How an automated parts availability lookup runs at a dealership

What the manual version actually costs

Parts counters run on interruption. A tech calls asking if a hydraulic filter is in stock. Ten minutes later a walk-in customer wants a price on a seal kit. In between, the phone rings again for a quote on a used loader. Each of these gets the same manual sequence: put the current task down, walk to a terminal, look up the part number, read the screen, then walk back or pick the phone back up. None of it takes long by itself, but a parts department fielding fifty or sixty of these calls a day loses real hours to the walking and the waiting, not the looking up.

That's the same math behind answering the phone all day: the cost isn't any single call, it's the interruption adding up across a shift, spread across parts and service scheduling, plus the sales desk, which fields trade-in questions.

How the automated version runs, step by step

The request still starts the same way, by phone or text. A voice agent or a text parser pulls out what it can: a part number if the caller has one, or a description if they don't. If the description is vague, the agent checks it against the dealer's own parts catalog before doing anything else, the same way a counter person would flip through a manual. It isn't searching the open web or a third-party parts marketplace; it's reading fields that already sit inside the DMS the dealer already pays for.

Once it has a part number, it queries the dealer's inventory record directly, inside DIS, CDK Global Heavy Equipment, e-Emphasys or Charter Software, whichever system the dealer actually runs. If the part is on the shelf, it reads back the count and the bin location, and quotes a price alongside it. If it isn't, it checks for an open purchase order or an incoming transfer and gives an estimated arrival date when the system has one.

When there's no estimate, or the part number doesn't resolve to anything in the catalog, the agent stops guessing and flags the request for a person, with the transcript attached so nobody has to ask the customer to repeat themselves. We think that handoff is the part most automation pitches skip over, and it's the part that decides whether the system holds up on a bad day.

Why this isn't the same problem an auto parts shop has

Most of what ranks for parts lookup today solves a different problem. PartsVoice describes itself as the largest open OEM parts locator in the US and Canada, built around more than 1,800 dealers that list inventory to each other. That's a network problem: whose shelf has the part, not just mine. PartsTech aggregates more than 225 suppliers into a single search for repair shops. That's a sourcing problem: which of many vendors can fill the order fastest.

A construction, ag or marine equipment dealer usually has neither problem. The inventory lives in one system, at one or a handful of branches, and it's already as accurate as the DMS is kept current. The actual gap sits between the request channel, a ringing phone, and data that already exists. Protractor Software, a shop management platform, builds its catalog lookup directly into the repair order screen rather than a separate portal, which is the more useful pattern to copy: parts availability wired into the workflow where the question comes up, not a second system someone has to remember to check.

The same interruption pattern shows up past parts. A dealer weighing this kind of automation is usually facing the same problem in equipment rental availability questions, just with rental units instead of shelf stock.

Where a person still has to step in

An agent can read a DMS record faster than a person can walk to a terminal, but it can't do everything the counter does. A part number that's been superseded needs someone who knows the cross-reference history, not just the current catalog. A vague description like 'the round thing under the loader' needs someone who has seen the machine. Negotiating price on a backordered part with a vendor still takes a person who can trade on volume and history, not just read a screen. Neither is a limitation worth apologizing for. It's just where judgment lives.

The less obvious limit is data freshness. If DIS, CDK Heavy Equipment, e-Emphasys or Charter syncs inventory overnight instead of in real time, an automated answer at four in the afternoon is only as current as last night's count. We'd audit that before automating anything else: how often the system of record actually updates, and whether the counter's answer still matches what's on the shelf. Where AI agents fail when they touch real systems is almost always right here, at the seam between the software and what's physically true.

Before letting an agent read from or write to any of these systems, it's worth working through how to evaluate an AI automation vendor, specifically what happens when the agent gets it wrong, and who sees the mistake first.

Pull last week's parts counter call log and sort it into two piles: calls that were a pure stock check, and calls that needed real judgment. That split is the actual size of the opportunity, before any vendor conversation starts.