CHICAGO — Artificial intelligence has moved from keynote slides into the weekly stand-up at companies that do not think of themselves as tech firms.
Controllers at regional manufacturers are using assistants to draft variance comments. Warehouse supervisors are testing routing suggestions. Insurance claims units are triaging incoming mail before a human opens the file. The pattern, in interviews with a dozen CIOs this month, is the same: small seats, narrow tasks, a demand for a measurable hour saved.
What they are not buying, several said, is a second data platform. After two years of overlapping cloud contracts, budgets are being clawed back toward core systems — identity, backup, the ERP already in place. Vendors that cannot sit on top of those systems are losing proofs of concept.
Labor groups have taken notice. A Midwest logistics local said it would not oppose tools that cut overtime if staffing levels on the dock were preserved. A white-collar union representing claims staff in two states has asked for written rules on when a model’s suggestion can be used in a denial letter.
Regulators are still catching up. The Federal Trade Commission has warned firms against overstating what automated tools can do in consumer-facing decisions. State privacy laws in California and several other jurisdictions already require disclosures that many mid-market legal teams are only now mapping.
The practical picture for 2026 is less “transformation” than selective automation. The companies moving first are the ones that can point to a boring, expensive process and a vendor willing to price against that cost.