AI Pilots Need to Show Up on the P&L: What Startups Can Learn from RetailClub
Executives across retail call AI a top priority, yet very few can say what it has returned. In a June 2026 Deloitte survey, in fact, of 200 retail and consumer products executives, 75% named AI a top strategic priority, but only 16.5% could quantify a return from it.
That gap is the backdrop for the RetailClub AI Festival this week, opening Tuesday at Huntington Beach, California, and running through Thursday, September 24.
The event involves an all-outdoor boardwalk lined with sponsor activations and its organizers, the founders of Shoptalk and Groceryshop, expect around 2,000 attendees and more than 150 speakers.
The promise: retail has to move past AI experimentation.
So the question hanging over the week is simple. Will retail’s AI pilots still be around after their first serious budget review?
Money isn’t an obstacle. At least not yet. Deloitte found that 82% of executives it surveyed plan to increase AI investment over the next 12 months, most of it still going to IT and data infrastructure, rather than to use cases that generate revenue or protect margin.
For founders who sell into retail, the sharper question is who on the buyer’s side owns the result. That person ends up deciding what gets renewed, and throughout firms large and wide, it isn’t the one who championed the pilot.
Nisum, one of the technology providers on the boardwalk this week, has built two sessions around exactly that problem.
The Gap Between Priority and Proof
The Deloitte numbers show how wide the distance is between intent and execution: about half of respondents invest less than 0.5% of revenue in AI, while enterprise-wide deployment sits between 7% and 10% for both retail and consumer goods companies.
Outside of IT, wide adoption never exceeds 36% in any function.
Ownership is perhaps where the survey gets most interesting. In 54% of the companies, AI strategy sits with technology leaders rather than with the P&L owners who have to deliver business results from it. Here, a founder can win a pilot on technical merit with the first group and still lose the renewal to the second without ever having sat across from them.
Wider research leans the same way. MIT’s Project NANDA reported in mid-2025 that 95% of the generative AI pilots it examined delivered no measurable profit-and-loss impact, as reported by Fortune, based on 300 public deployments.
Gartner, meanwhile, predicted in June last year that more than 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value and inadequate risk controls.
Those are management problems more than model problems. But in fairness, retail isn’t uniformly stuck. Most respondents in both sectors say the biggest impact so far has come in productivity and cost reduction, with retail participants far likelier than consumer goods ones to report revenue growth – 38% against 10%.
Deloitte’s advice to closing the gap is to put strategy in the hands of P&L owners and to treat scaling, not piloting, as the measure of progress.
Where the Vendors Are Placing Their Bets
Nisum isn’t alone in reading the room this way. Several of the technology providers on the boardwalk this week are pitching some version of the same pivot, from proving AI works to proving it pays.
The California-based global consulting firm, which says it has worked in retail and commerce for more than 25 years and has over 1,500 engineers and consultants, will be hosting two Breakthru Experiences at the festival, under the title From Pilot to P&L: An Unfiltered Guide for Retail Technology Leaders, presented by board member and Cartweel co-founder Sajid Mohamedy.
According to the company, the sessions will examine why some retail AI pilots fail to reach production and what separates the initiatives that continue into the next budget cycle.
Its retail work is organized around four areas, and two of the four, AI-powered margin growth and data and AI for retail performance, are named for outcomes rather than tools.
Other vendors are running the same argument through a different door. KIBO is title sponsor of an invitation-only lunch called The Retail AI Debates, where it says senior retail operators will pressure-test the industry’s biggest AI questions rather than sit through another round of demos.
Centric Software, a product lifecycle management vendor, is also pitching its AI tools as a way to turn experimentation into what it calls measurable business value across product creation, planning and pricing, while Intelligence Node, a pricing and market intelligence company, is framing its pitch around a similar line: that the retailers who win won’t just deploy more AI, they’ll deploy AI backed by better intelligence.
The shared premise is that the market, at least the part showing up to sponsor the festival, has decided demos no longer close deals. What closes them now is a number a buyer’s finance team can check.
Budget cycles are, it is argued, where an enterprise software contract lives or dies. A pilot comes with a sponsor and plenty of goodwill, while a renewal comes with a finance team, and finance teams renew what shows up in a spreadsheet.
Founders know the other side of that screen, since investors have been checking them for profitability. Customers, on the other hand, run a quieter version of the same check every year.
Agents Raise the Stakes
Agentic commerce makes the pilot problem harder. 29% of retailers and 40% of consumer goods companies have no defined approach to it yet, and between 50% and 60% of the executives surveyed across both sectors are already piloting capabilities and prioritizing API readiness.
Execution, in other words, is outrunning strategy.
The scale of what’s coming explains the rush. Deloitte’s global retail outlook says 68% of retailers expect agentic AI adoption within the next 12 to 24 months, and it cites forecasts suggesting agents could influence up to 25% of global e-commerce sales by 2030.
But shoppers are a less certain part of the equation. June 2026 research also found that 72% of merchants in the UK and U.S. agree consumers will adopt agent-led shopping faster than most merchants are prepared for. In the meantime, 27% of the consumers surveyed said they trust no organization to operate an AI shopping agent for them.
Retailers are being hurried along by customers whose trust they haven’t fully earned. And that’s as much a problem for small merchants as it is multinational firms.
Then there’s the supply side. Gartner estimated that only about 130 of the thousands of agentic AI vendors are the real thing, and it says many use cases sold as agentic don’t need agentic implementations at all.
Signal enough is that the rebranding of chatbots and automation tools has its own name, “agent washing.” A retail buyer who has already sat through one stalled pilot is going to hear every new agentic pitch through that filter.
What Can Founders Take From the Gap?
The evidence has limits. Deloitte’s survey is 200 executives describing their own companies; Gartner’s number is a forecast, not a measurement; MIT’s finding applies to generative AI pilots and counts success as measurable P&L impact, a demanding bar for early-stage projects.
Even then, the direction is consistent enough for founders to act on. Finding out early who owns the number, and getting that person into the conversation before the pilot ends, addresses the ownership gap directly; writing the success metric down in the first week, in the buyer’s language, addresses the unclear business value Gartner lists among the causes of cancellation, which is the one cause a vendor can actually influence.
And scope matters as well. Aditya Challapally, the MIT report’s lead author, said that the young companies pulling ahead “pick one pain point, execute well, and partner smartly.” A narrow use case gives a finance team something it can check, whereas a platform promise gives it a reason to wait.
If P&L owners take over AI strategy, and if vendors agree on measurable outcomes before the first deployment instead of after it, the retail AI pilots that survive this year’s reviews will start to look less like experiments and more like line items.
Featured image: Zyanya Citlalli via Unsplash+