By Guest contributor |
Merchandising sits at the very heart of retail. And yet, it has consistently been handed the lowest-grade software in the building. This is one of the biggest problems facing retailers today – and it goes much deeper than many realise.
The issue is not that merchants lack data. If anything, they are drowning in it. The issue is that the process of turning data into decisions remains almost entirely manual. Merchants are still acting as the human glue between disconnected systems – pulling figures from an enterprise resource planning (ERP) solution here, a business intelligence (BI) tool there, a planning spreadsheet somewhere else, then stitching it all together to run a single what-if scenario. By the time that scenario is ready, the market has already moved. Plans that take the best part of a week to build are partially stale before they are finalised. The cost of that latency is measured in lost margin, missed trends and reactive decisions made too late to matter.
Legacy machine learning systems tried to solve this. But they made things worse – and in a specific way. They moved fast, but they moved opaquely. A recommendation would arrive and no one could explain it. The demand forecast, the transfer proposal, the markdown suggestion – all of it from a black box, and merchants were asked to trust it or ignore it. That bargain was never going to hold. When a recommendation misses badly and you cannot figure out why, you lose confidence in the whole system. The spreadsheets come back out. We have all seen it happen.
Agentic AI is a fundamentally different proposition. The distinction is not speed (although the speed gains are real – what used to take three days now takes 40 minutes). The distinction is reasoning. It tells you what data it used, what assumptions it made, what trade-offs it considered and what it recommends you do next. Critically, it ingests live signals in real time, such as market data, inventory movement and external signals like weather patterns or upcoming events, and acts on them continuously, not in weekly batch cycles. You can question it and ask it to rerun the scenario with different constraints. And every override, adjustment and accepted or rejected recommendation becomes a training signal that makes the system smarter.
Leading retailers are already realising the benefits. Balenciaga, the global luxury fashion house, deployed Intelo.ai’s Collaborative Intelligence Agent Network to augment human decision-making and master the complexities of modern luxury retail. Clothing retailer The Children’s Place, improved sell-through by 42 per cent. These are not marginal gains – they are the compounding result of decisions being made faster, with better information and with institutional knowledge captured rather than lost.
That last point matters more than it might appear. One of the most underappreciated costs in retail is what walks out the door when an experienced merchant or planner moves on. The judgment, the heuristics and the hard-won understanding of how a particular brand’s customer behaves lives in people’s heads and in spreadsheets and then it vanishes. Agentic systems change that equation.
Every decision made within the system becomes part of a continuously learning, compounding institutional intelligence. The data is yours, the flywheel is yours – and every season it accelerates.
At Intelo, we use the model context protocol (MCP) and agent-to-agent (A2A) protocol to eliminate the manual ‘human glue’ between siloed retail systems. Internally, our merchandising agents use them to share tasks and live data instantly, replacing days of spreadsheet stitching. Externally, they act as a universal translator, connecting Intelo’s agents to existing enterprise systems – including legacy databases and the Microsoft stack – without complex custom IT work. The result is a secure, continuous network that frees merchants to focus on running the business.
That institutional capability is further strengthened by Intelo’s close partnership with Microsoft. Every Intelo agent runs natively on Microsoft Azure and is engineered to meet Microsoft’s security and compliance standards. For enterprise retailers already standardised on Microsoft, that matters: Intelo is available to purchase through Azure Marketplace against existing consumption commitments, and accessible directly within Microsoft Teams – meaning the path to deployment is shorter, the security review is familiar and there is no new infrastructure to evaluate.
As for where this is heading – the question I hear most often is whether we are moving towards fully autonomous merchandising. My answer is no. Merchandising is a blend of art and science – and AI will not replace the creative dimension. Instead, it will provide a PhD-level assistant that delivers more powerful analytics while freeing merchants to focus on the humanistic, curatorial judgement that no algorithm can replicate. The goal is a model where agents earn increasing trust through transparency, consistency and measurable outcomes, and where the boundaries of autonomous action expand in proportion to that trust. What the best merchants bring is exactly what retailers need more of, not less. Agentic AI does not diminish that. It finally gives it room to breathe.
Microsoft perspective
“Most merchants already run their day on the Microsoft stack, which makes Copilot the natural user interface for the modern workforce. The future of merchandising is agentic – where Intelo’s agents work alongside Microsoft’s tools, moving AI from analysis to action. By taking on the heavy lift, they ensure merchants spend less time in data and more time actually running the business.”
Kimberly McKinley
Director, Retail Industry Strategist, Microsoft
Noah Herschman is a strategic industry advisor at Intelo.AI
Discover more about enterprise solutions from Microsoft Partner businesses in the Summer 2026 issue of Technology Record. To get content like this on a regular basis, subscribe to the print edition or free digital edition of the quarterly Technology Record magazine.