91探花

A third of US companies have integrated AI into their supply chain operations 鈥� yet nearly three quarters of them still cannot track a shipment in real time. This mismatch exposes a truth about current AI investment: the software layer is sprinting ahead while the physical infrastructure lags 鈥� and that widening gap is where the money is disappearing.

The adoption numbers look encouraging in isolation 鈥� 36% integration across US companies in 2026 represents a huge shift from where the market was five years ago, when supply chain AI was largely the domain of large retailers and automotive manufacturers with the technical capacity to build and implement it. The tools have become cheaper, more accessible and better at integrating with existing systems. Vendor pitches have kept pace 鈥� promising real-time visibility, predictive disruption alerts, automated reordering, optimised routing and demand forecasting that actually works.

The 72% visibility discrepancy is the part that doesn鈥檛 fit that narrative. If AI supply chain tools were delivering on their core promise, the first thing you鈥檇 expect to see is companies gaining visibility into where their goods are and when they鈥檒l arrive. That鈥檚 the most basic operational requirement, and it鈥檚 the one most directly addressed by the technology. The fact that nearly three quarters of companies are still operating without it suggests the problem isn鈥檛 the software itself but the infrastructure the software is trying to work with.

听

The Physical Problem That Software Can鈥檛 Fix

听

Real-time shipping visibility requires data from every point in a shipment鈥檚 journey: the warehouse management system, the carrier鈥檚 tracking infrastructure, the port or border crossing, the last-mile delivery provider. In most supply chains, these systems don鈥檛 talk to each other. Some carriers use APIs, some use EDI, some use email and some use phone calls.

International shipments cross multiple carriers, multiple customs systems and multiple data standards before they reach their destination. An AI visibility tool placed on top of that fragmented infrastructure isn鈥檛 solving the data problem, it鈥檚 trying to compensate for it.

This remains the fundamental structural issue that supply chain AI investment has talked around 鈥� instead of solving. The vendors selling predictive analytics and demand forecasting tools need clean, complete, real-time data to make predictions worth acting on. If the underlying data is patchy, delayed or missing for entire segments of the journey, the predictions are built on partial information and the visibility gap persists regardless of how sophisticated the model is. Garbage in, garbage out still applies 鈥� it arrives in a much more expensive package.

The companies that have achieved real-time visibility have typically done it the hard way: by standardising carrier relationships, building or buying integration layers that clean and consolidate data across multiple sources and investing in the unglamorous work of data engineering that makes the visibility tools function correctly. That鈥檚 a different category of investment from buying an AI supply chain platform, and it鈥檚 one that most vendors aren鈥檛 selling.

We asked supply chain operators, logistics leads and AI practitioners who have deployed these tools to say what they鈥檝e actually delivered in practice.

听

Our Experts

听

听

• Nishith Rastogi, CEO and Founder, Locus
• Jim Bureau, CEO, Loftware
• Michelle Northey, Chief Product Officer, Loftware
• Josh Medow, CEO, Mercury
• Roger Bible, Director of Operations, ATC Driveaway
• Vitaly Koval, Co-Founder, GoGloby

听

听

Nishith Rastogi, CEO and Founder, Locus

听

听

鈥淭he gap between AI adoption and real-time visibility reflects a structural issue in how many retailers approach transformation. Too often, AI is layered onto fragmented carrier networks and legacy systems, with the expectation that it will compensate for inconsistent or delayed execution data. In practice, it cannot.

鈥淚n omnichannel environments, this challenge is most visible in last-mile delivery. Retailers are promising tighter delivery windows and faster fulfilment, but the execution layer 鈥� carrier integrations, store dispatch and real-time tracking 鈥� remains uneven. When shipment updates are delayed or exceptions are flagged too late, AI-driven insights become retrospective rather than actionable.

鈥淭his disconnect is already showing up in performance. Our research found that only 7% of businesses consistently meet fast delivery promises. At the same time, brands offering shorter delivery windows see a 104.5% increase in delays compared to those using wider windows 鈥� highlighting how fragile execution becomes without real-time control.

鈥淲hat retailers actually need is technology that closes the loop between planning and execution. That means ingesting live data across carriers and stores, understanding constraints in real time and enabling immediate intervention. AI delivers real value only when it is directly connected to execution. Without that, retailers are investing in better predictions about disruptions they still cannot prevent.鈥�

听

Jim Bureau, CEO, Loftware

听

听

鈥淢any companies are stuck in a pilot phase because AI is only as good as the product data, connected partner networks and processes behind it. In supply chains, fragmented systems and disconnected trading partners make scaling AI difficult. The organisations succeeding are embedding AI into core operational workflows and connected supply chain networks 鈥� not treating it as a side experiment 鈥� because real business value comes from execution, visibility and collaboration at scale.鈥�

听

Michelle Northey, Chief Product Officer, Loftware

听

听

鈥淔or many companies, the challenge isn鈥檛 a lack of AI ambition but a lack of data readiness. Decades of investment in disparate systems have created data silos, inconsistent standards and fragmented processes that make it difficult for AI to access reliable, contextualised information. The biggest roadblocks are poor product data quality, limited interoperability between systems and the absence of strong data governance practices. As companies look to scale AI initiatives, digitisation and effective data management will become critical 鈥� because AI can only deliver business outcomes when it鈥檚 built on accurate, standardised and trusted data.鈥�

听

Josh Medow, CEO, Mercury

听

听

鈥淢any companies are asking AI to solve a problem it was never designed to fix. In the world of shipping logistics, AI can identify patterns, synthesise information and enable teams to make decisions faster. What it cannot do is create reliable, real-time visibility when faced with incomplete, delayed or inconsistent data. If suppliers and carriers aren鈥檛 providing accurate information, AI will be producing predictions from inaccurate inputs.

鈥淲hen you鈥檙e dealing with temperature-sensitive shipments, an incorrect move isn鈥檛 just an inconvenience 鈥� it can jeopardise a clinical trial. After learning that broader AI models were surfacing outdated internal procedures alongside current ones, Mercury deliberately limited AI鈥檚 role. Today, we use AI to help teams efficiently access valid information and summarise shipment status, while employees remain responsible for operational decisions and exception management.

鈥淐ompanies using AI to accelerate human expertise, while investing in the people responsible for addressing disruptions, have the real competitive advantage.鈥�

听

Roger Bible, Director of Operations, ATC Driveaway

听

听

鈥淪upply chain visibility issues aren鈥檛 about data scarcity 鈥� they鈥檙e about data availability versus usability. Most often, we have a surplus of data. The difficulty lies in acquiring accurate and up-to-date information from various sources and synthesising it into an action-ready format that operational staff can readily use. Carriers, customers, manufacturers and tech platforms all operate on their own unique systems, and weaving these together continues to be the hard part.

鈥淎I can detect trends, automate repetitive tasks and predict more effectively, but it鈥檚 not a substitute for well-defined processes, dependable communication channels or superior data quality. In my experience, technology creates the biggest positive change when it streamlines operations 鈥� giving transportation teams visibility of assets, flagging risks early and speeding up decision-making. The best technological solutions support experienced people rather than seek to replace them.鈥�

听

Vitaly Koval, Co-Founder, GoGloby

听

听

鈥淓verybody bought the same promise: layer AI on top of your existing supply chain and get real-time visibility. What most companies got instead was a more expensive version of the same blind spots.

鈥淭his isn鈥檛 a smarter model problem. It鈥檚 an integration problem. Carrier data is fragmented, tracking systems are outdated, supplier reporting is inconsistent 鈥� and a prediction model sitting on top of that doesn鈥檛 fix any of it. It gives you a number pointing at faulty data with confidence.

鈥淲e see the same pattern across engagements. Companies get the pilot approved, and then the real gaps become apparent in production: in the integrations, in who actually controls the system, in all the ways a demo never shows you. AI doesn鈥檛 add value by sitting next to your ERP. It adds value once it鈥檚 genuinely wired into it.

鈥淲e recently helped a PE-backed industrial ERP platform replace a ten-person legacy team with five embedded engineers focused entirely on integration work. Output increased 3.6x. Supply chain operators don鈥檛 need another AI tool. They need the integration work done first.鈥�

听

For any questions, comments or features, please contact us directly.

听