Rachelle Butler
A version of this piece originally appeared on JBF Consulting’s blog: https://jbf-consulting.com/what-buyers-need-to-understand-before-they-invest/
As a logistics consultant, it is critical to stay close to what is happening across log-tech. Given the sheer number of vendors we monitor, I enlisted the help of ChatGPT Tasks to run a daily scan and notify me of new product announcements and release notes across many different providers.
One of those scans flagged a vendor introducing what it called a “new AI feature.” I read further, and the functionality sounded familiar. When a predefined condition occurred, the system triggered a prescribed response. That’s rules-based automation and is far from new.
Labeling this feature as AI implies it is able to learn, adapt, and adjust on its own, but in reality, a configured trigger produced a configured response.
The label changed, but the technology did not.
As someone who’s been on the other side of it, I’m sensitive to this distinction. Earlier in my career, in a product leadership role, I watched the AI hype cycle begin the way these things usually do: specific questions started showing up in RFPs. Sales asked me to document which AI capabilities were native to our platform and what was on the roadmap. We’d identified real use cases where AI made sense, but we were solving business problems, not adding AI for its own sake.
In my first draft of that RFP response, I described our optimization engine — sophisticated, heuristic-driven, and native to the platform for over 20 years — as an AI capability. Our development lead read it and pushed back immediately. The optimizer wasn’t learning or adapting. It was mathematical optimization, and we couldn’t represent it as anything else.
I think about that conversation often, because I now see the reverse pattern across the market. Automation and optimization that have existed for years are being repositioned as AI, sometimes because AI genuinely extended the capability, and sometimes because the terminology moved while the technology stayed put. Buyers rarely have an easy way to tell which is which from a demo or a spec sheet.
Sophisticated Log-Tech in Not New
Log-tech has automated work and supported complex decisions for decades, long before AI entered the conversation.
Rules-based automation runs logic that people define. A condition is met, and an instruction fires. The workflow can be layered and can automate a meaningful share of a process, but it’s still operating inside logic someone configured in advance.
Optimization has an equally long history in TMS and WMS platforms. A TMS optimizer evaluating how to consolidate and route shipments — balancing cost, service level, and constraints — can weigh far more combinations than a planner could manually assess, which is why the output looks intelligent. Traditional optimization typically leans on mathematical programming or heuristics to get there. A heuristic earns its place when the solution space is too large to solve exactly in a useful amount of time; it trades mathematical perfection for a strong, feasible answer on a deadline.
None of that requires learning from prior plans or interpreting unstructured information. It’s solving the problem it was built and configured to solve, which is extremely valuable in its own right.
Accurate language doesn’t diminish value. However, it does set the right expectations for what the system will and won’t do on its own.
Three Capabilities That Actually Meet the Bar
When something genuinely is AI, it tends to fall into one of three buckets, and each comes with a different maintenance profile:
Machine Learning (ML) finds patterns in historical data and applies them going forward. For example, improving an ETA prediction, or estimating whether a carrier will accept a load. Maintaining it means monitoring the underlying data and watching for drift as conditions change and the model’s accuracy degrades.
Generative AI interprets and produces content. In logistics that could mean: explaining why transportation costs spiked or summarizing what’s driving an exception. Its output quality depends on the model, the instructions it’s given, and what information it can access, all of which need ongoing testing as the business context shifts.
Agentic AI pursues an objective across multiple steps and may take action along the way. This includes investigating an exception, gathering information, and executing part of the resolution. This one extends maintenance furthest, into system access, permissions, integrations, escalation logic, and performance controls.
These often work in combination, and alongside components that aren’t AI at all. An agent might use an ML prediction to flag a likely delay, call a traditional optimizer to weigh response options, and apply predefined workflow rules to decide whether the chosen response needs human sign-off. Calling that whole chain “AI” obscures which piece is doing what, which is problematic when each piece carries different failure points and different upkeep.
Follow the Work, Not the Label
Take a shipment likely to arrive late. The technology might predict the delay, recommend a response, or act on that recommendation directly by securing alternate capacity or updating the plan.
Following the work means tracing not just what happens next, but what produced it. An ETA built only from historical shipment data is a different animal than one that also factors in current location and live operating conditions. The difference might come from a pattern an ML model picked up, an optimizer weighing defined trade-offs, or a rule dictating the response.
Each has distinct implications. A prediction hands the planner more information but leaves the decision with them. A recommendation narrows the decision space, though someone may still need to weigh a customer relationship or an operational wrinkle the system can’t see.
Once the technology starts acting on its own, the organization needs clear answers on what it’s authorized to do, where it stops, and when a human needs to step back in.
This is also where maintenance requirements surface. Predictions need clean data and ongoing accuracy checks. Recommendations need to keep reflecting the objectives and constraints they were built around. Anything that acts needs defined access, guardrails on how that access gets used, and oversight of what it actually does. Skip this step, and there’s no real way to know whether a feature is solving the problem you think it’s solving, or what it’ll cost you to keep it working after go-live.
The Evaluation That Actually Matters
A fair read on any AI claim in a logistics technology deal requires understanding what the feature does, what’s actually producing the result, and what it demands once it’s live — data and integration dependencies, security and access requirements, internal resourcing, and where responsibility sits between vendor and buyer. Those factors, not the marketing language, determine whether the capability you saw in the demo holds up in production.
AI can add real value to logistics technology, the same way automation and optimization have for decades. Buyers should be able to identify where value is coming from and what will be required to sustain it before allowing the promise of AI to influence a technology decision.
If you are evaluating AI as part of a logistics technology investment and need help separating the claims from the underlying capability, reach out to JBF Consulting. Their Strategy Practice can help.
Author Bio
Rachelle Butler is a Director, Strategy at JBF Consulting with more than 15 years of experience spanning logistics operations and technology. She partners with shippers to assess, design, and implement solutions that align operational needs with long-term business direction.
Rachelle’s background includes roles in product leadership, consulting, implementation, and post-deployment client success at e2open, BluJay Solutions, and LeanLogistics. She began her career in the United States Marine Corps, where she gained foundational experience in transportation coordination and logistics operations. Rachelle brings a practical, real-work approach to helping clients realize meaningful value from their operational investments.