AI that runs in production, not in a pilot
Appfront is an AI agency from Amsterdam. We build custom software in which AI is a genuine component rather than a label, integrated with the systems where your people already work.
- Runs on your own documents and data, using an existing language model that we set up around them
- Built to run in production, not to stay stuck in a pilot
- Integrates with the ERP, CRM or record system where the work already happens
Organisations we've helped
From carrier to employers' organisation. What these clients have in common is that the work had to run in production, not in a trial setup.
What we've delivered
Three systems running in production every day, with the figures our clients report themselves. The first is the AI project; the other two show that we finish software and get it adopted.
AWVN
75% faster at answering a collective agreement question. More than 10,000 collective agreements spanning thirty years, made searchable for the advisers who work with them.
GVB
60% shorter turnaround when drafting traffic messages, and 45% fewer incomplete diversions in passenger information.
De Haas Road Cargo
90% less manual work in the order process, with real-time visibility of packaging and returns for the planners.
How we work
We start with the work that is currently done by hand and the data you already have. Only once it is clear where the time or the errors are do we decide whether AI is the right tool for the job.
Then we build it as part of software your people use every day, with an integration to the systems where the work already happens.
What we build with AI
For us, AI is not a standalone service but part of software your people use every day. These are the six applications we deliver most often.
Reading and processing documents
Invoices, contracts and forms that are currently retyped by hand. The system reads out the fields you need and sends doubtful cases to a person.
Searching your own data
Your knowledge sits in thousands of documents that nobody can keep track of any more. A search layer answers a question in plain language and shows which records the answer is based on. This is how the collective agreement application for AWVN works.
Predicting from history
A machine breaking down, weekly demand, the risk of a late order. A prediction only works on the screen where the decision is made.
Classifying and routing
Sort incoming work, from mailbox to ticket flow, by type, urgency and handler. Rules that currently live in someone's head become explicit and testable.
Conversations and speech
An assistant that answers questions based on your own documentation, or speech that ends up as text in the right record.
Integrating with what you already use
A model only delivers value when its output ends up in the ERP, the CRM or the record where the work happens. We build those integrations ourselves.
Meet the team
At Appfront, we believe technology and innovation go hand in hand. Our team of experts is the heart of our success. Each member brings a unique set of skills and a shared passion for smart digital solutions. Meet the driving forces behind Appfront:
Thirty years of collective agreement data, made searchable
AWVN advises employers on terms of employment and manages more than 10,000 collective agreements spanning thirty years. That knowledge sat in an Excel file nobody could keep track of any more; what you found depended on who was searching. We built the application that lets advisers search and substantiate that history. They now answer a collective agreement question 75% faster on average, and more importantly, every piece of advice rests on the same data and the same methodology.
What happens to your data
With AI, this is usually the first question, and rightly so. We process data in EU data centres, require two-factor authentication on all company accounts and record per project who can access which data. Where a sector standard applies, such as NEN 7510 in healthcare, BIO in government or DORA in the financial sector, we work to it. During onboarding we share architecture overviews and pentest reports on request.
What we don't say: that we are ISO 27001 certified. That process is in preparation, nothing more. You can read exactly how we have set things up in our information security policy.
Custom AI, an off-the-shelf AI tool or no-code?
Building AI is not always necessary. Here is how the three routes compare.
| When it fits | What to watch for | |
|---|---|---|
| Custom AI | The AI has to work with your own documents and data, and the result has to end up in your ERP, CRM or records system. | More work upfront: data in order, integrations and maintenance. After that, it runs within your own workflow. |
| Off-the-shelf AI tool | General writing and search tasks for individual employees, for example with ChatGPT Enterprise or Microsoft Copilot. | Knows nothing about your processes and integrates only to a limited extent with your systems. Arrangements about where data is stored differ per vendor. |
| No-code AI builder | A simple chatbot or an automation with an AI step, for a clearly defined task. | Limits when it comes to custom logic, permissions and checking the output. Hard to develop further. |
We will tell you honestly whether AI is the right tool here, and if not, what is.
Frequently asked questions about building AI
What does Appfront build with AI?
Software in which AI is one component: reading out and processing documents, searching your own data, forecasting based on history, classifying incoming work and assistants built on your own documentation. We also build the integrations with the systems where the work happens.
Do we have to train a language model ourselves?
Usually not. We set up an existing language model around your own documents and data. Where necessary, a custom model or an additional step is added, but that only becomes clear once we know your data and your question.
Where does our data stay?
We process data in EU data centres and record per project who can access which data. Where a sector-specific standard applies, such as NEN 7510 in healthcare, BIO in government or DORA in the financial sector, we work to it.
Isn't an off-the-shelf AI tool enough?
Check that first. For general writing and search tasks for individual employees, an off-the-shelf tool is often enough. Custom makes sense when the AI has to work with your own data and the result has to land in your own systems.
How do we get started?
With a half-hour conversation in which you tell us which work is currently done manually and what data you have. You get an answer from someone who builds it. After that, we choose a first, clearly defined use case together.
What if AI is not the right solution for us?
Then we will say so. Sometimes a standard integration or a better-organised process is the better step. We can build that too.
Curious whether AI would pay off for you?
Tell us which work is currently done by hand and what data you have. We'll tell you honestly whether AI is the right tool here, and if not, what is.
Submit your AI challenge
Tell us which work is currently done by hand and what data you have. We'll tell you honestly whether AI is the right tool here, and if not, what is. You'll get an answer from someone who builds it.
Prefer to email directly? You can reach us at info@appfront.nl.
Still comparing options? In our overview of AI agencies in the Netherlands we are listed alongside other providers, with their strengths noted.
Where to find us
Our office is in the Houthavens, a fifteen-minute walk from Amsterdam Centraal station.