ABI Research and FourKites surveyed 490 supply chain leaders to learn why companies are missing out on supply chain tech ROI
Key Points:
Earlier this year, MIT gave the much-hyped AI industry a harsh dose of reality. Their research found that only 5% of AI pilot programs achieve rapid revenue acceleration. The vast majority stall, delivering little to no measurable impact on P&L.
Fortune reported that “MIT’s research points to flawed enterprise integration. Generic tools like ChatGPT excel for individuals because of their flexibility, but they stall in enterprise use because they don’t learn from or adapt to workflows.”
We found something similar. Shippers are racing to deploy AI in their supply chains, but they’re mostly focusing on things that don’t move the financial needle.
ABI Research and FourKites surveyed 490 supply chain leaders, revealing that decision support and customer service are the most common use cases for AI, followed by demand forecasting and inventory management. Those are planning and communication use cases — better forecasts, faster customer responses, cleaner dashboards for leadership. Meanwhile, far fewer leaders say they’re applying AI to automated issue resolution or risk management, aka the places where money gets stuck.
Most teams are still using AI to predict or report. Fewer are using it to resolve problems that slow cash, inflate inventory, or erode margin. It’s no surprise that so few shippers are seeing an impact on their financial statements.
When inventory is late, stuck, misrouted, or quarantined, it still sits on your balance sheet. You’ve already paid for it, but it’s not generating revenue. When there’s a delivery issue, a quantity mismatch, a labeling error, or missing proof of delivery (POD), customers delay payment or take deductions. Cash you expected this month slides into next month.
Similarly, every time a load misses a window and you chase it with premium freight or change allocation to another DC, the fix shows up in your cost-to-serve. That pressure hits gross margin first, and pricing power rarely covers it.
For many shippers, these are recurring, predictable balance sheet impacts. They’re just spread across transportation, logistics, AP/AR, and customer service, so they don’t always get treated as a single financial problem.
Part of the challenge in resolving these issues is that ownership is unclear. A late shipment that triggered a retailer fine, plus a deduction, plus a payment delay — who owns that? Logistics? Customer service? Finance? Sales? AP? Legal? Too often, nobody is watching the full chain end-to-end.
To complicate matters further, the data is messy. Resolving a claim or rate dispute requires freight audit data, temperature readings, appointment logs, signed PODs, rate tables, and contract terms. This is not one clean dataset.
So AI gets pushed toward areas that are easier to pilot and easier to defend internally, even though those areas may not be where the cash problems live. The consequence is that the same issues continue to drain working capital and margin, quarter after quarter.
Large shippers and retailers routinely have ongoing disputes with carriers, 3PLs, and suppliers over accessorials, detention, and OS&D. Those disputed invoices sit unapproved, which delays payment. On the other side, when you owe a customer a credit, AR can’t close cleanly until the details are settled. It’s not unusual for this unresolved bucket to sit in the low single-digit millions at any given time for a multi-billion-dollar manufacturer. That’s real to treasury and working capital planning.
Large customers take deductions for missed appointments, bad labeling, and incorrect ASNs. Sometimes valid, sometimes not. The longer it takes to assemble documentation, the harder it is to challenge the deduction, and companies end up writing it off. Across a year, that becomes millions in preventable margin erosion for brands selling into big-box retail or grocery.
Then there’s inventory that’s technically “on hand” but not actually available. You have product on a truck in the yard, or in receiving waiting on QC, or in the wrong DC. The system shows stock, but customer service can’t promise it. That drives last-minute expediting, split shipments, and elevated working capital because you buy or divert more inventory than you technically needed. You’re paying to hold inventory and paying again to go get more inventory because the available-to-promise picture is wrong.
This isn’t going to sink the company, but it’s a steady drag on cash generation and margin every quarter.
AI can gather the documents a human would normally chase across TMS, WMS, carrier portals, and email — POD, temperature trace, check-in timestamp, contracted rate. Then it generates a clean case file for AP, AR, or a retailer portal. Instead of a week of back-and-forth, you’re at the point of resolution in hours.
When a high-priority shipment is at risk — temp excursion, truck delay, customs hold — AI can flag it and help with the next move: reroute to a closer DC, alert customer with revised ETA plus mitigation plan, escalate unloading priority at the facility. You reduce downstream penalties and avoid having sales negotiate credits after the fact.
Ultimately, it’s about cutting the cycle time between “there’s a problem” and “the problem is closed.”
Instead of pitching “AI for supply chain,” frame it in terms of measurable frictions. How many days do your open transportation and invoice disputes sit unresolved? How much is sitting in deduction review with your top three retail customers? What percentage of your inventory is physically in network but not available-to-promise? How much are you spending on premium freight as a share of total freight?
Fund AI that reduces those numbers directly. Customer service chatbots demo well, but they don’t move cash the way dispute closure and risk prevention do.
The industry has spent a decade trying to get better at forecasting demand, and that work should continue. But the real strain on cash right now isn’t only “did we forecast correctly?” It’s “did we execute cleanly enough that we actually got paid on time, avoided penalties, and didn’t have to carry more stock than we can afford?”
Going back to the MIT research, they also found that the biggest ROI came from back-office automation — eliminating business process outsourcing, cutting external agency costs, and streamlining operations. That tracks with what we’re seeing. AI is finally good enough to sit in those messy handoffs like shipment to invoice, invoice to payment, or load to claim, and shorten the distance between problem and resolution. That’s where the 5% who are seeing real P&L impact are deploying it.
Want to learn more?
