In 2025, agentic AI started to alter the retail landscape.
Brands and retailers that in 2024 thought generative AI was the most important technology to follow this year turned their zeal toward agentic AI. And though agentic is a type of AI that technology professionals have been discussing for several years as a possibility, this year it started to penetrate reality a bit more.
So what is agentic AI?
Related Articles
According to IBM, “Agentic AI is an artificial intelligence system that can accomplish a specific goal with limited supervision.” That is to say, an AI model that has been trained to use context to complete tasks autonomously. In many instances, agentic AI is powered by individual agents, which handle specific tasks needed to meet the larger goal.
Many technologies today have started to emerge under the guise of agentic — but many of them are still in their nascent stages. Some of that is because the technology isn’t fully baked, but some of it is because the end users — whether supply chain operators, store managers or consumers — aren’t ready to hand the reins to an AI model completely.
You May Also Like
Agentic AI on the Back End
AI has long been influencing global supply chains and the way many large brands and retailers do business. But the goal of agentic AI is to optimize those processes even further.
Companies like software giant SAP have been touting agents that come together to handle tasks autonomously — but even the executives who helped build them know that it’s still early days on convincing end users to adopt agentic.
Still, the German company and so many of its competitors — including Salesforce, Blue Yonder and others — are continuing to develop agents targeted at streamlining myriad functions inside organizations. One of the most widely discussed use cases in the industry today is supply chain agents.
Blue Yonder, for instance, has been working to deploy agents aimed at upping efficiency in the warehouse. It has also built an inventory operations agent, aimed at flagging and solving mismatches between supply and demand, and a logistics operations agent that helps schedule drivers and optimize routes, among others.
Meanwhile, SAP has revealed plans to consolidate a variety of agents under its AI assistant, which it calls Joule, so the fleet of agents remains interconnected across the business journey. Earlier this year the company shared that it will launch a system called Supply Chain Orchestration, which uses internal and external signals to proactively track supply chains with AI, alerting operators to any issues preemptively. Orchestrator then surfaces its insights to Joule, so the human operator can make decisions on issues in natural language. In the first half of 2026, SAP plans to integrate three new, supply chain-focused agents into its systems: a production planning and operations agent, a change record management agent and a supplier onboarding agent.
But text-based agents aren’t the only way agentic technology has cropped up in the supply chain. Voice agents have already started placing calls on behalf of companies seeking information about shipments, scheduling drivers and coordinating warehouse functions.
The agents are trying to obtain simple information, then making decisions based on that input, which means that this application is one example of a place where agentic AI has started to make a real impact, rather than needing to be babysat by human operators.
DHL Supply Chain and Flexport, two logistics players, have both publicly spoken about voice agents they have deployed. DHL partnered with HappyRobot to allow voice agents to make calls on its operators’ behalf for functions like “appointment scheduling, driver follow-up calls and high-priority warehouse coordination.” The organization expects it will save millions of minutes spent on the phone and avoid hundreds of thousands of emails annually.
Flexport said in the summertime that it is using voice agents to streamline cargo management, including by gaining information about shipment statuses. At the time, chief executive officer Ryan Petersen told WWD’s sister publication Sourcing Journal that the organization is “having AI make thousands of phone calls every day on behalf of Flexport to get statuses, to get your cargo moving on your behalf.”
Agentic Commerce Rises to Public Consciousness
Agents’ core promises aren’t just limited to back-end processes — and technology companies, brands and retailers alike have spent much of this year trying to prove that to consumers.
This year has seen the rise of what many companies call agentic commerce, which is to say consumers leaning on AI to take care of product discovery, or even transactions, for them. Still, many of the agentic commerce systems on today’s market require some degree of human intervention and ideation.
Competition for consumer eyes has continued to heat up between companies like Google, OpenAI, Perplexity and Amazon, all of which have switched on so-called agentic capabilities in the last 12 to 18 months.
OpenAI has launched multiple shopping-specific upgrades over the past few months, the most recent of which focuses on finding particular, high-spec products on behalf of consumers. The tool, which it calls shopping research, can find similar products based on images, suggest gifts for consumers’ friends and family, compare features of products and find matches for complex occasions and needs.
OpenAI researchers said shopping research was designed to “index on all the both implicit and explicit requirements” a user shares when looking for specific items, surfacing dozens of results from high-quality product sites, based on reviews, product descriptions and more.
Google revealed just ahead of the holiday season that it had upgraded Gemini to surface better shopping results, based on users’ queries. It also noted that it has enabled voice agents that are able to call local brands and retailers’ stores to determine whether they have specific items in stock — and at what price point. From there, an agent aggregates the results of the phone calls to give the user a rundown of where they might find the item so that they don’t strike out going to multiple retailers without additional information about what they might find.
Amazon has been taking a different path — rather than allowing other generative AI models, like ChatGPT, Perplexity Comet or Gemini to use its product catalog to surface results to users, it has blocked many other companies’ crawlers, seemingly in favor of building out its own agentic AI strategy.
Shoppers are beginning to trust AI more when it comes to transacting, but many still falter at the idea of AI using their personal financial information. A number of payment providers, including Stripe, Visa, Mastercard and Google Pay, have been working to show consumers that agentic commerce can be safe when processed through a trusted provider.
Google recently launched a Buy for Me function that enables users with Google Pay accounts to have an agent complete a transaction with their permission and on their behalf. OpenAI has been working with Stripe on an openly shared set of guidelines they call the Agentic Commerce Protocol, or ACP, aimed at ensuring users’ data remains safe when making transactions through AI models. OpenAI itself has started allowing transactions on its ChatGPT platform with certain retailers, like Walmart and Etsy. Amazon has introduced a Buy for Me function that allows users to ask the company to seek out a product from a different retailer and buy it using their Amazon credentials and payment methods.
Where We’re Headed
If consumer and end user trust for agentic AI — in commerce, at the office, in the warehouse and otherwise — increases, some estimates show that the way people shop and work could change drastically.
But, as executives at some of the largest technology providers have themselves admitted, it could take a little while to convince legacy technology users — and the everyday consumer — of agentic’s true merits. For all the talk that circles the retail industry about AI’s promise, the fact is that Americans still fear job loss and replacement because of AI.
An August poll conducted by Reuters and Ipsos showed that more than seven in 10 Americans said they hold concerns about AI’s ability to put people out of work permanently. Pew Research data from September shows that 57 percent of Americans believe the general risks associated with AI’s proliferation are “very high,” and an additional 26 percent of Americans believe it has “medium” risk.
For agentic AI use cases to become truly democratized, technology companies and workplaces are likely to need to work on connecting directly with humans. While the technology behind agents will, no doubt, continue to change and grow, organizations that retain the highest benefits from it are likely to be those that bring people along for the changes technology incites.