From self-service portals to AI supported service in your portal

Self-service means: figure it out yourself

Most customer portals have become a digital warehouse. Thousands of products. Hundreds of manuals. Spare parts you can only find if you know the exact article number. Invoices somewhere behind three menus. And a search function that returns five pages of results, none of them relevant. That is not self-service. That is sending your customer into your 'digital' warehouse with the message: go and find it yourself. Or it is like letting a customer leaf through your filing drawer until they have found their own invoice. Because you are not there to sort it out for them.

The point is not that self-service was a bad idea. It was in fact a good first step. The point is that companies are complex, and that complexity grows as you digitalise more parts of your organisation. And that complexity has seeped one-to-one into the portal. That very complexity is why these companies exist and deliver value to their customers. But if you unleash that complexity on your customer uncontrolled, you are not delivering service. You are creating more work. And that is exactly the opposite of what a customer portal should do.

Your customer now lives in a different world

While your customer portal still looks exactly as it did five years ago, the world around your customer has changed fundamentally. Your customer uses AI tools such as ChatGPT or Claude for their own work every day. They ask a question and get a clear answer within seconds. No menu structures. No search results. Just the answer. Only those answers are not yet context-aware.

And then your customer logs in to your portal. They navigate a menu structure of countless pages. They look for a manual for a product your customer bought three years ago and get a catalogue of 200 documents or manuals. Your customer wants to know when their next delivery is scheduled and has to go through three screens. That contrast gets sharper every day. Your customer expects a frictionless, intelligent experience, comparable to the experience they increasingly have when interacting with AI tools. And you are still offering them a complex digital filing cabinet.

Your customers do want to arrange things independently. That is not the problem. The problem is that they no longer want the self-service version we chased for years. They want service that is smart enough to understand their question without them having to learn how your portal works first.

From self-service to personal assistant

Imagine the following. Your customer logs in and simply asks a question. "What is the status of my last order?" "I need the installation manual for the unit we had fitted in January." "Can I repeat the same order as last month?" And the system answers. Immediately. Correctly. Without searching, without navigating, without menus, with the right context. That is where we are heading. An AI assistant in your customer portal that functions like a personal account manager. One that is always available, knows everything about the customer and everything about your business.

Not a generic chatbot giving standard answers. An assistant that understands the context of this particular customer. That knows which products are installed, which contracts are running, which price agreements apply. And that on the basis of that context gives exactly the right information at exactly the right moment. We do not show what the customer may not or need not see. We do not display a catalogue of 3,000 products when only 12 are relevant. We offer manuals at the moment there is a need for them. Not earlier. Not all at once.

Ready for an AI-ready customer portal?

Customer context changes everything

The key lies in customer context. Because we know who logs in, we know who the customer is. And because we know that, we can personalise everything. The AI assistant retrieves the order history in real time. It knows the contract terms. It knows which products are installed at this customer and can immediately find the right manuals, spare parts and maintenance schedules for them. The customer does not have to search. The assistant searches for them.

Where a traditional portal shows a list of 200 manuals, the AI assistant shows exactly one: the one for the product the customer has a question about at that moment. Where a traditional portal offers a generic product catalogue, the assistant shows only what is relevant for this customer. Including specific prices and delivery times. The difference is fundamental. This is not a slightly better search function. It is a completely different approach to customer interaction. The difference between a library without a librarian and an expert who takes exactly the right book off the shelf.

The technical backbone

This only works if the AI assistant has access to the right sources. That is where the technical foundation sits. We connect the assistant to the company's ERP system. Whether that is Exact Online, Microsoft Dynamics 365 or another system: that is where the truth about customers, orders, invoicing and stock lives. Through those connections, and by properly rationalising the data that comes out of them, the assistant knows the customer at the same level as the company's best account manager.

We also open up the knowledge base. Often that is a SharePoint environment or another document database such as a PIM system with product manuals, technical specifications and installation instructions. Through a RAG architecture (Retrieval-Augmented Generation) the assistant searches those hundreds or thousands of documents and finds exactly the relevant fragment. Not the whole document. Precisely the answer to the question.

The third layer is identity-based context filtering. On the basis of the customer profile, the system determines which information, products and prices are visible. You do not want customer A to see customer B's price agreements. And you do not want an assistant sharing information the customer is not entitled to.

This is the kind of bespoke software that makes the difference between a demo and a production system. It calls for integration, governance and an architecture built for your specific situation. As we wrote earlier about agentic AI strategy: the real value arises when you weave AI into your own processes, customer logic and operational systems.

Help with the technical approach? The coffee is ready

The choice in front of you

Self-service was a good first step. But it was an intermediate step. Your customers' expectations have changed fundamentally. They do not want a portal where they have to search themselves. They want an assistant that understands their question and gives the right answer immediately. The technology to build that exists. Now. That shift is already under way. The question is not whether your customer portal is going to change. The question is whether you are the one who does it, or whether your customers look elsewhere in the meantime. We see this daily at the companies we speak to. The organisations investing now in AI-driven customer interaction are building a lead that will be hard to close. Not because the technology is so complicated. But because the trust and the customer relationship you build with it cannot simply be copied.

Stop sending your customers into your warehouse. Give them an assistant that knows who they are.

Want to talk through AI in your customer portal? Book a conversation

Frequently asked questions about AI in customer portals

What is the difference between a traditional customer portal and an AI customer portal?

A traditional customer portal is menu-driven: the customer navigates through pages themselves, searches for products and leafs through documents. An AI customer portal turns that around. The customer asks a question and the AI assistant gives the right answer immediately, based on customer context, order history and relevant documentation. The customer does not have to know how your system works.

Which systems do I need to connect for an AI assistant in my portal?

The core is a connection to your ERP system (such as Exact Online or Microsoft Dynamics 365) for customer, order and invoicing data. You also connect a knowledge base (often SharePoint or a document database) for product manuals and technical documentation. The AI assistant combines these sources to give context-aware answers.

Is customer data safe with an AI-driven assistant?

Yes, provided you set up identity-based context filtering. That means the assistant shows each customer only the information that customer is entitled to. Price agreements, contracts and order history are strictly separated per customer profile. This is not an optional layer but a hard architectural requirement in every implementation.

How quickly can an AI assistant in my portal be realised?

A first working version can usually be realised within 6 to 12 weeks, depending on the complexity of your ERP connection and the size of your knowledge base. Most implementations start with a defined scope, for example order history and the most frequently asked questions only, and are then extended step by step.