Imagine a customer visiting your website with a specific problem to solve. They understand the application, the operating conditions, and what they need the product to achieve. What they don't know is which of your hundreds or thousands of products will do the job.
Today, they might search through product categories, apply filters, compare technical specifications, and download several datasheets. If they still can't identify a suitable solution, they'll probably need to contact your sales or technical team.
But what if they could simply describe their requirements to an AI product adviser built into your website?
Rather than expecting customers to understand your catalogue, the AI product adviser could interpret their requirements, ask relevant questions, and guide them towards suitable product categories, technical information, and potential solutions.
For manufacturers selling complex products or with hundreds of products, this could represent a significant change in how your website supports the buying process. It could also create opportunities to improve enquiry quality, reduce repetitive technical enquiries, and help customers identify products they might otherwise overlook.
At DotCentric, we've been exploring how this could work in practice. Our work has raised some important questions about what's possible, what the technology needs to deliver, and where manufacturers should concentrate their investment...
Why conventional product search has its limits
Most manufacturing websites are designed around a fairly straightforward assumption – customers know enough about the product they're looking for to find it.
That works well when someone has a part number, knows the product family, or understands the terminology used in your catalogue. But it becomes harder when customers know what they need to achieve but aren't sure which product will meet their requirements.
Think about an engineer looking for a fastening solution for equipment exposed to vibration, frequent maintenance, and demanding operating conditions. They understand what they need in terms of a part or product to fix things, but they perhaps have limited knowledge of the manufacturer's available product range and how to look for it.
A conventional website search engine can help them find information containing particular words or phrases. Filters can narrow the results using known technical attributes. But neither necessarily helps the customer establish which product or product category they should investigate in the first place.
This is particularly pertinent to manufacturers with large product portfolios, where different product families may serve similar applications, technical requirements vary, and some solutions involve multiple compatible components.
What could an AI product adviser actually do?
Well, here’s how it might work…
An AI product adviser would sit within your website, drawing on your product catalogue, technical documentation, and supporting content.
A customer could visit your website and describe what they’re looking for in everyday language, without having to know your own specific terminology or product classifications.
The AI adviser would interpret the customer’s request and identify relevant areas of your product catalogue. The adviser could then ask follow-up questions about what the customer wants, their operating environment, required performance criteria, or other selection measures.
As the customer’s requirements became clearer, the AI adviser could narrow the options and direct them towards relevant product families, individual products, specifications, and technical resources.
For example, a customer might begin by asking for a suitable access solution for a piece of industrial equipment. The adviser could establish whether the application involves frequent opening, security requirements, environmental exposure, or particular space constraints, before suggesting which product categories warrant further investigation.
This is a vastly different, and much better, experience from entering a search term and receiving a list of results that you need to wade through.
Bringing an AI product adviser to life
We’ve recently been talking to a manufacturer with a broad portfolio of engineered products, where we identified an opportunity to introduce AI-assisted product discovery as part of a wider digital transformation.
The manufacturer's catalogue included products with very different levels of complexity, from relatively straightforward components to specialist systems requiring detailed technical consideration, so it was perfect for an AI product adviser approach.
One specific challenge the manufacturer had was helping customers identify which product categories or classes might be appropriate for their application before asking them to evaluate individual products.
To address this, we explored an approach that would combine natural-language interaction with semantic search, structured product information, and the manufacturer's existing technical content. Our proposed solution was to use retrieve relevant product and content information, alongside a large language model to interpret customer questions and provide guidance.
Rather than allowing the AI to generate answers solely from its general training, the adviser would use retrieval-augmented generation (RAG), supplying the model with relevant information retrieved from the manufacturer's own sources, including product data, supporting content, and technical documents. Not only this, but it would also combine semantic search with conventional keyword search, recognising that different types of queries require different approaches - a customer searching for a precise part number has very different needs from someone describing an unfamiliar application, for example. The concept also included follow-up questions, safeguards against unsupported product claims, and clear routes to sales or technical specialists when the customer's requirements couldn't be resolved confidently.
An AI product adviser like this could revolutionise your website and digital customer experience.
But there are several considerations manufacturers should address before investing in this type of capability…
The biggest challenge is probably your product information
An AI product adviser can only provide dependable guidance if the information available to it is sufficiently accurate, complete, and relevant.
Manufacturers often hold extensive technical information across product information management systems, website content, PDFs, engineering documentation, and other internal sources. The difficulty is that information created for different purposes isn't always structured in a way that supports application-led product selection.
Fo example, a product record might contain dimensions, materials, and performance specifications without clearly describing the applications for which the product is suitable. Technical documents, on the other hand, may contain valuable guidance that isn't readily accessible through your existing website search. Relationships between products, components, and compatible systems may also be difficult to establish.
An AI product adviser needs access to more than just individual product descriptions. It needs enough information to understand the relationship between a customer's requirements and the solutions you, the manufacturer, can offer.
Improving your underlying product and content information is therefore a prerequisite for developing an AI product adviser experience.
That means identifying gaps in product attributes, improving supporting content, and ensuring your system can retrieve reliable information about applications, specifications, and product relationships. It also means recognising that not every product category will be equally suitable for AI-assisted selection from the outset.
How do you stop AI recommending the wrong product?
This is one of the most important questions for any manufacturer considering an AI product adviser.
An incorrect recommendation for a product might lead to a return or an unhappy customer. For some industrial applications, recommending an unsuitable component could have much more serious consequences.
Using your own product information through RAG can help reduce unsupported answers, but it doesn't eliminate the risk of incorrect interpretation or recommendations. The experience needs safeguards appropriate to the products and applications involved.
These safeguards might include restricting recommendations to products supported by retrieved technical information, checking critical requirements against structured specifications, and preventing the system from making claims that the source material doesn't substantiate. The adviser also needs to recognise when it lacks enough information to make a useful recommendation.
In some situations, asking another question will help. In others, the appropriate response may be to direct the customer to a technical specialist, ideally passing on the requirements already collected so the customer doesn't have to start again. For products with strict compatibility, safety, or engineering constraints, deterministic selection rules and validated configuration tools may be necessary alongside the AI advisor experience.
An AI product adviser can help customers understand their options and navigate complex information, but it shouldn't be expected to replace engineering validation where that is required.
Could an AI product adviser make your sales team more effective?
Manufacturers with complex portfolios often rely on sales engineers, technical advisers, and customer service teams to help customers identify appropriate solutions.
Some of those conversations require specialist expertise and should remain with experienced people. Others involve answering familiar questions, explaining product differences, or helping customers find information already available within the business.
An effective AI adviser could help customers resolve more of those initial questions independently.
It could also improve the quality of conversations that do reach the sales team. Rather than receiving a general enquiry asking which products might be suitable, a technical specialist could receive a more developed description of the application, requirements, and options already considered. There might also be an opportunity to introduce customers to relevant product families that they wouldn't otherwise have discovered through conventional navigation.
For manufacturers working through distributors or other indirect sales channels, the experience could support product research and specification without changing the established route to purchase.
These are potential benefits rather than guaranteed outcomes. Their value will depend on the products, the customer journey, and how successfully the adviser performs.
For Heads of Marketing and Digital, that makes defining the commercial objective particularly important. Reducing unnecessary technical enquiries, improving qualified leads, increasing successful product selection, and supporting product specification are different objectives requiring different measures of success.
Intrigued? Where should manufacturers start?
The temptation with any new AI capability is to begin by evaluating technology platforms. But manufacturers will get more value from first understanding where customers struggle with product selection and whether an AI-assisted experience would genuinely improve things.
Here are four areas worth examining to see if an AI product adviser could work for you.
1. Identify the product-selection problems worth solving
Look at the enquiries your technical teams receive, the questions customers repeatedly ask, and the stages of product selection where people need additional support. These can help identify where an adviser might provide value.
2. Assess whether your product information is ready
Review the completeness of technical attributes, application guidance, compatibility information, and supporting documentation. Establish what would need to improve before the system could provide reliable guidance.
3. Choose a manageable starting point
A defined product category or application area provides an opportunity to test the experience using real customer requirements, without attempting to cover your entire catalogue from the outset.
4. Establish how you'll test accuracy and commercial value
Testing should examine whether the adviser identifies relevant solutions, asks appropriate follow-up questions, handles uncertainty correctly, and knows when to involve a specialist. Customer testing and measures such as successful product selection and enquiry quality can help establish whether the experience is delivering value.
The AI product adviser approach we've proposed at DotCentric follows these principles, using a proof of concept to validate the experience, technical architecture, and product information requirements before wider deployment.
That allows manufacturers to learn where the technology performs well, where additional work is required, and whether the commercial case supports further investment.
Ready to take on an AI product adviser?
For manufacturers with complex products or a huge product portfolio, an AI product adviser could help turn your website into a more useful source of guidance during the buying process, which in turn will help to increase overall sales, even if they happen elsewhere.
The opportunity is particularly interesting where customers begin with an application or engineering problem in mind, rather than a product name, and where finding a suitable product solution currently requires considerable search effort or specialist assistance.
At DotCentric, our approach is all about bringing together the customer experience and the technical foundations needed to make more advanced product discovery possible for manufacturers.
If you're considering how AI could improve your product discovery, technical enquiries, or customer self-service across a complex manufacturing catalogue, just get in touch - we'd love to talk.
