AI Consultant

The client distributes lubricants, mainly Valvoline® and Revline®. Its business is B2B sales; it doesn’t sell retail, and car owners buy oil from its partner stores. Once the distributor started advertising on Instagram, DMs and comments filled with questions: which oil for this particular car, what goes into the transmission, which coolant, what it costs, and where to buy it. Among them were wholesale and partnership requests, as well as reposts, stickers, and spam.

An accurate answer takes several steps: identify the exact engine or transmission variant, not just the model; know the OEM approval; find a product in the lineup with that approval and viscosity; name the recommended retail price and point to a partner store in the customer’s city where availability can be checked. The sales team works mainly with wholesale buyers, yet an accurate answer to a car owner builds the brands’ reputation and brings the buyer to the distributor’s partners.

YYTECH’s team built the AI consultant and its admin panel:

• an agent connected to the CRM’s chat channels, which bring in the distributor’s Instagram DMs and comments;
• a fluids knowledge base built from the Valvoline® product portal, with an online check of current approvals;
• integration with the distributor’s catalog: products, recommended prices, stock, accounts, and partner stores;
• a vehicle lookup that lists recommended fluids for each configuration, and an approval-based selection calculator;
• an admin panel: inbox, client type, CRM record, handoffs to sales reps, catalog, and analytics.

The consultant has been in production since July 2026, and development continues.

Selection takes five steps: car → engine variant → approval → product → recommended price and where to buy. A customer may type with mistakes (“chevy malibo” is a Chevrolet Malibu) or give a VIN or mileage. If the model had several engines in that model year (unit injectors or common rail, before or after a facelift), the consultant asks which one the customer’s car has, because the approval depends on it. It then picks engine oils, gear oils, and coolants that meet the approvals and specifications.

Messages a customer sends within a few minutes are merged into one request: three short replies get one answer.

In a conversation about a 2023 Peugeot 508, the consultant picked Valvoline® SynPower ENV C2 0W-30, which meets the factory approval PSA B71 2312, and gave the recommended retail price per 1 L bottle and per 5 L jug. Once the customer named their city, it listed partner stores with addresses, phone numbers, and opening hours, plus online retailers. When the customer quoted the approval from the service book, the consultant switched to a product that meets it.

B2B requests matter most to the distributor. The consultant recognizes supply and distribution requests, wholesale inquiries, partnership offers, and web form inquiries asking for a dedicated account manager. It asks for contact details and passes the request to a sales rep with a short summary.

Out-of-scope questions are handled separately: questions the consultant can’t answer yet, such as a product the distributor doesn’t carry, an approval dispute, or a topic not yet in the knowledge base. These also go to a sales rep, and the knowledge base and scenarios are extended based on them. The sales rep sees the handoff reason in the CRM and in the admin panel and can return the conversation to the consultant. If the model fails to respond, the consultant doesn’t guess: it hands the conversation to a sales rep.

The vehicle lookup shows a model’s configurations and the recommended fluids for each vehicle system. For a 2011 VW Passat CC 2.0 TDI it found six configurations and eight engine oils, all in the distributor’s lineup; the 5W-40 oils fit only some of the engines. Had the consultant not confirmed the engine variant, it could have recommended the wrong oil.

The catalog holds 347 products from 10 brands, including 223 Valvoline® and 39 Revline®. Each product has its viscosity, approvals and specifications, recommended retail price, and stock.

The admin panel shows every conversation: messages with processing tags, client type (retail, B2B, partner, influencer, spam), the CRM record with its owner and status, and handoffs to sales reps. Analytics show how many conversations the consultant resolved on its own, how many it passed to a sales rep, and how often customers keep the conversation going.

Over two weeks, from September 28 to October 11, 2026, the consultant handled 99% of requests on its own, 572 of 579:

• 87% answered by the consultant: an oil that meets the OEM approval, the recommended price, and partner stores, with no sales rep involved;
• 10% spam, thank-yous, stickers, and the like;
• 3% passed to a sales rep: 7 B2B leads and 8 out-of-scope questions.

8 B2B leads, all identified: wholesale, supply, partnership, and web form requests. The consultant asks for contact details and passes the lead to a sales rep; for the eighth lead, the customer left no contact, so that request counts among the consultant’s answers.

In the same two weeks, sales reps sent 6 messages on Instagram and the consultant sent 559.

Measured about two months after ad traffic ramped up (active advertising since August 2026). The consultant is still being refined.

A request is the set of messages the consultant processes in one go. There is no comparison with the period before launch: requests were not counted then.

Car owners who message the distributor on Instagram get an oil that meets their car’s approval, the recommended price, and the addresses of partner stores, all in the distributor’s name. The sales team focuses on wholesale and out-of-scope questions. Work on the consultant continues; the next plans are to use the same selection logic in an assistant for sales reps and in an agent for marketplaces.