# Adoption is an operating-model question > 37% of organisations see any EBIT impact from AI, unchanged in a year. Ownership, failover and funding decide the rest. Four questions before the next AI purchase. Source: https://barkingcat.ai/thinking/adoption-is-an-operating-model-question/ ## Thinking · October 2026 · Position Retailers that get value from AI are the ones that change who owns the work, who is paid for the result and what happens when the machine gets it wrong. The model they buy matters far less than any of those. UK retailers already know this, which is what makes the budgets interesting. When IMRG asked its members in 2026 which retailers would get the most value from AI over the next two years, 42% chose those that use AI to support people rather than replace them, 29% those with the highest quality data and 22% those that embed AI into existing workflows. Two per cent chose the retailers that allocate the most budget. Meanwhile NVIDIA’s 2026 survey has nine in ten retailers increasing AI spend this year, half of them by 10% or more. NVIDIA sells the chips, so read that as directional, but the direction is clear enough: the industry believes value comes from the operating model and is spending on the technology. The results follow the belief rather than the spend. McKinsey’s State of AI 2026, published in August from 1,719 respondents in 97 countries, has 37% of organisations attributing any EBIT impact to AI, essentially unchanged on a year earlier. The share McKinsey classes as high performers is 6%, also unchanged. McKinsey’s Global Institute puts the reason plainly in its May 2026 study of AI and work in Europe: incremental improvements at the task level rarely translate into meaningful gains, and the value sits in redesigning the workflow around the technology. ## Where adoption breaks Ownership is the first place. In a Statista survey for OMMAX of 250 European decision-makers in April and May 2026, 48% said AI execution was owned by IT or engineering and 7% by business units. Of the AI initiatives that failed, 79% failed in the pilot or between pilot and production. OMMAX and its co-sponsors sell AI transformation and integration, so the framing flatters them, but the shape is familiar to anyone who has run a programme. A pilot owned and measured by IT ends where IT’s remit ends, which is usually one step short of the trading P&L. Failover is the second. In February 2024 Klarna announced that its AI assistant had handled 2.3 million conversations in its first month, two-thirds of its customer service chats, the equivalent work of 700 full-time agents. In May 2025 its chief executive told Bloomberg that cost had been too dominant a factor, that the result was lower quality, and that customers would always be able to reach a human. By its Q3 2025 results Klarna was describing the assistant as doing the work of more than 853 agents and saving $60 million, presented as rebalancing rather than reversal. The practice’s reading, which is inference: the technology did not change much between those three statements. The operating decision did. What moved was the design of the handover, the point at which a query is recognised as needing a person and how quickly it reaches one. Readiness for AI-first service is a property of that design far more than of the customer’s appetite. Funding is the third. McKinsey’s July 2026 survey of 1,205 executives found 32% of self-described first movers had reallocated 20% or more of their resources year on year, against 10% of late movers. Late movers were 2.6 times more likely to say funding followed internal politics. Most retail businesses do not lack AI ideas. They lack a mechanism for moving money and people behind the three that matter. ## The pattern is older than AI I have seen this argument before, with a different technology in the middle of it. At a national bed retailer where I ran ecommerce, there was real internal competition between the stores and online, and store teams were inclined to treat the website as a rival rather than a tool. The answer we reached for was in how sales were credited. First we used store visits to estimate digital’s contribution to the stores. Then we went further and allocated online sales to stores by matching each store’s postcode to the delivery postcode of the online order. The aim was to report online and physical retail as one system and to take the heat out of the competition that made store teams resist the website. That was an operating-model change made to let a technology be adopted. AI adoption in retail is the same problem with a faster clock. The tool will be bought regardless. Whether it touches the P&L depends on whose numbers it improves, and whether those people were in the room when it was chosen. ASOS’s chief technology officer said much the same in June: the most significant challenge in scaling AI is organisational rather than technological, and because the underlying technology is not proprietary, the advantage lies in how well and how fast it is deployed. ## Four questions before the next AI purchase Who owns the outcome, by name, with a budget, inside the business unit whose numbers it should move? Which line of the trading P&L moves, by how much, and over what period, written down before the pilot starts? What happens when it is wrong? Who notices, how fast does the customer reach a person, and who is accountable for the handover? What stops so that this can start? If the answer is nothing, the resource reallocation McKinsey measures has not happened, and the pilot will end where pilots usually end. ## Sources - IMRG, Tech and AI Connect, July 2026, p.32 (retailer survey; sample size not stated). - NVIDIA, State of AI in Retail and CPG 2026, January 2026 (vendor survey, directional). - McKinsey, The state of AI in 2026: On the road to ROI, August 2026 (1,719 respondents, 97 countries). - McKinsey Global Institute, Agents, robots, and us: How AI reshapes work and skills in Europe, May 2026. - OMMAX, The AI Trends Report 2026, June 2026 (Statista survey for OMMAX, Ibexa and Make, n = 250, five European markets; vendor-produced). - Klarna, press release of 27 February 2024; Bloomberg interview with Sebastian Siemiatkowski, 8 May 2025, as reported by CX Dive and Fortune, 9 May 2025; Klarna Q3 2025 results as reported by CX Dive, November 2025. All Klarna figures are self-reported. - McKinsey, Why accelerated resource allocation matters in the age of AI, July 2026 (n = 1,205, 94 nations). - McKinsey, The rise of the agentic shopper: ASOS’s AI investment, 18 June 2026. - Store attribution of online sales at a national bed retailer: first-hand. ## Contact steve@barkingcat.ai Barking Cat Advisory Limited. Registered in England and Wales, company number 17338003.