Document extraction
Convert PDFs, emails, scans and forms into structured data. Validated line by line against the printed total, running daily, with a human in the loop for edge cases.
Custom AI
First working AI solution in your process, within 4 to 8 weeks
No slide decks, no loose demos. We take one concrete process, build an AI solution on your own data and prove it delivers value before you commit to anything long-term.
- Free scan · 30-60 min
- Fixed price per module
- Governance & GDPR by design
Incoming invoice.pdfSupplier recognised · 3 pages
ReadLine items extracted14 lines · total matches printed amount
ValidatedEdge caseUnknown article no. → for review
Human decidesLoaded into ERPWith audit trail and source reference
DoneProcessed today: 212For review: 4Avg. lead time: 38 s
1source of truth
instead of 12 spreadsheets
instead of 12 spreadsheets
EU-hosted · audit trail
human decides on doubt
human decides on doubt
Why this approach
Four stories we hear often. Recognise one?
We have ChatGPT licences, but nobody uses them for the real work.
Generic chat doesn't know your documents, customers and rules. We build AI that does connect to your systems, data and workflow.
Our people retype the same reports from the ERP every week.
Document extraction and structured output. Built once, it runs daily, with a confidence score and human review for edge cases.
We did an AI pilot before. Then it stalled.
Pilots that aren't designed for production remain experiments. We start with production as the end state and build towards it in a controlled way.
Everything I read about AI is hype. What actually works for my organisation?
Honest answer in a 30-minute scan. Sometimes AI is the solution, sometimes a plain integration or a better form. We'll tell you.
If one of these hits home, that's a good starting point for an AI Opportunity Scan.
What we build
Six patterns we apply every day
Not one template for everyone, but proven building blocks we adapt quickly to your data, context and workflow.
Chat over internal knowledge
A knowledge assistant that knows your documents, manuals and processes. Answers with source references and exact calculations instead of estimates, with role-based access.
Agents with tools
AI that doesn't just answer but takes action: fetching data, preparing quotes, routing tickets, entering data. With an audit log and approval steps where they belong.
Quote from request, drawing or specification
Requests arrive as PDF, Excel, photo or drawing. An intake reads them, matches positions against your price logic (which the AI never sees) and prepares a draft quote in brand style. The sales rep checks and sends. Proven for access systems, in development for glass office spaces and custom timber.
Work order, inspection and certificate files
For companies with regulated assets: automatically completing work orders, certificates and maintenance history per installation or component, flagging gaps and expired inspections. The same pattern as our document validation in aviation.
AI copilots in your own software
AI embedded in your web application: as a side panel, smart field or natural-language search. Part of the interface, not next to it. For example from drawing to draft quote in one click.
Proven, not theoretical
What is already running
Production
Food & Agri · Document AI
6,800 supplier invoices from PDF into one procurement database
A tomato grower with sites in Belgium and the Netherlands had six years of invoices as PDFs, in dozens of layouts, via email and WeTransfer. A pipeline reads them automatically from SharePoint, recognises supplier and entity, extracts every line and checks the total against the printed amount. Edge cases reach the buyer as proposals, approved with one click. On top of the database: a procurement dashboard with price monitoring, categories and data quality.
6.873PDFs in the archive processed
100%checked against printed total
Proof of concept
Aviation · Document validation
Thousands of pages of maintenance documentation per aircraft engine checked
With every purchase of an engine or part, the full history must be right: every day, flight cycle and flight hour accounted for, in unstructured packages of thousands of pages. Errors hide in details humans miss: a wrong serial number, a non-matching period, a missing signature. We designed a pipeline that classifies documents, extracts key fields, builds a timeline per component and bundles deviations for the engineer. Their own chat environment on the client documentation is already running.
5 → 1days of review per package (target)
Humanreviews every deviation
Proof of concept
Engineering & Installation · Quote agent
From request with drawing to draft quote in brand style
A supplier of access and parking systems received requests as PDF, Excel, Word and photos. An intake reads the request with vision, matches positions against a confidential price model that never reaches the agent, and renders a draft quote as Word and PDF. Validated against real quotes to the euro. New: a dimensionally accurate 3D product visualisation generated from the calculation model.
€66.362test quote matched to the euro
Minutesinstead of days
Names anonymised for client confidentiality.
How to start
Start small, prove value fast
Four steps, each with a clear result and a decision point. You're not committed to anything until the value is proven.
Free
AI Opportunity Scan
A 30 to 60 minute conversation with your team. We determine which process costs the most time and whether AI is the right solution.
- 3-5 promising use cases
- Impact, feasibility and fit
- Advice on the first step
Free 2-4 weeks · max. 2 feedback rounds
Proof of concept
One defined process, your real data, a working first version and an evaluation set that measures whether it works. At most two feedback rounds, then a go/no-go.
- Built on your documents
- Measurable result
- Honest go/no-go
€5,000 - €10,000 fixed price
Working module
Delivered production-ready for one agreed amount: authentication, monitoring, cost control, fallback paths and audit log. AI that keeps running without us.
- Fixed scope, fixed price
- Live in 4-8 weeks
- Documentation and hand-over
from €450 per module / month
Maintenance & development
Hosting, monitoring, model updates, evaluations and a monthly review. AI usage is billed separately based on use.
- EU hosting included
- Model updates and evals
- Next use case lined up
Prices exclude VAT and AI usage (tokens). Complex integrations or multiple modules are quoted upfront, always fixed-fee per phase. No open-ended contracts.
Approach
Four phases, honest about week one
We start small, validate fast and only build for production once the proof of concept has proven its value.
1Week 1
Use-case scan
Problem and data mapped, AI weighed against classic automation, success criterion agreed.
2Week 2-4
Proof of concept on real data
Working version on your documents and systems, an evaluation set with metrics, and an honest conversation about what does and doesn't work.
3Week 4-8
Production build
Authentication, monitoring, cost and rate limits, fallback paths, audit logs and human-in-the-loop where needed.
4Ongoing
Further development
Models and data change. Monthly review, periodic evaluations and the next use case once the first one works.
Security & governance
Control is part of the design, not an afterthought
Your most sensitive information passes through our hands: financial data, customer details, contracts. That's why traceability, access control and human decision points are built in from day one.
EU hosting and EU endpointsNo training on your dataRole-based accessLogging and audit trailHuman decides on doubtCost and rate limits
Full autonomy is rarely the best starting point. A hybrid setup with AI support and review moments builds trust faster and improves adoption.
Frequently asked questions
What organisations often ask us about AI
We already have a ChatGPT Enterprise licence. Why you?
ChatGPT is an excellent personal tool, but knows nothing about your documents, customers or systems. We build the layer underneath: embedded in your workflows, with your data and the right guardrails. In short: ChatGPT helps someone type an email faster. We make sure that email is no longer needed.
What about data privacy and the GDPR?
We host in the EU by default. Models from Anthropic, OpenAI and Google run via EU endpoints under terms where your data is not used for training. For sensitive data we deploy open-source models on your own infrastructure. We arrange the data processing agreements together with your DPO.
Which model do you use?
It differs per use case. Document extraction works well on Claude or GPT, classic classification can run on a small open-source model, for agents we often pick Claude for its strong tool use. We isolate the model choice behind a gateway, so you're not locked into one vendor when prices or policies change.
How do you measure whether it really works?
Before we build, we define an evaluation set: a few hundred examples for which we know the right answer. On that we measure accuracy, cost per call and lead time. When models update, we rerun the evals. In production we log every call, so we spot drift and regressions early.
Do you build in our repo or your own environment?
In your repo, on your cloud, with your team as co-developers, or fully hosted and managed by us in the EU. Either way we don't leave behind code only we understand.
What if AI isn't the right solution for our case?
We'll tell you honestly. A well-built form, a validated database or a plain API integration often solves more than a language model. We recommend what solves the problem, not what sounds most exciting.
AI Opportunity Scan
Discover your AI use case with the most impact
Book a free 30-minute scan. We explore where custom AI can reduce manual work and where it makes no sense. Reply within one business day, from an engineer.
- Reply within one business day, from an engineer
- Free and no strings attached
- Your data stays within the EU
Thanks, we received your message.
An engineer reads your message and replies within one business day.