Gam Dias is an AI and Data Strategist who has spent the last 20 years helping both Fortune 500 companies and innovative startups develop data strategies and AI driven products. The Retail Hive’s Ed Lawson spoke with him to get his thoughts on the potential for agentic AI in retail: what will we see first, and what should retailers do now to be prepared?
Gam Dias
Ed Lawson
What is agentic AI?
Agentic AI refers to autonomous systems that not only suggest actions (like traditional recommendation engines), but can execute them on behalf of a user. Think of it like an intern: capable, fast, but new—able to follow goals, make decisions, and act.
So, is agentic a natural evolution or a big new step?
It’s both. We’ve had product recommendations for over a decade. The difference now is agency: AI can reason, make decisions, and carry out tasks. That’s a fundamental shift. We move from generative AI (which produces content) to agentic AI (which achieves goals).
What industries will lead?
I see retail and travel most likely. Regulated industries like banking or healthcare will move slowly due to compliance constraints. Retail and travel are less encumbered, making them ideal for early agentic applications.
The reason is friction: planning a trip or finding the right outfit still involves too much manual effort. Agentic AI can simplify those journeys instantly. You’re already seeing GPTs in ChatGPT that plan trips or offer fashion advice. This is just the beginning.
Will agentic AI change how we buy, or how retailers sell?
Both. Consumers will have personal agents that shop for them. Retailers will deploy their own store agents that handle everything from triaging customer queries to proactive reordering. Eventually, we’ll see agent-to-agent commerce where bots negotiate, transact, and resolve issues.
So, a shopper’s agent visits a site, not the shopper themselves?
That’s right. Your agent will browse, filter, compare, and even buy on your behalf. Retailers will need to recognise and cater to these agents as if they were customers.
Forecasting, stock management, procurement—agents will outperform humans in scale and speed. That doesn’t eliminate humans, but refocuses them on creative and strategic tasks.”
What about the obstacles of Regulation, Reliability, and Risk?
We’re in a hype cycle, and agentic systems still hallucinate. Standards are emerging—like A2A (Agent-to-Agent by Google), MCP (Model Context Protocol), UTCP (Universal Tool Calling Protocol), and Agent Registries like NANDA,, but they’re not yet fully baked. We’ll need these to ensure interoperability and trust.
Security is another issue. Today’s cybersecurity teams protect a building. With agentic systems, it’s like defending an entire city. Attack surfaces multiply.
When will this be real?
Some early use cases are already live, but full-scale, consumer-facing implementations? In retail and DTC, we’ll see those for Christmas 2026. In the meantime, organisations should focus on preparation: clean data, streamline processes, remove exceptions, and audit security layers.
So, what, if anything, should retailers do now?
Start experimenting! Small retailers should explore plug-and-play tools like Shopify’s integrations or SDKs from startups like Data Sapien. Large retailers should audit internal systems, prepare product data, and consider where human-in-the-loop workflows still matter.
Think of this data problem as much as a technology one. If your catalogue, pricing, or categories are messy, your agents will perform poorly.
Organisational impact: what changes?
Retail org structures are built around human limitations. Category managers, buyers, merchandisers—they’re siloed because humans can only manage so much. But an agent can handle all categories, all the time.
Forecasting, stock management, procurement—agents will outperform humans in scale and speed. That doesn’t eliminate humans, but refocuses them on creative and strategic tasks.
Is agentic AI a distinct overlay, or will it be embedded and provided by major tech providers?
It’ll be both. Shopify may offer agentic features as overlays. Bigger retailers may build bespoke agents or consolidate existing tools. Walmart has been working on this for some time. Agentic functionality will emerge in platforms as SDKs and embedded features, depending on scale and technical maturity.
What is the role of standards?
Standards like MCP act as the new API layer, allowing agents to safely and securely access backend systems. These protocols describe how agents authenticate, what data formats to use, and how to interact with functions like order creation or returns.
But standards take time. Remember how long EDI took to mature? Until these stabilise, most likely mid-2026, true interoperability remains limited.
A new retail paradigm then?
Today, customers choose from existing supply, but agents can generate demand for things that don’t yet exist. Imagine thousands of agents requesting a pink laptop. Retailers can monitor that demand and decide whether to fulfill it. That’s a game-changer.
Final thoughts: humans still matter
Agents are great at tasks, but they lack creativity, empathy, and common sense. In complex emotional situations, only humans can respond with the care and the needed nuance. So, let agents handle the routine, and free up humans to be brilliant at what only we can do.
About Gam Dias
Gam Dias is an AI and Data Strategist, Enterprise Applications Product Manager, and Customer Empowerment advocate. Gam is an adjunct professor at IE Business School in Madrid where he teaches Data and AI Strategy on executive programs, has been a trusted advisor to IBM, Walmart Labs, and Applied Materials, and is the author of Agents Unleashed and the Data Mindset Playbook. Read Gam’s blogs on agentic:
Milo and the Retail Marketplace
Want to know more?
Join our upcoming Digital Café with Gam Dias to unpack all things agentic AI!
When? 11th September 2025 (12pm GMT)
Where? Taking place on Zoom, so you can dial in from the comfort of your home or desk!
Who? All of our Cafés are informal, retailer-only sessions. Get involved below…
