Point AI at what was impossible
Ludvig Strand has one of the more unusual seats in the AI conversation. As merchant tech and AI future analyst at Axel Johnson Group — one of Sweden's largest family-owned groups, with over 300 companies from Axfood to Dustin — his job isn't to ship one more AI project. It's to lift the gaze across the whole portfolio and ask what the collapsing cost of cognitive work does to how we organize, market, and work. His central warning: most companies apply AI through what he calls skeuomorphism — the same instinct that made early electric factories just install lamps on the ceiling instead of doing what Ford did and reinvent the assembly line. The hard part isn't seeing AI's value. The hard part is seeing that most of what you're building with it is just lamps.
His fix is to go back to first principles and ask what was economically nonsensical last year that's now on the table. Talking to every supplier instead of the top 5%. Reviewing every contract. Doing the work that was impossible, not just automating the work you already do. That reframe carries into governance too: agents aren't software, they're closer to labor, so the old control-first IT playbook has to give way to agentic governance — short-lived identities, spending limits, and continuous reassessment as the baseline. The METR benchmark went from 22 minutes of automated software work in early 2025 to over two days by mid-year. If your strategy still assumes the 22-minute world, you're planning against yesterday's model.
