From visibility to execution: the next frontier of supply chain AI

This article draws on reporting by Peter MacLeod, originally published in Logistics Business and featured in the May 2026 edition of Logistics Business magazine.

The artificial intelligence (AI) conversation in supply chain has grown up. The question is no longer whether AI belongs in supply chain operations. It's whether it’s actually doing anything useful once it gets there.

Supply chain leaders aren’t arriving at these conversations with curiosity anymore. They're arriving with expectations. They want to know what AI is delivering inside real operations, not what it might deliver in theory. For Aadil Kazmi, Head of AI Product Development at Infios, it's the only distinction that counts. “Everyone is definitely interested in AI,” he said. “The conversations we’re having are really around two themes: how we’re bringing AI into our products and services, and how those capabilities actually empower people rather than replace them.”

Why embedded AI is different

There's a version of enterprise AI that sits alongside your operations: a new dashboard, a separate interface, a tool your team has to remember to open. And then there’s embedded AI, which works where your people already work.

Infios has made a deliberate choice to build the second kind. “We build agents that work where people are already used to working,” Kazmi explained. “People don’t have to learn a new screen or change their workflows. The agents meet them where they already work.”

In an industry where operational systems are deeply entrenched and change management is notoriously slow, reducing friction isn’t a nice-to-have. It’s the difference between adoption and abandonment.

Equally important is the question of control. AI agents at Infios are designed to augment human decision-making, not circumvent it. If an agent encounters an unfamiliar scenario or crosses a predefined threshold, it escalates to a human operator, much like an exception being flagged for manual handling in a warehouse. “Our agents aren’t looking to automate the full human,” Kazmi said. “Quite the opposite. We take a very ‘human-in-the-loop’ approach, where the person is always in control.”

From chatbots to agents: what actually changed

Part of the challenge in any AI conversation is language. Terms like LLMs, AI and agents get used interchangeably, which creates confusion and erodes trust.

“In the beginning, we had simple LLMs delivered as chatbots,” Kazmi explained. “You ask a question; it gives you an output. It’s not doing anything more than that.”

These early systems were limited to their training data. They couldn’t interact with live operational environments, trigger workflows or take action in the real world. AI agents represent a meaningful step forward. “An agent is really an LLM with access to tools and reasoning loops,” he said. “Tools allow it to interact with the real world: fetching order statuses, triggering workflows, even communicating externally. And reasoning loops allow it to gather information, reassess and improve its output.”

The difference is significant. An agent isn’t just answering questions. It's participating in processes.

That's also what separates today’s deployments from what came before. When a shipment goes dark with no GPS signal, no electronic data interchange (EDI) and no carrier communication, a chatbot reports the problem. An agent understands the downstream impact across orders, warehouse tasks and customer commitments, and starts resolving it. In real deployments, that gap between detection and recovery has shrunk from hours to minutes.

Three outcomes operators actually care about

The technology is evolving fast, but the business case ultimately comes down to what can be measured. Based on current deployments, Infios identifies three areas where AI agents are delivering tangible value.

1. Increased capacity

   Many customers couldn’t perform check calls across all loads because it was too expensive and time-consuming. With AI agents handling driver check calls, they can now achieve full coverage without adding headcount.

2. Lower cost to execute

   One customer deployed Infios’s order entry agent and now processes inbound orders in seconds versus what previously took upwards of 10 minutes manually. The cost of executing that workflow fell enormously.

3. More consistent customer experience

   With agents handling check calls, order entries, reporting and exception detection, the end customer receives a more reliable, predictable experience across every interaction.
   

These aren’t isolated efficiency gains. They point to something larger: AI is beginning to reshape the economics of supply chain operations, not just optimize tasks within them.

The real cost of slow response

Supply chains have always absorbed disruption. But the frequency and scale of that disruption has shifted fundamentally. The numbers are stark: $184 billion in annual costs per large company attributed to supply chain disruption. A billion-dollar weather event every two weeks. The Suez Canal corridor moving $9.6 billion in goods daily when it’s open. A tariff announced on a Tuesday that reroutes $34 billion in trade by Friday.

These aren’t outliers anymore. They're the baseline. What’s changed is how fast organizations need to respond, and how much it costs when they can’t. According to a 2024 IDC study, 83 percent of supply chains can’t respond to disruptions within 24 hours, with an average response time of five days. And when disruptions do hit hard, McKinsey data shows they can cost businesses up to 45 percent of a year’s profit over a decade. That model doesn’t hold up when disruption is constant.

"What AI has been able to do is supercharge teams' abilities to sense disruptions, decide what to do and act in real time," Kazmi said. But speed of response depends on a foundation being in place first. "The precursor to all this is integrating agents into the systems where your data lives, and deploying them safely and reliably." Without that integration layer, even the most capable AI has nothing to act on.

Visibility was the goal. Execution is the opportunity.

For the better part of the last decade, the dominant priority in supply chain technology was visibility: knowing where goods are, what’s happening across the network, getting signal out of complexity. That work is largely done. Most organizations have crossed that threshold.

“Visibility was probably one of the most important vectors to manage over the last decade,” Kazmi said. “Most organizations have already overcome that hurdle.”

The new problem is different. “What good is visibility if you can’t act on it? That’s the gap we’re filling.”

This is what Infios calls Intelligent Supply Chain Execution™: a model that connects orders, warehouse operations and transportation into one coordinated system, with AI embedded directly into the workflows that drive decisions and actions. Predictive AI anticipates disruption before it hits. Generative AI turns fragmented signals into clear context. Agentic AI takes action across systems to resolve issues. Conversational AI makes all of it accessible to every operator, not just technical teams.

The result isn’t a smarter dashboard. It's a different way of operating: sensing what’s happening, deciding the best response, acting immediately and learning continuously.

AI as a co-worker, not a replacement

Successful AI adoption is as much about people as it is about systems. The organizations getting the most from AI aren’t treating it as a replacement for human judgement. They're treating it as an extension of the workforce: one that handles repetitive or time-sensitive tasks while humans focus on higher-value decisions.

That framing requires careful thinking about roles, responsibilities and trust, and it places real responsibility on technology vendors to make the integration seamless. “Partnering with a vendor that integrates deeply into your stack is key. You need to bring the agents right to where your teams are already working to reduce the friction of change management.”

That's not just a vendor pitch. It's a design principle that determines whether AI actually gets used, or sits unused alongside the systems your team already relies on.

What comes next: context and memory

Current deployments are delivering measurable return on investment (ROI). But the next step is more ambitious. Kazmi believes the next meaningful leap in enterprise AI will come from improvements in memory and context management.

“The next jump will come from bringing deeper context to these agents,” he said.

Context, in this sense, goes beyond raw data. It’s the situational awareness, priorities and accumulated reasoning that underpin how experienced humans make decisions. embedding that level of understanding into AI systems could move the technology from reactive assistance toward proactive, strategic support: anticipating what matters before it becomes urgent, and acting on it accordingly.

The window is open, but not indefinitely

For an industry long defined by complexity and constraint, the convergence of AI, execution capability and connected data represents a genuine inflection point. Not another wave of innovation to track. A structural shift in how supply chain operations work.

“The world is no longer predictable,” Kazmi said. “Execution is becoming trickier and trickier to manage.”

In that environment, the advantage isn’t visibility. It's the speed and intelligence of your response. The organizations that act now, fixing one high-impact workflow, then another, building from there, will enter 2028 with trained models, proven workflows and a meaningful operational lead.

AI can’t prevent the next Suez Canal blockage. But it can fundamentally change how you respond when it happens. As Kazmi puts it: “It allows us to bring visibility and transparency into the promises that we make every single day.”

That’s what Intelligent Supply Chain Execution™ is built for. Execution without interruption.