An AI agent built to answer customer questions turned out to know something the sales team needed. Here’s how a freight forwarder turned carrier data into earlier warnings, saved revenue, and what that means for your own agentic AI deployment.
Earlier this year, I worked with a freight-forwarding company on the personal data its customer-service AI agent collected through calls, emails and texts. It was a familiar starting point focused on what information the company needed from someone asking about a shipment, and what would happen to that information afterward.
The agent would send delay notices as they arose, but customers still contacted the company about shipments they thought were missing. As we worked through those inquiries, most of which were truly just delayed, I started looking at what the company could learn before a customer needed to ask.
What the carrier could tell us
The freight-forwarding company partnered with delivery carriers, and those carrier companies had vehicle schedules and maintenance plans that could give the freight forwarder earlier notice of a capacity problem. A carrier might be down trucks in an area for forty-eight hours and knowing that would help the customer-service agent explain a delay sooner. The carriers were already sending a feed, but the majority only ran twenty-four hours ahead, even though the data was available a full week in advance.
That capacity information would also be useful to the sales team, which was pushing business in areas where delivery capacity was about to drop. More orders meant a longer backlog, more frustrated customers, and penalties under contractual discounts and volume pricing. Salespeople were also trying to maintain client relationships while pushing orders the company would struggle to deliver on time. It’s a bit like a dentist’s front desk booking Friday cleanings after the hygienist called in sick for Friday. Someone knows there’s a problem, but the person making the next commitment hasn’t heard about it yet.
We had an opportunity to get that information to sales while they could still do something with it.
What we had to agree
I helped establish the scheduling integration and notifications from the customer-service agent to sales. My role involved bringing together customer service, sales, the Office of General Counsel and the CTO’s people across the US, Canada, Germany, France and Spain.
We had to agree how carrier information could be shared and used by the agent, including the automated notifications. The personal-data questions that brought me into the project were still there. We also worked through privacy requirements across the markets, data localization, anonymization and accessibility. Germany’s data-protection requirements made it the hardest of the five. The US was the simplest.
Those decisions had to fit both legacy workflows and the new agent processes. When a carrier’s schedule changed, the information needed to reach sales in a form they could use. So off we went, defining requirements with the people responsible for the information, the systems and the resulting business activity.
Once the integration and notifications were running, sales could redirect its efforts around delivery constraints. Great news all round: the company preserved revenue that would otherwise have been lost through contractual discounts, and salespeople could avoid pushing more business where they already knew delivery would be delayed. Sales was shocked and grateful. One salesperson said he could now spend his time chatting up clients or taking them to lunch instead of apologizing and getting discounts signed off by his boss. A win in my book!
Where I would look in your organization
As you prepare for your agentic AI deployment (or at your next deployment meeting, if you’re already underway), ask your business counterparts what information the agent already knows that their teams could use. In this case, carrier capacity was relevant to a sales conversation. In yours, the connection may be less obvious. The litmus test for me is to ask, “If we’d known that yesterday, could we have done something differently?”
Take one example and work through it together:
- What would they have done differently?
- Which information would have helped?
- What needs to be agreed before the agent can share it?
- How will it reach the person who can act?
That gives you a specific opportunity to assess, with the affected teams involved from the beginning. It also gives your policy work a business outcome to aim for.
P.S. I’ve anonymized this use case to protect my client confidentiality.
Image: AI-generated.
.jpg)
