Claude for Dealers / Learn / AI for Used Car Managers

Variable Ops · Used Car Desk

AI for Used Car Managers: Descriptions, Aging Reports, and Appraisal Prep with Claude

Your inventory export already holds the answers. The problem is that nobody has two hours on Monday to read 180 rows and decide what to do about each one.

By Gray Scott, co-founder of Claude for Dealers · Updated September 2026 · 8 min read

Where AI fits on the used car desk

Buying and pricing cars is judgment. Claude does not replace it. What Claude removes is the reading, sorting, and writing that sits around that judgment: the descriptions nobody has time to write well, the aging report nobody has time to study, and the recon spreadsheet nobody updates.

Job 1: Vehicle descriptions that are accurate

Generic AI descriptions are easy to spot. "This stunning SUV is ready for your next adventure." They also create risk when the tool invents a feature the car does not have.

The fix is a description Skill with two rules. First, features come only from the build data or window sticker you paste in. If it is not in the data, it does not get written. Second, your store voice comes from a knowledge file, so 184 descriptions sound like one store.

Write the vehicle description for the unit below.
Use ONLY the features listed in the build data. If a feature is not listed, do not mention it.
Lead with the three things a shopper for this model cares about most.
No cliches. No "stunning," "loaded," or "won't last."
120 words. End with one line on our reconditioning standard from the knowledge file.

BUILD DATA:
[paste]

Job 2: The Monday aging action list

Export inventory. Give it to Claude with your rules. Get back a short list of the units that need a decision, sorted by urgency, with the reason next to each one.

The rules are the whole game. Write them down once:

• Photo minimum. For example, flag any unit with fewer than 30 photos after day 3.
• Reprice cadence. Flag any unit with no price change in the window your policy sets.
• Recon target. Flag any unit in recon past your target days.
• Exit policy. At your day limit, the list asks for a retail or wholesale decision.
• Missing data. No price, no comments, no VIN decode, stock photos.

Job 3: Appraisal prep

Before the appraiser walks the car, Claude can assemble what you already have into one summary: the history report you paste in, service history from your own RO export, your current stock of the same model with days on lot, and your recent sales of that model. It does not produce a number. It produces the page the appraiser reads so the number is better informed.

Job 4: Recon chase

Paste the recon tracker. Ask for a list by vendor and stage with days in stage. Claude drafts the chase message to the body shop or the detail vendor. One minute instead of fifteen, every morning.

Job 5: Merchandising QA

Give Claude the listing export and ask it to find problems: descriptions that mention a feature not in the build data, prices that differ between feeds, duplicate stock numbers, units with a sold status still listed. It reads every row. People skim.

What Claude cannot do here

If you ask Claude what a 2021 Tahoe is worth, it will give you an answer that sounds confident and is not grounded in today's market. Bring the market data to it. Do not ask it to be the market data.

Common questions