote. Actually read it against what a custom-built alternative would cost over three years. Most CFOs haven't done this math since the volume discount news broke.

What to Actually Do About It

Stop treating supplier diversification and software modernization as two projects. They're one project with two names. Before you add a fourth or fifth supplier, ask whether your current systems can even see supplier three clearly. If the honest answer is "sort of, in a spreadsheet, updated weekly," you don't have a resilience strategy. You have a resilience aspiration. Start with visibility. Map your actual data: contracts, lead times, quality history, currency exposure, geopolitical risk per region. Get that into something an agent can reason over, not something a person has to manually refresh. Then build the automation layer, whether that's AI agents that flag risk before it becomes a delay, or a fully custom system built for how your business actually operates rather than how a 2015 ERP vendor imagined it should. If your supply chain contracts and payments need cross-border trust without a bank in the middle, that's where blockchain development and smart contracts stop being a buzzword and start being a real answer for provenance and settlement between regional suppliers who don't fully trust each other yet. This is a build problem before it's a buy problem. Most off-the-shelf supply chain software was designed for the old world of one dominant supplier and predictable lanes. It wasn't built for what Swedish companies actually need now.

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.