The boards are asking for an Agentic AI strategy. The vendor pitches are everywhere. And at NRF, the melodrama was dialed to eleven. Every conversation at the conference was framed as if retailers who haven’t deployed AI agents are already circling the drain.
Here’s the confession you won’t hear in those vendor pitches: most retailers are not ready. They don’t understand what Agentic AI actually is. They can’t evaluate it properly. And they are making the exact same mistakes they made with every previous technology wave, i.e. waiting too long, picking the wrong partners, or worse, rushing into pilots without a real strategy.
But here’s what else is true.
It is not too late. Not even close.
In our latest episode of Confessions of Supply Chain Executives, Omar Akilah, SVP of Product at Infios, and Aadil Kazmi, Head of AI Product Development at Infios, a team working with over 5,000 customers across 70 countries, laid out the most practical, no-hype roadmap for Agentic AI in retail commerce I’ve heard.
And, here’s what every executive should take from the discussion.
First, Get the Definition Right
Before you can evaluate whether you’re ready, you need to understand what Agentic AI actually is because most executive-level definitions stop one step short.
Generative AI was the unlock that allowed machines to understand unstructured data: natural language, conversational text, the kind of information that previously required a human brain to parse. That was transformative. But Agentic AI goes further.
It adds action.
As Kazmi described it: Agentic AI is the combination of sensing, reasoning, and, most critically, acting.
The old call center robot could press one or press two. Today’s systems can understand intent and execute a response. They can modify an order, reroute a shipment, reschedule a delivery. It is not the difference between good and better. It is the difference between a recommendation and a result.
Supply Chain Execution Is the Perfect First Target
There is a debate in retail about where Agentic AI delivers first. At least there is for me anyway.
Customer-facing applications or backend operations?
The answer, as both guests made clear, is ultimately both. But execution is where the ROI is fastest and most measurable, and the reason comes down to three structural realities.
First, supply chain execution has never been more volatile. Tariffs, port disruptions, labor shortfalls, carrier failures, etc. are all variables that can blow up a customer promise. Second, most retailers have already made the infrastructure investments to capture the data. The data is there. What’s missing is the ability to act on it in real time. Third, the use cases are concrete and the metrics are clear in a way that’s harder to achieve on the customer-facing side.
Kazmi illustrated this with a scenario that every supply chain leader will recognize.
For example, a shopper checks out on a Friday with an estimated delivery of Wednesday. Somewhere between that checkout and the doorstep, a freight driver misses a connection. In the old world, that downstream impact, a broken promise, an angry customer, a manual exception to manage, unfolds slowly and often invisibly. With an Agentic AI framework, an agent detects the anomaly, reasons through the downstream implications, and acts, deciding to reroute the order from a different warehouse, proactive notify the shopper, and reconfigure the delivery promise before it fails.
That is the promise. And it is already happening.
The Readiness Bar Is Lower Than You Think But the Real Barrier Is Culture
Here is where both guests said something that is going to be controversial, and that every executive should hear clearly: you do not need a unified data lake or a single data model to get started with Agentic AI.
Modern large language models are capable enough to make sense of what “order status” means across disparate tables, systems, and definitions without requiring months of data harmonization work first. The infrastructure bar is genuinely lower than the last three technology waves demanded.
But that does not mean the path is simple.
The real barrier, as Kazmi put it, is internal. It is culture. It is change management. It is the organizational clarity to answer three questions: What problems are we solving? What is the process for solving them? Where does AI fit into that process?
Companies that skip that work and go straight to technology deployment are the ones who end up in pilot purgatory.
The Three Options And Why Most Retailers Will Choose Wrong
When it comes to actually adopting Agentic AI, Kazmi outlined three paths:
Option 1: Build it internally. Stand up a dedicated AI team from scratch. Realistic timeline: 12 months or more. Cost: significant. Risk: high, because you are building both the capability and the institutional knowledge simultaneously.
Option 2: Buy off-the-shelf AI tools. Fast deployment, immediate surface-level ROI, and a near-certain ceiling. Horizontal, generic tools cannot handle the specificity of your order life cycle — the multi-agent, interconnected workflows that let you actually operate a supply chain autonomously. This is where pilot purgatory lives.
Option 3: Partner with a domain-expert vendor. A vendor that brings not just the Agentic AI tooling but deep expertise in your specific industry, one that can help you define the strategy, redesign the workflows, train the models to your operation, and connect individual agents so they communicate horizontally, not just vertically.
The nuance Akilah added is important: even with the right vendor partner, retailers should be running a parallel track to upskill their internal teams. The goal is not permanent dependency. The best-case outcome is that after 12 months, you have both a working Agentic AI operation and an internal team that understands how to run it.
Therefore, worst case, you still have the skills in-house if the partnership changes.
The 30-Day Roadmap
For supply chain executives ready to move, Akilah offered a refreshingly simple 30-day framework:
Get clear on strategic outcomes before you touch a single technology. Identify the two or three areas where you are hurting most, e.g. labor volatility, order visibility gaps, carrier management failures, etc., and define the ROI of solving them. Then look for the lowest-hanging fruit: the menial, repetitive tasks that agents can absorb immediately and that free your best people for the decisions that actually matter.
The instinct to overcomplicate this is the enemy. The answers, as Akilah put it, are usually right in front of you.
The Confession: Automate Visibility First
When pushed to name the single highest-impact action a retail executive could take after hearing this conversation, both guests converged on the same great answer: automate visibility across the order life cycle.
Map every step from checkout to delivery. Identify every handoff, every point where a human, a carrier, a warehouse, or a system touches the order. Deploy agents at each of those dependent steps. Then connect them so they communicate. The result is a unified, always-on system that knows where every order is, can detect anomalies before they become broken promises, and can answer any customer, through any channel, with a real answer.
The downstream metrics will unveil the story: you will get a real sense of all the hours spent on inbound “where is my order” calls, customer satisfaction scores around delivery, and the staff hours currently devoted to manual configuration and workflow management.
All of them move, and move fast, when visibility is automated.
The Uncomfortable Truth
The boards are asking the wrong question. “What is your Agentic AI strategy?” is not the right frame. The right question is: “What strategic objective do you need to achieve, and how does Agentic AI unlock it?”
Retailers who approach this as a technology problem will get a technology answer (and a bill with little to show for it). Retailers who approach it as a strategy problem, with AI as the new unlock, will look back in two years and wonder how they ever operated without it.
Watch the full Confessions episode with Omar Akilah and Aadil Kazmi of Infios to hear their complete framework for Agentic AI readiness, including the real questions to ask any vendor before you sign that next contract.
Available wherever you get your podcasts:
Apple Podcasts | Spotify | Soundcloud | Amazon Music | YouTube
Be careful out there,
– Chris, Anne, and the Omni Talk team
Music by hooksounds.com
*Sponsored Content*
Omni Talk® is the retail blog for retailers, written by retailers. Chris Walton and Anne Mezzenga founded Omni Talk® in 2017 and have quickly turned it into one of the fastest growing blogs in retail.