The KPMG Global Tech Report 2026, published in June and drawing on insights from more than 250 consumer and retail leaders across 27 countries, contains a finding that most retail technology teams have quoted selectively. Nine in ten consumer and retail leaders say they are embedding agentic AI into their workflows. That is the number most press coverage led with. The same report documents that weak governance, fragmented technology stacks, and ongoing shortages of AI-native talent are the three obstacles that continue to slow progress across the same organizations. Ninety percent adoption. Documented execution failures. Both are true simultaneously.
That is not a contradiction. It is the most accurate description of where retail AI actually stands in 2026. The sector has crossed the threshold from experimentation to declared commitment. It has not yet crossed the threshold from declared commitment to sustained operational returns. The gap between those two thresholds is where most retail technology budgets are currently sitting, and it is the gap that determines whether the $50 million or more that most large retailers are now investing annually in digital technologies produces the $250 million in returns that the top performers are reporting, or produces a capability that looks like AI on paper and performs like a pilot in practice.
Ninety percent of retailers say they are embedding AI agents. Forty-five percent report returns of $250 million or more from their digital investments. The distance between those two numbers is the execution gap. What fills it is not technology. It is the three decisions that most retailers make incorrectly before the first agent goes live.
What KPMG Found When They Asked 250 Retail Leaders About AI
The report covers a sector moving at an unusual pace. Over half of the consumer and retail companies surveyed are investing $50 million or more annually in digital technologies. Sector confidence is high: 93 percent of leaders believe advanced systems will drive their future competitive advantage, and 86 percent say technology investments frequently improve overall business value. Additionally, 94 percent of those polled agree that managing an AI agent will become a necessary workforce skill in the next five years, while 87 percent say they are actively adding AI-native roles to their hiring strategies.
The returns, for those who are achieving them, are substantial. Forty-five percent of retailers and brands polled report financial returns of $250 million or more from their digital investments. One retailer documented a 30 percent increase in conversion rates from AI-driven personalized recommendations. The KPMG authors noted that retailers who have moved AI from isolated pilots to scaled operations are seeing measurable impact in forecasting accuracy, supply chain efficiency, and increasingly automated front and back-office operations.
However, the report is equally direct about the other side of the picture. Many organizations still struggle to measure technology investments against business outcomes, highlighting an execution gap despite strong confidence in AI’s strategic value. The three obstacles the report identifies as most consequential are not technology problems. They are organizational and architectural problems that precede any technology decision.
The Three Obstacles Between Adoption and Returns
The Governance Problem Nobody Is Measuring
Weak governance is the first obstacle the KPMG report identifies. In the context of AI deployment in retail, governance means something specific. It means the organization has defined who owns each AI output, what the escalation path looks like when the AI produces a result that contradicts human judgment, how performance standards get set before deployment rather than discovered after it, and what the accountability structure looks like when an AI-driven decision produces a bad outcome for a customer or a store team. Most retail organizations that have declared AI adoption have not answered those four questions in writing before go-live. The result is that the AI runs, produces outputs, and nobody is certain whether those outputs are producing the intended business result or simply appearing to do so because nobody defined the measurement framework in advance.
As I described in the context of the Starbucks AI inventory deployment analysis, the pattern of deploying AI without a defined performance standard and then abandoning it when performance is unclear is not a failure of the technology. It is a failure of governance design that preceded the technology decision by months.
The Tech Stack Problem That Precedes Every Agent Decision
Fragmented technology stacks are the second obstacle. An AI agent is only as useful as the data it can access and act on in real time. A retail organization where pricing data lives in one system, inventory data lives in a second system, customer transaction history lives in a third system, and those three systems do not share a common data layer cannot deploy a useful AI agent regardless of how much they invest in the agent itself. The agent will produce recommendations based on incomplete context, and the organization will spend its AI budget debugging data pipeline failures rather than scaling AI-driven outcomes.
Consequently, the 52 percent of retailers in the KPMG report planning major budget increases for cybersecurity and the 49 percent boosting data investments are making the correct sequencing decision. Data infrastructure and security are not adjacent to AI deployment. They are prerequisites for it. The retailers who close the returns gap first are the ones who funded the data foundation before they funded the agent layer.
The Talent Problem That Compounds Both
AI-native talent shortages are the third obstacle, and the one with the longest lead time to solve. The KPMG report found that 94 percent of retail leaders agree that managing an AI agent will become a necessary workforce skill within five years, but 87 percent are only now adding AI-native roles to their hiring strategies. That sequencing gap means the organizations investing in AI agents today are simultaneously short the people who know how to configure, monitor, and improve those agents over time. The technology can be purchased. The operational knowledge of how to run it well cannot.
In deployments I have worked on across the United States and Latin America, this talent gap shows up in a specific pattern. The AI gets deployed. The vendor provides training. The internal team learns to operate the tool as it was configured at launch. When the tool needs to be reconfigured because the business context changes, such as a new product category, a new pricing strategy, or a new customer segment, nobody internally knows how to do it. The vendor gets called. The reconfiguration takes weeks. The AI sits in a state that no longer matches the operational reality it is supposed to support.
What the 45% Getting Returns Are Doing Differently
They Define the Business Outcome Before the Technology Selection
The retailers reporting $250 million or more in returns from digital investments did not start with a technology selection. They started with a specific business problem: reduce forecast error by X percent, reduce stockout frequency in Y category, increase conversion in the AI recommendation lane from Z percent to a defined target. The technology selection followed from the problem definition, not the other way around. That sequencing means the performance measurement framework exists before the technology goes live, which means the organization knows within 90 days whether the investment is working rather than discovering the answer in a year-end review.
They Build for the Agent Layer, Not Just the Application Layer
The shift from AI applications to AI agents is not cosmetic. An application surfaces a recommendation. An agent acts on it, monitors the outcome, and adjusts its behavior based on what it learns. Building for the agent layer means the data infrastructure supports real-time feedback loops, the governance framework includes rules for when the agent can act autonomously and when it escalates to a human, and the technology stack has API accessibility that lets the agent read from and write to multiple systems without manual intervention. As I described in the shopper agent analysis, the retailers who grew sales 59 percent faster were not running better AI recommendations. They were running agents that could complete transactions, not just suggest them.
They Treat AI Outcome Measurement as a First-Class Operational Metric
The retailers who struggle to measure AI returns are measuring AI activity rather than AI outcomes. They track how many AI recommendations the system generated, how many customers interacted with the AI feature, and how many sessions included an AI touchpoint. The retailers generating returns track what changed as a result: conversion rate before and after AI introduction in a specific category, forecast accuracy delta in a specific distribution center, customer lifetime value in cohorts exposed to AI-driven personalization versus those not. Furthermore, as I described in the AI-referred traffic analysis, AI-referred shoppers convert at higher rates than any other acquisition channel, but only retailers tracking the right attribution model can see that signal clearly enough to optimize for it.
What This Means for LatAm Retailers
The KPMG report draws on global data, but the execution gap it documents is sharper in Latin American markets for a structural reason. The tech stack fragmentation that slows AI deployment in large U.S. retailers is more pronounced in LatAm markets where legacy POS systems, regional ERP configurations, and locally customized retail management software have accumulated over decades without a modernization cycle. An AI agent that needs clean, real-time data from inventory, pricing, and customer systems faces a longer path to production in a LatAm store environment than in a U.S. retail chain that has already run a platform consolidation.
That does not mean LatAm retailers should wait. It means they should sequence correctly. Data infrastructure investment comes before agent deployment. Governance framework design comes before technology selection. AI-native talent development starts now, because the five-year window the KPMG report describes is already running. The LatAm retailers who use 2026 to build the foundation will enter 2027 ready to deploy agents that produce measurable returns. The ones who use 2026 to deploy agents onto fragmented stacks without governance frameworks will spend 2027 explaining why the investment did not produce the outcome the vendor projected.
The Number That Defines the AI Conversation in 2026
The KPMG report contains two numbers that belong in the same sentence more often than they appear together. Nine in ten retail leaders say they are embedding agentic AI into their workflows. Forty-five percent report returns of $250 million or more from their digital investments.
$250 million.
That is the return the retailers who got the execution right are reporting. The distance between 90 percent adoption and 45 percent returns is not a technology gap. The technology exists. The distance is a governance gap, a data infrastructure gap, and a talent gap that most retail organizations have not yet closed because they prioritized the announcement of AI adoption over the architecture of AI returns.
The retailers who close those three gaps in 2026 will be the ones reporting $250 million returns in 2027. The ones who do not will be reporting adoption rates at their next board meeting and wondering why the returns are not following.
If you are building the governance framework, data infrastructure, or talent strategy for AI deployment in your retail organization, connect with me here or reach me on LinkedIn. I am happy to walk through the deployment framework we use across the U.S. and Latin America.
Adriana Rivas is a retail technology executive and AI strategist, and the founder of a U.S.-based hardware company specializing in self-service kiosks, POS systems, electronic shelf labels, and digital signage deployed across the United States and Latin America. She is the award-winning author of How to Implement Self-Service Without Failing (Amazon #1 Hot New Release, Silver Nonfiction Book Award 2025) and recipient of the Gold Stevie® Award, Thought Leader of the Year 2026. She is also recognized by Thinkers360 as the #7 Global Thought Leader in Retail and a Certified Master Expert in Retail.
Pingback: Retail AI Adoption: Why the Real Problem Is Not the Platform