Who answers the fitment call? Voice AI for Shopify auto parts stores

"Will this part fit my car?" is the question an auto parts store hears most. On the website a fitment app handles it: the buyer picks year, make and model and the catalog narrows down. On the phone, the answer depends on who is free to pick up.

When nobody is free, the buyer may not wait. In a CallRail survey of 1,000 U.S. consumers published in September 2025, 78% said they had taken their business elsewhere after failing to reach a company by phone.¹

This is the long version of our launch post on LinkedIn. It is written for two readers: Shopify auto parts merchants who miss fitment calls, and the agencies that build and run those stores and keep hearing "can AI answer our phone?"

Why a fitment call is not a normal support call

Most AI phone agents are built for questions that have one answer per store: opening hours, the shipping policy, where an order is. A fitment question has a different answer for every vehicle. "Do you have front brake pads for a 2014 F-150?" can depend on the trim, the drivetrain, sometimes the cab style.

That has two consequences.

• The answer has to come from the store's own fitment data, not from what a language model "knows" about cars. A model will produce a confident answer either way, and a confident wrong answer comes back as a return.
• "I can't confirm that" is a valid answer. A good agent says it cannot confirm the fit and passes the question to a person instead of guessing.

Where the answer lives: vehicle tags

Most Shopify parts stores already have fitment data. It is what powers their year / make / model search. Apps like EasySearch keep it as vehicle tags on each product, and the same data can drive a phone conversation.

What an agent needs from those tags:

• Year, make and model are the minimum to match a product.
• Trim and drivetrain are where the fit usually splits.
• One format across the catalog. Tags in mixed formats, or tags that disagree with the product title, are where any automated answer goes wrong.
• Untagged products treated as universal. Fluids, tools and accessories should not trigger a "what car do you drive?" interrogation.

A practical step for agencies: before any voice or chat agent touches a client's catalog, run a fitment audit. Which products have no tags? Which tags use a different format? Where does the title name one vehicle and the tags another? The audit pays off on its own, because it is the same data that makes the storefront search work.

How the call should go

This is the flow we built Inclamo around. It also works as a yardstick for any voice agent you evaluate for a parts store.

1. The caller names a part and a car. Until the agent has the year, make and model, it asks for what is missing, one question at a time.
2. The search runs in the store's own catalog, so everyday wording works: "shocks", "ball joint", "roof basket". If several products match, the agent reads back a short list and asks which one the caller means.
3. The verdict comes from the tags: fits, doesn't fit, unknown, or fits on a condition. If the fit depends on four-wheel drive, cab style or trim, the agent asks the caller instead of deciding for them.
4. Price and availability come from the store, and only after the caller has confirmed which product they mean.
5. Anything it can't confirm goes to the team. That means a transfer to the store's number within the hours the merchant sets. If nobody picks up, the agent takes a callback request and reads the number back digit by digit.

The other calls: orders and returns

Fitment is not the only reason people call a parts store.

Order status. If the caller's phone number matches a customer in the store, the agent can greet them by name and see their order history. Otherwise it asks for the order number and ZIP code. It tells the status, the carrier and the tracking number, and never reads out an order to an unverified caller.

Returns. The agent records a return request for the team to process. It never tells the caller a return is done when only the request was recorded.

What it should not do on a first rollout: edit or cancel orders, issue refunds, change inventory, or quote shipping and return policies. Those terms differ in every store, so they go to a person.

A checklist for evaluating any voice AI vendor

If a client asks you to pick a phone agent for their parts store, these questions separate a demo from something you can put in front of their customers:

• Does it read the store's fitment data directly, or do we have to maintain a separate knowledge base?
• What does it say when the tags don't confirm a fit?
• How does it handle conditional fitment: drivetrain, trim, cab style?
• When does it quote a price?
• How does it verify a caller before reading out order details?
• Can it change orders, issue refunds or edit inventory? For a first rollout, the safe answer is no.
• What happens when it can't answer: a transfer, a callback, voicemail?
• Is recording optional, is the caller told, and how long are recordings and transcripts kept?
• What does the merchant see after each call?
• Can we test it on the store's own catalog before any customer hears it?

How the Inclamo private beta works

1. Install the app from the Shopify admin. Read-only permissions: products, orders, checkouts.
2. We scan the catalog. You see which products the agent can match and which it can't, before any call.
3. Pick a voice and make a test call. The store gets a phone number. Dial it yourself and test on your own catalog; we adjust the scenario together.
4. Go live. Put the number on the site or forward the store's existing number to it.

After every call the merchant sees the transcript, a summary and the recording if recording is on. Recordings and transcripts are kept for 30 days and deleted when the app is uninstalled.

Today the beta covers Shopify stores in the US, with calls in English. EasySearch tags are supported now. If a store uses its own tag format, we adapt the agent to it.

The offer: complete the beta application and get the first month free. It starts when the agent goes live, not during setup. No credit card required.

For agencies

If you build or run Shopify stores for auto parts clients, the beta is a low-risk way to find out whether a voice agent suits a client's store. The catalog scan shows what the agent can and can't match before a single customer hears it.

Apply with your client's store URL, or write to [email protected] and we will start with the scan.

Footnotes

1. CallRail consumer survey, published September 2025; 1,000 U.S. consumers. Cross-industry and self-reported, not an ecommerce or auto-parts lost-sales rate. ↩