Customers may know what occasion they need an outfit for and what cut they prefer, yet be unable to name the brand or designer. If a search requires an exact product name, even customers who can describe their needs may struggle to find what they want. Business sales teams often face a similar situation: customers explain what they want to use a product for, then ask a sales representative to help choose a model. For businesses looking to use AI to support this work, a product catalog alone is not enough.
Intellicon Solutions recommends first documenting how customers typically describe their needs and the conditions in which products are suitable, then turning the way experienced sales representatives clarify requirements into a set of steps others can follow. In the first phase, an AI Agent can help internal sales teams find product information. There is no need to launch a customer-facing shopping service from the outset.
Why Can’t Customers Find Products When They Can Describe the Intended Use?
According to TechOrange, every item on the luxury resale platform The RealReal is unique and is not restocked once sold. Even when shoppers know the occasion they are dressing for, the cut they prefer, or the era of style they want, they may still struggle to find an item because they do not know the brand name. Following a pilot, the platform worked with Google Cloud to expand Ask TRR to all members, allowing them to describe their needs in a conversation and receive recommendations based on what they are looking for and their personal preferences.
The report shows businesses that product searches can start with customers describing their needs. However, it does not provide conversion rates or actual cost changes following the launch of the recommendation service, so it cannot support promises about the results a business might achieve with its own implementation. We recommend starting with this question: when a customer explains the intended use, can the sales team use the company’s existing information to identify products worth comparing and explain the conditions each is suited to?
For example, suppose a customer asks a component supplier for a product suitable for a humid environment. The sales representative cannot simply search for “humid” in product names; they also need to clarify the installation location, the equipment it will be used with, and dimensional constraints. This is a hypothetical product selection scenario, not a case from the report. When preparing information, businesses can start with a product category that frequently receives this type of inquiry and document the language customers actually use, rather than expecting them to understand the company’s model naming conventions from the start.
What Information Do Product Records Need Beyond Specifications?
A specification sheet can list dimensions, materials, and model numbers. But if a business wants an AI Agent to help find products based on intended use, it also needs to explain which usage conditions those specifications suit. We recommend having sales teams first document the terms customers use to describe their needs, then asking product specialists to explain the applicable conditions. A customer’s request for something “more durable” should not be treated as a specific technical requirement without clarification.
The information can be organized into the following categories so the people responsible know exactly what to add:
- How customers describe their needs: Keep the descriptions of intended use and informal terms customers actually use, and note the corresponding formal names. If a phrase could have more than one meaning, also list what needs to be clarified.
- What uses each product is suited to: In addition to suitable applications, specify compatible equipment, installation requirements, and situations in which the product is not suitable, so sales teams can explain product differences using the available information.
- Which specifications can be compared: Organize model numbers, dimensions, materials, and other details into tables with consistent column names and units. Identify the corresponding models in usage scenario documents so colleagues do not have to guess which product a document refers to.
For example, EgentWrX provides knowledge bases that an AI Agent can search for information organized by the business, with Wiki, vector, and table types available. Vector retrieval finds content based on similarity in meaning. Businesses can start by organizing specification sheets and usage scenario descriptions into categories for an AI Agent to search for product information. Having these materials ready, however, is not the same as having a complete shopping recommendation service.
Before organizing the information, we recommend that managers ask sales representatives and product specialists to review the same product document together: using this document alone, can they explain why a particular product suits the customer’s intended use? If the explanation is insufficient, ask the product specialists to fill in the gaps first rather than expecting an AI Agent to infer the conditions in which the product is suitable.
Turn Experienced Sales Representatives’ Questions into Steps Others Can Follow
When customers cannot name a product, presenting a list of model numbers may still leave them unsure how to choose. We therefore recommend asking experienced sales representatives to document the first question they would ask when an inquiry is unclear, then determine the next question based on the customer’s answer. Avoid starting with a questionnaire that asks customers to fill in every specification field.
In the hypothetical scenario above, the first questions could be where the product will be used and what equipment it will be used with. If the customer has already identified the equipment, move on to installation dimensions rather than asking about the intended use again. If the customer does not know the dimensions, ask them to obtain the information from on-site colleagues before continuing the product comparison. Questions about appearance or brand preferences can wait until the essential requirements are clear, so the focus on what customers like does not overshadow practical constraints.
Businesses can document standard work steps and rules in EgentWrX Skills for an AI Agent to use when applicable. Colleagues can also specify which Skill to use during a conversation. The question sequence described above, the follow-up questions to ask after each answer, and what to ask first when information is missing can all be organized into a Skill. Product specifications, meanwhile, belong in the knowledge base and should be prepared separately from the process for clarifying requirements.
In the first phase, we recommend having an AI Agent help sales teams organize requirements, identify products to compare, and explain the differences. Sales representatives should still confirm the intended use and commercial terms before formally recommending a model or providing a quote. As a next step, ask the sales representatives and product specialists responsible for the selected product category to jointly prepare a specification sheet, a usage scenario document, and a question sequence. This gives IT and AI project leads concrete material to discuss, rather than just a request to “build a product chatbot.”
FAQ
Can we start AI product search with only our existing product catalog?
Yes. You can start with your existing catalog without rewriting every document, but we recommend adding information about intended uses and usage limitations. If the catalog contains only model numbers and specifications, ask sales teams to document how customers actually phrase their questions, then have product specialists add the applicable conditions so an AI Agent can use the information to find products that meet customer needs.
What should we ask first when a customer’s needs are vague?
Start with the essential requirements that affect product selection, then ask about preferences. For example, clarify the usage setting, the equipment the product will be used with, and dimensional constraints before asking about appearance. We also recommend having experienced sales representatives document how follow-up questions should change in response to different customer answers. This progressively narrows the search without requiring every specification field to be covered at once.
Should the first phase be for sales teams or open directly to customers?
We recommend starting with internal sales teams to help them organize requirements and find product information. This allows the initial effort to focus on product information and question sequences, while sales representatives can clarify conditions customers have not fully explained. Once the business decides which customer-facing services it wants to offer, it can separately plan customer interactions and the subsequent shopping process.
References
- How Do You Sell Faster When Every Item Is One of a Kind? Luxury Resale Platform The RealReal Uses Two AI Systems to Transform Warehousing and Search — TechOrange, November 20, 2026, by Ariel; describes approaches to item processing and conversational search but does not provide separate performance data for the recommendation service following its launch.



