Skip to content

AI agency in Zwolle

We build an assistant that answers questions using your own knowledge and systems, and keep it up to date. That’s how we developed Dex for PostNL.

We developed Dex for PostNL, a self-learning chatbot that serves as a central point for employees’ questions

Dex is connected to PostNL’s existing systems, so the knowledge base does not need to be stored and maintained in multiple places. This gives service desk staff more time to handle complex issues.

Read the case study

Why do colleagues keep asking the service desk the same questions?

The answer is somewhere in a procedure, a policy or a system, but no one can find it quickly. An assistant connected to your own sources answers straight away, shows where the answer came from and passes on questions it can’t answer.

Book a meeting

Answers from your own sources, for staff and at the counter

Discuss your challenge

An internal service desk

For PostNL, we developed Dex, connected to existing systems, as a central point for staff questions.

Answers at the counter

For NFZ, we built a virtual assistant that provides contract information and policy terms directly at the pharmacy counter.

A language model for the team

For Fokker Services Group, we built Anthony Chat, an LLM application that saves time and supports productivity.

Staff who can put it to work

For AFWC, we delivered AI training in which participants practised with their own tasks and built an agent.

An assistant is only as good as the sources behind its answers

  1. Sources in order

    Which documents and systems are authoritative, who maintains them and who can see what.

  2. Testing with real questions

    The assistant answers questions your service desk has already received, and your team assesses the answers.

  3. Live and kept up to date

    In Teams, on your intranet or in your own environment, with logging and a regular review of new questions.

Can an assistant really take work off our team’s hands?

Test it in a conversation

A good fit if

The same questions keep coming up, the answers are in documents or systems, and someone keeps those sources up to date.

Not a good fit if

The answers aren’t documented anywhere, or every question needs a bespoke response from a specialist. In that case, the first step is to document the knowledge.

What determines the scope

How many sources and systems need to be connected, whether answers vary by role, which channel the assistant will be available in, and who will manage it.

What if the assistant makes something up?

It only answers using the sources we connect and cites them. If it doesn’t know, it says so and passes the question on.

An FAQ page, a standard chatbot or an assistant using your own sources?

An FAQ page

Choose this if you have a limited number of questions that rarely change. Easy to write well and maintain.

A standard chatbot

Choose this if you mainly need predefined paths, such as a form or a referral, without using your own knowledge sources.

An assistant using your own sources

Choose this if the answers are in your own documents and systems. Our advice: start with one team and the questions that come up most often.

Since 2016, we’ve been building AI assistants that employees use every day

Data Science Lab has been operating since 2016 and is a Microsoft and Databricks partner. We’ve worked with PostNL, NFZ and Fokker Services Group, and share what we learn at our Techdays.

Book an introductory meeting

The questions your service desk gets every week are the starting point for the first conversation

Stef Stoelwinder

Stef Stoelwinder

Business Development

Stef is your point of contact. In the first conversation, we’ll discuss which questions keep coming up and where to find the answers.

Should the assistant start by answering HR, IT or process-related questions?

Tell us who asks the questions and where the answers are currently kept. We’ll reply by email or arrange an initial conversation.

Enter a valid email address.

Let us know how we can help you.

Frequently asked questions

An AI company like Data Science Lab builds and maintains applications that take over or speed up tasks. This page focuses on an assistant that answers employees’ questions using your organisation’s own knowledge and systems.

You can find all our applications under AI applications.

No. We work from Westerdoksdijk in Amsterdam. We work online with organisations in Zwolle and the rest of Overijssel, and visit for workshops and testing sessions.

None of the case studies here took place in Zwolle; they show what we build.

With retrieval-augmented generation: for each question, the assistant first finds the relevant passages in your sources. The language model then formulates an answer using only those passages and includes a source reference.

You can read more about this technique in LLMs and knowledge management.

Yes, if the systems are connected. Dex, for example, is connected to existing systems at PostNL, so knowledge does not need to be maintained in multiple places.

We set permissions for what each role can access.

Where your employees already work: Teams, the intranet or your own web page. An assistant in a separate interface tends to be used less.

For a chatbot for customers or reception staff, see Building an AI chatbot.

We use a fixed set of test questions answered by your own people. We run the tests again after every change to the sources or model. Once the assistant is in use, we can see which questions did not receive a good answer.

Those questions determine what we update next.

You need agreements on where the model runs, what is logged and who can access which sources. Employees must know they are talking to an AI assistant.

Read more about the AI Act.

We agree on that in advance: us, your team, or us initially and then your team. The department that owns the sources keeps them up to date; we or your IT team manage the model and its integrations.

Read about how we set this up under MLOps.