Supply chain operators: Here's how to make AI and data work for you

Most logistics businesses are not short on data. They have transport management systems, warehouse management systems, ERPs, CRMs, spreadsheets, CSV exports, and sometimes handwritten con notes still making their way through the back office. 

The data exists. None of it talks to anything else, and that disconnect is costing more than most operators realise.

Take proof of delivery. A customer emails asking for a copy and someone in the office needs to search the system or local records to find a document, verify it, and reply. It's a simple enough task on its own. 

However, when that process happens dozens of times a day, across a team, it adds up fast. Some businesses are spending close to $100,000 a year just to have someone completing that one task alone. 

This is the kind of challenge and inefficiency I recently spoke with James Kerr about on the Supply Run Podcast by UTenant. It's an area that most conversations about digital transformation in logistics skip over. Everyone wants to talk about AI. Very few want to talk about the condition of their data.

The Integration Problem No One Wants to Own

Every logistics operator I speak with runs some version of the same operation. 

They have a TMS that does some things well, a WMS that handles others, a CRM sitting off to the side with customer information that never quite lines up with what is in the TMS, and Excel spreadsheets bridging the gaps that nobody has gotten around to fixing. 

The tools are fine individually, but together they create a mess of information where the data you want to get at is buried and takes valuable time to mine out.

The instinct when things get unwieldy is to look for a better piece of software. Sometimes that is the right call, but more often, the answer is a layer of connection above the tools already in place. Use the best of what is out there and build something that brings the data into one place and does something useful with it.

Think of it as a control layer rather than a new system. The CRM stays where it is, the TMS stays where it is, but there is something sitting above them that connects the data, automates the repetitive work, and gives you actual visibility over what is happening in the business. 

Knowing how long it takes to get a full truck's load to the dispatch area, or how long a driver waits at a delivery location before they can unload, that kind of operational intelligence is available when the data is connected. Most businesses are generating it without ever being able to see it.

The harder version of this problem is when a TMS is so dated that it offers no way to connect to it at all. No API, no data export, no integration point. Those systems exist, and they are more common in transport than most people expect. 

For organisations that find themselves in this position, they sometimes recognise that the only practical path forward is building a custom service that connects directly to the underlying database and extracts what is needed.

This can be a great approach, however, there is a version of this conversation that goes off the rails quickly. A business hears about what AI can do and decides to try and solve everything at once. They hand over a list of seven problems and want all of them addressed. The instinct is understandable, but the approach tends to fail.

AI needs good data to work. If you want to automate how a truck is loaded, making sure heavy items are not stacked on fragile ones or that pallets reach the dispatch bay in the right sequence before the truck arrives, the AI needs historical records of how that loading has been done. 

Without that data in a usable format, the system can only guess. In logistics, a bad guess tends to show up as damaged goods or a truck that leaves half empty.

The businesses that get genuine value from AI are the ones that start with a specific, contained problem, connect the relevant data, and measure what changes. Fix one thing, prove the return, and the next investment becomes straightforward. The savings from the first project tend to fund the second one.

Where the Wins Are

The practical starting point is almost always administration and communications. An AI agent that watches incoming emails, identifies a delivery status enquiry, finds the consignment note, checks the GPS position of the truck, calculates an estimated arrival time, and drafts a reply. 

That is happening in logistics operations right now. The customer support person reviews the draft and sends it. What took five minutes of manual lookup happens in seconds, and one person can handle ten times the volume.

That same logic extends to warehouse operations. A forklift operator who knows a truck is 30 minutes out and has a sequenced list of pallets to stage in the dispatch area can have everything ready when the truck backs in. The truck loads and leaves. The difference between that and an operation where the driver waits 45 minutes while stock is being located comes down to whether the data is connected and the workflow is designed around it.

These wins compound. Four hours saved here, eight hours saved there, and it adds up. The business builds a track record that makes the next improvement easier to justify.

When businesses come to us about a digital project, they ask about cost, timeline, and what the system will do. The question almost nobody asks is what does 'finished' look like? The honest answer is that it will not be finished. That is just the reality of how this technology moves. 

Models that could not solve certain problems six months ago can solve them now. And even then, the ones available today will be superseded by something more capable before the year is out. A business that treats its data infrastructure as a set-and-forget installation will fall behind the one that treats it as an evolving asset.

That does not mean an endless, expensive build. It means having a foundation that can accept new capabilities as they become available, and a practice of looking at the operation regularly to ask where the next improvement can come from.

Where to Start

The most useful thing a logistics operator can do before talking to any technology provider is spend time looking honestly at where time and money are going. That means identifying where people are doing repetitive manual work, where errors happen most, and where information gets lost between systems.

That analysis tends to surface one or two problems that are both high-cost and straightforward to address. Solving one thing well and proving what it returns builds the foundation for everything that follows. 

Staff benefit because their days get easier. Owners benefit because the numbers improve. The next investment becomes a decision based on demonstrated return, not theoretical promise.

The data is already there. The question is whether it is working for your organisation.