Efficient matching of candidates to vacancies with our automated classification model
Olympia
AI agency in Arnhem
We build AI agents and applications that work within your workflow, and manage them. For Olympia, we built a classification model for matching.
For Olympia, we developed a classification model that automatically matches candidates with job openings
Olympia, which operates in the temporary staffing sector, wanted to use a data-driven approach to match candidates with suitable job openings more efficiently and accurately. The aim was to improve its workflow and the service it provides to candidates and clients. We developed an automated classification model for this.
Can an AI agent take over a step in our workflow, and where do we still need a person?
At a provincial authority, grid operator or healthcare organisation, applications, invoices and questions pass through the same hands every day. An agent or model can read them, classify them and suggest a response. A member of staff decides what matters.
Reading, classifying and suggesting: AI that takes over a step in the process
Matching and classifying
For Olympia, we developed a classification model to match candidates with job openings more efficiently as part of their workflow.
Answers from internal documents
For Fokker Services Group, we built Anthony Chat, which uses a RAG model to retrieve answers from internal documents.
Only the discrepancies
For Fokker Services Group, an AI model predicts the expected invoice amount, and a dashboard highlights significant discrepancies.
Management after launch
We maintain what we build: updating prompts and models, monitoring quality and making adjustments when the process changes.
We define the handover between the agent and the employee before development begins
-
The step and the decision
Which step does AI take over, which decision stays with an employee, and how do they see the proposal?
-
Build and run in parallel
We build the agent or model and initially run it alongside the existing workflow.
-
Go live with management in place
When it goes live, logging, monitoring and an owner are in place. We also agree who handles ongoing management: your team or ours.
Does our task need an AI agent, or is conventional automation enough?
A good fit when
Staff read and assess many similar documents every day, such as applications, invoices or CVs, and previous assessments have been recorded so the results can be checked against them.
Not a good fit when
The rule is fixed and can be fully specified. In that case, a workflow or a rule in your existing system is easier to manage than an agent.
What determines the scope
How many systems the agent works with, whether personal data is involved, how many steps it takes independently, where it runs and how much management we handle after launch.
What if the agent makes a mistake?
That is why, for decisions that matter, it makes a recommendation rather than a final decision. Every step is logged, so you can see what it did and a staff member can intervene.
An AI feature in your software, your own agent managed by your team, or one managed by us?
A feature in your software
Choose this if your case management system, ERP or CRM already handles the task with a built-in AI feature, and it is clear where the data is stored.
Your own agent, managed by your team
Choose this if your IT or data team can monitor applications in production. We build the agent and hand it over with documentation and tests.
Your own agent, managed by us
Choose this if the process changes frequently and you have no dedicated team for it. Our advice: agree on how it will be managed before building it, through management and continuous improvement.
If you have us build an agent, we can manage it too
Based on Westerdoksdijk in Amsterdam, Data Science Lab has worked with data and AI since 2016 as a Databricks Partner and Microsoft Partner. For organisations in Arnhem and Nijmegen, we work on site or online.
The first conversation is about the workflow step AI should take over
Stef Stoelwinder
Business DevelopmentStef is your first point of contact. In our first conversation, we discuss which step in your workflow AI can take over, in which systems, and who will manage it.
E-mail info@datasciencelab.nl
Phone +31 (0)20 3636 724
Which step in your workflow takes up the most of your team’s time in manual review?
Describe the step, the systems and documents involved, and who currently makes the decision. We’ll reply by email or arrange an initial conversation.
Message sent
Thanks for your message. We’ll be in touch as soon as possible.
Frequently asked questions
An AI agency builds applications that use language, documents or data to carry out or prepare a step in a workflow, then keeps them running. We do this as part of our AI applications.
If you need a predictive model, take a look at our data science agency in Arnhem. For advice on where AI is worthwhile, see AI consultant Arnhem.
No. We mention the Olympia case study as an example of a model embedded in a workflow: matching candidates and job openings with a classification model. We have not yet published a case study from Gelderland.
For organisations in Arnhem and Nijmegen, we work on site or online from our office on Westerdoksdijk in Amsterdam.
Automation follows fixed rules: if this, then that. An AI agent uses a language model to break a task into steps, use tools such as a search function or a connection to a system, and choose the next step itself. That makes it useful for a wider range of tasks, but also harder to predict.
Read more about agents under agentic AI.
By defining, for each step, what the agent can do independently and what must remain a proposal. It might be allowed to code an invoice, for example, but not to approve a payment. Every call and decision is logged, and exceptional cases go to an employee.
That is also how invoice checking works for Fokker Services Group: employees now only need to check the exceptions.
Article 4 requires providers and deployers of AI systems to take measures to support their staff’s AI literacy, taking account of their knowledge, experience and the context of use. This is how it reads in the amended text as at 27 July 2026.
Your legal adviser can interpret what this means for you. We help with training and adoption for the people working with the agent.
Arnhem is home to TenneT’s Dutch headquarters, Alliander and Gelderland’s provincial government offices on the Markt. Organisations like these receive requests, reports and documents every day that someone has to read and pass on.
An agent can read a document, extract the information and propose how to handle it. Whether that is permitted and worthwhile needs to be assessed for each process.
We agree on that in advance. Management means updating prompts, models and connections, monitoring quality against a fixed set of examples, and intervening when a system or process changes. Your team can handle this, or we can.
We only manage what we have built ourselves. For models, we use MLOps: version control, monitoring and retraining.
When a task can be handled with a fixed rule or a form, or when there are no documented examples against which to assess the outcome. In that case, start with a conventional workflow or by documenting decisions.
If you first want to find out which task would deliver the most value, start with advice on AI in Arnhem.
Agents, models and management: where can you find the rest?
- AI applicationsAll the applications we build: prediction, images, text, generative AI and agents.
- Data science consultancy in ArnhemPredictive models for grid management and healthcare in Gelderland, built for production.
- MLOpsVersion control, testing and deployment, monitoring and a set schedule for retraining.