AI's Biggest Winners Won't Be AI Companies

Most 2026 AI coverage reads as a race between model labs. The economic story is somewhere else: in trucking yards, staffing agencies, and small factories nobody calls "AI companies."

Daniel Kornum argues the operators with the most to gain from AI are not the ones building or buying the most AI. They are the ones running at the thinnest margins, where a single percentage point of cost reduction compounds into an outsized profit lift. His long-form essay, published on X and mirrored at Flowtivity, frames AI transformation as three levers: revenue, cost, and risk. For software companies, revenue is the dominant lever. For 3%-margin operators, cost is.

The arithmetic: a manufacturer running at 3% margins can convert a sub-1% reduction in coordination costs into a greater-than-25% increase in profit. A 30%-margin software company applying the same AI to the same 1% cost line barely moves its P&L. The lever is identical; the impact differs by an order of magnitude because the denominator is different.

What changes the denominator is the layer AI attacks. Kornum's argument centers on coordination costs, the work of scheduling, dispatch, invoicing, parts reconciliation, shift handoffs, and customer follow-through that low-margin operators have historically absorbed as permanent overhead. In trucking, that is dispatchers rerouting loads when a truck breaks down. In staffing, that is recruiters working a 200-person placement list against a 60-job pipeline. In manufacturing, that is planners reconciling a parts shortage against a customer promise date. None of it shows up as a line item. It is the friction between line items.

AI's industrial story, on this reading, is that those coordination costs are no longer structural. A scheduling model can reroute a small fleet faster than a dispatcher on a cell phone. A staffing agency's resume-to-match loop can run before a recruiter's morning coffee. The work has not disappeared; it has moved from phones and spreadsheets into a system that runs in seconds.

Kornum's second claim is about timing. The operators who adopt first bank the margin uplift and reset their cost position before competitors force the efficiency back into lower prices. In a commoditised trucking market, a 1% cost advantage does not stay a 1% cost advantage for long. Carriers without it cut rates to keep loads. The early mover's gain becomes the industry baseline within a few contract cycles. The same dynamic applies in staffing: a 25% lift in recruiter productivity is a temporary moat, not a permanent one, because the price of a placed worker converges to the new cost structure.

The math is illustrative, not measured. Kornum is a single author writing an opinion essay, not a research team publishing in a peer-reviewed venue, and the original piece names no customers and reports no deployment results. The 1%-to-25% worked example is a thesis device. A reader who acts on it should pressure-test the denominator against at least one named operator's actual books before treating the leverage as confirmed.

The investment picture the essay sketches is still useful as a frame. Capital that flows into AI for low-margin operators is buying a transient rent, not a permanent efficiency. That is a different risk profile from "AI makes the software company 30% more productive," which is mostly a permanent moat story. Operators, B2B software buyers, and the funds writing checks to vertical-AI startups should price the two differently.

Watch the next 12 months for the first public P&L from a small or mid-cap trucking carrier, staffing firm, or industrial distributor that has rebuilt its coordination layer on AI and reported the result. That filing is the test of the thesis: whether the 3%-margin operator can hold the margin uplift, or whether competition compresses it back into lower prices.