What heavy-duty parts distributors asked about AI order automation

At one heavy-duty distributor, about fifteen feet of shelving under the parts counter is packed with catalogs, binders and loose paper. Counter staff reach for it every day to find a part or work out a cross. A council member described that shelf during Episode 5 of the CVSN Business Innovation Accelerator, a recorded series in which the Commercial Vehicle Solutions Network puts a technology company in front of distributors and suppliers from its advisory council.

Revalgo founder and CEO Ajay Kamble was the guest. He spent most of his career as a CIO and CTO in distribution, and he described Revalgo's job simply: take demand from whatever channel it arrives on, whether email, phone, fax, text or the counter, build the complete order, and work out how to fill it from the distributor's own stock or the right supplier. The council spent the hour testing how that holds up in their business.

No part number, wrong part number

Requests often arrive with a photo, a wrong number, or nothing specific at all. Ajay's answer was that the system doesn't need a part number. It's trained on the distributor's own history of what customers asked for and how each request was filled, so it works conversationally.

Crosses got the hardest questions. One member pointed out that an 80-micron filter on an engine that needs 20 microns means a failed engine. Ajay explained that the model learns, category by category, which specs can never flex and which can, like brand or color, and it learns which customers insist on the brand they named and which just want the cheapest part that works.

For something like drive axle brakes on a 2014 Freightliner, where one shoe fits four brands in four sizes, he was direct. About 80% of typical requests can be automated. The rest still need a person asking "did you mean this?"

The lost sales your ERP never sees

An ERP only records what someone entered. If a customer sends an RFQ for ten line items and orders five, the other five usually disappear. Because Revalgo ingests the original request, it keeps them, and it can show whether the same products keep slipping away across customers. That points to either a pricing problem or a stocking gap.

What comes next

The council described a younger customer base that doesn't want to call anyone. They want the price and they want it now. Ajay's view is that this makes the knowledge sitting in people's heads and in those binders urgent to capture, because online ordering only works for customers who already know the exact part number.

He also offered a prediction, which he labeled his own opinion: within a couple of years, buyers will have AI agents that want to talk to the distributor's agent in real time. His question to the council was whether they'd be ready.

The full session goes further into non-stock sourcing, how long the system takes to learn a distributor's business, and where the ROI comes from. Watch Episode 5 here.

‍