Our Take
The real constraint is not AI capability—it's organizational discipline to say no.
Why it matters
Enterprise AI pilots are cheap; shipping integrated, measurable systems at scale is not. IKEA's framing of the prioritization problem will resonate across retail, logistics, and supply chain ops, where vendor noise drowns out execution.
Do this week
CDO / CTO: audit your active AI projects this week and kill the bottom 30% by headcount spend—you're likely funding activity, not outcomes.
IKEA's Digital Chief Identifies the Real AI Risk
In a conversation with McKinsey, IKEA's chief digital officer framed the challenge facing large enterprises investing in artificial intelligence. The risk, he said, is not that AI fails to work—it is that organizations undertake too many initiatives without finishing any of them. "The risk of doing everything but maybe nothing" captures the diffusion problem: unlimited budgets and vendor pitches create the illusion of progress while measurable delivery stalls.
IKEA, a global retailer managing complex supply chains, inventory, and customer experience across dozens of markets, faces the same vendor pressure and internal enthusiasm as every other large enterprise. The company has chosen to be selective about which AI initiatives receive resources and executive attention, rather than attempting to optimize every function in parallel.
Execution Bandwidth Is the Actual Bottleneck
Most enterprises do not lack AI talent or access to models. They lack the organizational capacity to integrate AI into shipping products, measure outcomes, and iterate based on live customer or operational feedback. The diffusion problem is visible everywhere: marketing departments running chatbot pilots, supply chain teams spinning up demand forecasting POCs, customer service exploring agent automation, all at the same time, all competing for the same data engineering and ML ops resources.
IKEA's framing matters because it names the constraint that spreadsheets and vendor pitches obscure. A company can run 20 AI pilots and generate 20 reports. Only 3 or 4 will produce operational or revenue impact. The other 16 will consume budget, engineering time, and executive credibility, leaving the organization in a worse position to fund the winners.
Retailers, in particular, sit at an intersection of high customer expectation, thin margins, and complex logistics. An AI initiative that does not reduce cost per transaction, improve inventory turns, or decrease fulfillment time is expensive overhead, not innovation.
How to Triage Your AI Portfolio
The prioritization problem is not solved by frameworks. It is solved by honest accounting: which projects have a named owner with authority to ship, a measurement plan tied to a live metric, and realistic resource commitment for the next 12 months? Projects without all three should be killed or merged.
Practitioners should also separate exploratory work from committed initiatives. A small team can run 3 or 4 true experiments in parallel. The rest should be serial, queued, or outsourced. IKEA's implicit message is that the organization willing to say "not now" to 70% of AI asks will outship the organization trying to do everything.
The customer experience wins come later, but only if execution discipline comes first.