💬 The Portal Is Becoming a Conversation

Why AI could replace traditional customer portals—and make structured product data more valuable than ever.

A customer conversation this week changed the way I think about the future of product information.

We were discussing a fairly conventional project: a new WordPress website, a WooCommerce integration and a connection to their PIM. Nothing unusual.

Then the conversation turned to AI.

The customer asked whether their own customers—the people specifying and buying their products—could speak to an AI assistant that understood the entire catalogue.

Not a basic chatbot offering scripted answers.

An intelligent product assistant with access to every specification, technical detail, image, document and relationship between products.

The answer was yes.

A customer could say:

“I need profiles suitable for an outdoor canopy project.”

The AI could ask where the project was located, what environmental conditions applied and whether there were particular structural or aesthetic requirements.

It could then identify suitable products, explain the reasoning behind its recommendations and generate a professionally formatted datasheet containing the selected products, specifications, dimensions and images.

No menus.
No filters.
No export configuration.
No twelve-step download process.

Just a conversation that ends with the customer receiving exactly what they need.

There was a brief pause.

Then the customer asked:

“So we do not need to build a customer portal?”

That is the moment worth examining.

🖱️ Thirty years of putting buttons around databases

For most of the past three decades, business software has been built around user interfaces.

Portals, dashboards, search screens, configuration pages, reporting tools and export wizards all exist for broadly the same reason: people need structured pathways through large amounts of information.

A customer portal exists because no person can hold an entire product catalogue in their head.

Search filters exist because a customer cannot examine ten thousand products simultaneously and identify the five that meet a particular requirement.

Export tools exist because users need a way to specify which products, attributes, images and documents they want to receive.

These interfaces are useful, but they are also compensations for human limitations.

Every screen has to be designed. Every option has to be programmed. Every workflow has to be tested. Every change has to be maintained and supported.

Now put an AI assistant in front of the same product data.

The AI does not need to click through a portal. It can examine the entire catalogue.

It does not need an export wizard. The customer can simply describe the required output.

It does not need a conventional search interface. It can interpret requirements, identify relevant products, ask clarifying questions and explain its conclusions.

The interface is no longer a collection of screens.

It is a conversation.

👋 What begins to disappear

This transition will not happen overnight. Traditional interfaces will remain useful for many tasks, particularly where users require precise control or need to review large volumes of information visually.

But their role is likely to shrink.

💬 Customer portals become product conversations

Instead of logging in, navigating categories and applying filters, customers speak to an assistant that understands the catalogue as well as an experienced salesperson.

In some respects, it may understand it better. It has perfect recall of every specification, compatibility rule, product relationship and supporting document.

📄 Export configurators become simple requests

A customer can ask:

“Create a PDF containing these five products, including full specifications, images and dimensions.”

The AI can assemble the information, apply the correct template and produce the document.

The customer does not need to understand the data structure behind it.

🔎 Search becomes consultation

A conventional search might begin with the phrase “outdoor profile” and return hundreds of results.

A conversation is different:

“I need a product for outdoor use.”

“What environmental conditions will it face?”

“It will be installed near the coast.”

“In that case, these three products are suitable for marine environments.”

That is not simply search. It is guided product selection.

📊 Report builders become questions

Instead of creating filters and selecting columns, users can ask:

“Which products have not been updated in the past six months?”

“Show me every product added since January that is missing an image.”

“Which products have incomplete technical data?”

The AI does not necessarily require a dedicated reporting interface. It requires reliable access to the data and a clear understanding of the question.

🧠 The technology underneath becomes more important

It would be easy to conclude that removing the portal makes the underlying system simpler.

The opposite is true.

When people navigate a portal, they regularly compensate for poor data.

They recognise that “aluminium” and “aluminum” refer to the same material. They notice that a product image is incorrect. They understand that two slightly different attribute names mean the same thing. They work around missing descriptions and inconsistent classifications.

People are remarkably good at quietly correcting bad systems in their heads.

AI is less forgiving.

For an AI assistant to reason reliably about a product catalogue, the information must be clean, structured, complete and properly contextualised.

It needs:

• consistent attribute names and values;

• accurate technical specifications;

• clear product classifications;

• meaningful relationships between products;

• current images and documents;

• comprehensive product descriptions; and

• rules governing compatibility, suitability and compliance.

This is why PIM does not become less important in an AI-driven world.

It becomes the foundation.

The single source of truth is no longer merely an architectural preference. It becomes the knowledge base behind every AI response, recommendation, report, export and customer conversation.

Companies that have invested in organising, enriching and maintaining their product data will be in a strong position to benefit from AI.

Companies whose data remains scattered across spreadsheets, duplicated across websites and trapped inside an ERP will encounter a more difficult reality.

AI does not repair poor information simply because it can express an answer confidently.

Garbage in still produces garbage out—only now the garbage may be unusually articulate.

🏗️ From a data layer to a knowledge layer

The next stage goes beyond storing product attributes.

What is emerging is a product knowledge layer: the information and context that allow an AI to interpret the catalogue intelligently.

That includes questions such as:

• Which industry does the customer operate in?

• Which regulations and standards apply?

• Which terminology do customers use?

• Which terminology do engineers use?

• Which products are compatible?

• Which products are commonly specified together?

• What problems is each product designed to solve?

• What are the common reasons a product may be unsuitable?

• Which questions should be asked before making a recommendation?

This knowledge is often present within a business, but not within its systems.

It sits inside the heads of experienced salespeople, engineers, product managers and customer-service teams.

Capturing that knowledge is what turns a general-purpose AI model into a useful business assistant.

It is also why we encourage each OneTimePIM customer to give their AI assistant its own identity.

The assistant is not simply a generic chatbot attached to a database. It is shaped around the customer’s products, market, terminology, workflows and accumulated expertise.

Prompt templates, agent tasks, business rules and customer-specific context may sound like software configuration.

In reality, they are institutional knowledge being made available to AI.

Every corrected translation, refined prompt, product-selection rule and unusual customer query makes that knowledge layer more valuable.

And unlike a conventional portal, that value can compound over time.

🧭 How businesses should prepare

For businesses that depend on product information, there are four practical implications.

1. 🧱 Treat your PIM as strategic infrastructure

Your ERP remains essential for managing transactions. Your CRM remains essential for managing customer relationships. Your website remains essential for presenting your business.

But your PIM increasingly holds the product knowledge that connects them all.

It will power websites, marketplaces, catalogues, sales tools, technical documents and AI assistants.

It should be funded and managed accordingly.

2. ✅ Treat data quality as AI readiness

Cleaning product data is no longer merely an administrative exercise.

Every inconsistent value, missing attribute and poorly defined relationship creates another opportunity for an AI system to provide an incomplete or incorrect answer.

The businesses that improve their data now will gain a meaningful advantage later.

3. 🚪 Question the next portal project

Before committing six months and a substantial budget to another customer-facing portal, ask whether the same objective could be achieved more effectively through a conversational interface connected to clean PIM data.

A portal may still be necessary.

But it should no longer be the automatic answer.

4. 🧩 Capture the knowledge your systems do not contain

Identify what your most experienced employees know that has never been formally recorded.

Which questions does your best salesperson always ask?

How does your technical team identify an unsuitable product?

Which product combinations are commonly recommended?

Which specifications appear acceptable on paper but create problems in practice?

That is the knowledge that will distinguish a genuinely useful AI assistant from one that merely repeats product descriptions.

🔮 The end of the portal era

Traditional portals are not going to disappear tomorrow.

But the direction of travel is becoming clear.

Customers will increasingly expect to ask for what they need rather than learn how a supplier’s interface works.

They will expect systems to understand intent, ask intelligent questions, recommend appropriate products and generate useful outputs immediately.

The technology visible to the customer may become simpler.

The product knowledge underneath it will need to become much better.

That is the opportunity.

The future of product experience may not be another portal with more menus, filters and configuration screens.

It may simply be a knowledgeable conversation.

And the last customer portal you build may genuinely be the last one you need.

---

Dr Pat Violaris is Managing Director of OneTimePIM, a Product Information Management platform developed by a team that has been working with complex product data since 1992. He holds a PhD in Expert Systems—which is what artificial intelligence was commonly called before it became fashionable.