Salesforce made agentic commerce official on July 6th, rolling out Agentforce Commerce, AI agents built to carry shoppers from product discovery through checkout across ChatGPT, Google’s AI Mode and eventually Gemini.
While the company pointed to last year’s holiday season –when AI-influenced purchases made up a fifth of global online sales– as evidence that the shift is already underway, Google has been building toward the same moment from a very different angle. The tech giant has in recent history been pushing its own Universal Commerce Protocol as a common standard for how AI agents transact with retailers across platforms.
Most of the attention on that wave has gone to the online, customer-facing side of agentic AI. Less visible, however, is a problem sitting one step earlier in the chain. Whether the product an AI agent is trying to sell is actually on the shelf, priced correctly and displayed the way the brand paid for is the open question now.
An elephant in the room remains, which is that while the focus in popular culture may stil be on the technology’s impact on the web, physical stores still dominate global retail, commanding approximately 79% to 81% of total worldwide spending.
For brands that span the gap between physical and online, technology is now an essential part for companies small and large, providing the connective tissue between both worlds so the customer experience can be mapped and stock levels can be visualized in real time.
All of this has made AI one of the most closely-watched developments in retail. For physical and digital stores, it’s being used to optimize omnichannel experiences and compete in crowded markets. For consumer goods brands specifically, it’s also being pitched as a fix for a longstanding execution problem, in the shape of limited visibility into what’s actually happening with the products once they reach the shelf.
Here, effie.ai is one company positioning itself as the first to address that gap, building leading-edge agentic AI technology which the startup describes as the first AI agent built specifically for retail execution, capable of handling every task of a store visit, except for the hands-on merchandising itself.
Turning shelf data into action
In today’s retail market, speed is of the essence. Tight margins and steep operating costs mean that retail brands need visibility into their supply chains, warehouses, and storefronts at all times to properly manage supply and demand.
The pressure isn’t abstract. Retailers worldwide lost an estimated $1.73 trillion last year to what analyst firm IHL Group deems “inventory distortion”, with out-of-stocks alone accounting for roughly $1.2 trillion of that figure. For a brand that doesn’t control the stores its product sits in, a single day of an empty shelf can be gone before anyone at headquarters even knows it happened.
But, at the same time, social media and globalization entail that food and fashion trends can spread like wildfire across an international customer base. The brands that can launch and ship products or respond to seasonal demand stand to win big.
At the same time, without a robust software infrastructure and intelligent automations, brands won’t be able to capture such competitive advantages. This is part of why NVIDIA’s third annual State of AI in Retail and CPG survey found supply chain efficiency to be the most commonly cited reason retailers are turning to agentic AI, ahead of customer-facing use cases.
Among the firms building for that demand, effie.ai says retailers can turn shelf data into immediate actions and verified execution.
Led by CEO Ruslan Okhrimovych, the company, which already has worked with companies including Coca-Cola and Unilver, provides a structure that allows humans and AI agents to operate in a real-time closed loop, meant to let consumer goods brands capture the speed and efficiency of AI.
Such an approach leads to the consolidation of store-level requirements from multiple sources, analyzing real-time context and creating tailored action plans without losing oversight or control, thanks to guardrails that keep verification tasks and reports in the hands of HQ. In particular, its AI agent process includes five steps: consolidating store-level requirements from compliance documents, planograms and monthly priorities; reading the shelf in real time via image recognition; building a store-specific action plan; verifying execution during the visit; and reporting back to headquarters automatically.
As reported by the company, that shift cuts merchandiser time per visit by 56% and supervisor time by 55% with zero days of training needed for field staff to adopt new tasks, as the AI agent, rather than a manual, guides work in the moment. Effie.ai also claims a 2-4% sales uplift tied to its platform.
Making in-store interactions visible
Whether selling in their own stores or through other retailers, the companies that actually create the products need to pay extremely close attention to how their goods and brand are represented to customers.
Retail execution refers to the activities that ensure a brand’s products are available, visible, and correctly presented at the point of sale, from a promotional display unit at a supermarket to the digital imagery that accompanies an online product listing. It’s a discipline that has existed for decades, but one that’s traditionally run on a lag: a field rep visits a store, fills out a report, and headquarters finds out what happened days or weeks after the fact.
Staffing adds another layer to that lag. The Promotion Optimization Institute, a CPG and retail-execution research group, found that 88% of companies face execution challenges tied to merchandising complexity and high field-staff turnover. In sum: much of the workforce doing this job is still learning it when they leave.
The problem this creates is rarely a dramatic failure, but rather thousands of small ones that never individually register. Analytics from NielsenIQ built its On-Shelf Availability Barometer specifically to measure not whether a warehouse has stock, but whether that stock is actually reaching the shelf. Inventory systems can show a product “in stock” while it sits in a backroom, on the wrong shelf, or behind a mispriced tag. That distinction is the crux of the execution problem: a brand can have no supply issue at all and still lose the sale. Meanwhile, the lag is exactly what shows up in the numbers.
For multinational brands, this challenge is even more pronounced. Tracking thousands of SKUs across hundreds of thousands of stores means the $1.2 trillion in annual out-of-stock losses are not concentrated in a handful of failures, but spread thin across millions of small, individually invisible gaps. In this paradigm, a gap that lasts a few hours in a single store rarely shows up in a weekly or monthly audit report at all, which is precisely the kind of blind spot that’s pushed brands toward continuous, automated monitoring rather than periodic human visits.
Nestlé Ukraine had a mission to address this lack of visibility and turned to effie.ai as its partner of choice in the industry’s first full-scale deployment of an AI agent for retail execution.
In order to monitor shelf activity, SKUs, and store visits in near real time across 45,000 visits each month, the product team at effie.ai built a digital twin that fused GPS, timestamps, image recognition, and more. This resulted in Nestlé achieving near-perfect data accuracy across its stock, moving this figure from 75% to 97%, per the company’s account.
The company also reports shelf audit time fell by 60%, and that this data accuracy helped to improve the success of sales initiatives by ensuring stock for promotions was available 95% of the time, up from a baseline of 84%. With this digital foundation in place, we can expect that the next phase of development will bring agentic AI into the picture to support autonomous, intelligent decisions for retail execution.
An internal commitment to agentic AI and leading-edge technology
For effie.ai, the use of agentic technology isn’t only something for the consumer goods brands it works with. The company, in fact, has also turned it into its own engineering process.
Effie.ai says it adopted an agentic engineering model internally at the start of this year, embedding AI agents across its software development lifecycle to support engineers in analysis, implementation, testing and iteration, rather than treating AI as a coding assistant bolted onto an otherwise unchanged workflow.
That shift mirrors a broader challenge McKinsey has been tracking across the software industry. In a May 2026 report, the consulting firm found early adopters of what it calls the “agentic SDLC” are seeing larger engineering teams give way to smaller pods supervising AI-driven execution, with roles shifting away from manual coordination and testing and toward architecture, code review, and oversight of the agents themselves.
By its own account, effie.ai has seen a version of that same shift play out internally: it has cut its prototype delivery timeframe from months to weeks, and shortened the evaluation process for new implementations to a couple of days.
“It has not only accelerated delivery but also improved the way we approach problem decomposition, experimentation, and solution design,” said chief product officer Oleksandr Hryn.
For a company selling retail brands on the idea that agentic AI can compress a similarly long, manual process into something closer to real time, being able to point to its own engineering org as an early adopter is a useful proof point.
Closing the loop
For consumer goods companies, brand perception is built one shelf at a time, and effie.ai’s bet is that agentic AI can finally close the gap between what a brand decides in a boardroom and what actually happens in a store.
And as Salesforce, Google and others push agentic AI to the front of the customer shopping experience, companies like effie.ai are making the equivalent case for what happens before a customer ever shows up: that the same agentic approach can plan and correct what’s happening at the shelf, not just react to a report of it days later.
How far the industry is willing to hand that job to AI agents, rather than using them to inform human decisions, is still an open question. In fact, a Deloitte survey of more than 3,200 IT and business leaders, found only 21% of organizations consider their governance for agentic AI mature, even as adoption accelerates.
That gap is part of why effie.ai still frames its product around keeping verification and sign-off with human teams at headquarters, rather than letting agents act on their own. Whether that balance holds as the technology matures will likely be decided by the same pressure that’s driving the rest of the industry: brands that can act on what’s happening in their stores in real time.
