An AI demo can make almost any workflow look simple. The harder question is what happens when that tool meets your customers, your carriers, and the way your brokerage actually operates.
In the Port TMS roundtable AI in Freight: What’s Real, What’s Next and Who Wins, Mark Groves, Chris Jolly, and Matt Harding discussed that gap between a compelling pitch and useful daily work. Their perspectives differed on where automation belongs, but a clear theme emerged: evaluate AI by the work it improves and the service it helps your team deliver.
Start with a task your team needs help completing
Chris described the opportunity to spend less time on repetitive tasks and more time on sales, customer conversations, and other activities that move the business forward. Matt pointed to the many steps and potential failures in the process from quote to cash.
That gives brokerage leaders a useful starting point. Before comparing vendors, identify a task that creates repeated work. Where is someone collecting the same information, moving it between systems, or chasing an update? What would a successful result look like?
The answer should be specific enough to evaluate. A tool that summarizes account information before a customer call serves a different purpose from one that contacts carriers or updates shipment records. Each needs its own expectations and review process.
Ask where the work goes after the AI finishes
The roundtable explored tools that work within an operating platform and tools that sit outside it. The practical issue is the handoff: how does information reach the person or system that needs it next?
For a brokerage evaluating a tool, that means asking where call outcomes, shipment updates, and exceptions appear. If staff must copy the result elsewhere or reconstruct the conversation, include that work in the evaluation.
An impressive interaction is only one part of a useful workflow. Your team still needs a clear next action.
Understand cost before deciding on value
Mark emphasized that AI has costs and that buyers need to understand what they are paying for. His examples also show why a single headline price is not enough to evaluate an entire workflow.
A practical evaluation should connect spend to a defined result. What does the tool complete? What still needs staff attention? How much time does the team spend reviewing, correcting, or following up?
For an initial pilot, choose a small set of measures such as time spent per task, completion rate, and follow-up work. These are suggested evaluation measures, rather than results reported by the panel.
Keep people involved where judgment matters
Chris was open to automation while expressing reservations about delegating some carrier conversations. Matt highlighted the difficulty of automating trust and information that has never been captured digitally.
The discussion also recognized that communication preferences vary. Some established carrier relationships work well by email; other situations benefit from a conversation.
The takeaway is to define the human handoff as carefully as the automated task. Know when an employee needs to review an exception, resolve uncertainty, or speak directly with a customer or carrier.
Use preparation as an immediate opportunity
One of the strongest areas of agreement was customer preparation. The panel discussed using AI to research prospects, organize market information, and prepare for business reviews.
That work can help a broker arrive with more relevant questions. It still needs review. Matt noted that polished AI output can contain problems, so presentation quality should never replace checking the underlying information.
Start with a real task, review what happens, and expand based on the result. The goal is a team with more time and better information to serve its customers.