Service 03
AI for small and mid-sized companies
We do not sell "AI transformation". We identify specific tasks where AI saves real hours, implement them with human verification, and measure the outcome.
Services
Where AI actually pays off
- Hundreds of invoices and delivery notes keyed in by hand every month.
- The same customer question answered twenty times a week.
- Information buried in contracts, manuals and procedures nobody can search quickly.
- Proposals written from scratch when 70% of the content is always the same.
- Complaints and requests waiting on manual triage.
What we implement
Every use case starts as a measured pilot. If it does not save time, it does not go to production — and we will tell you.
Document data extraction
Invoices, delivery notes, contracts and forms read automatically, with validated data flowing into your systems and human review wherever the risk justifies it.
Deliverables
- Pilot on your real documents
- Measured and reported accuracy rate
- Exception queue for human review
Internal assistants over company documentation
Your team asks in plain language and gets answers with the source cited, drawn from your procedures, contracts and manuals — not from the open internet.
Deliverables
- Organized knowledge base
- Assistant with source citations
- Role-based access control
Request triage and drafting
Emails and requests classified, routed and answered with a drafted reply, leaving the final decision to a person.
Deliverables
- Classification rules
- Drafted replies with human approval
- Response-time metrics
Commercial content support
Proposals, tender responses and datasheets generated from your templates and history, so your team reviews instead of writing from scratch.
Deliverables
- Templates and content library
- Assisted generation with review
- Time saved per proposal
Governance, GDPR and training
Clear rules on what data can be used, where it is hosted and who validates the output — plus practical training for the team.
Deliverables
- AI usage policy
- Risk and GDPR assessment
- Role-based training sessions
How we evaluate
Pilot
before any long commitment
Accuracy
measured on your own documents
Human
in the loop wherever risk exists
EU
data hosting available in the European Union
How we work
A predictable path, from diagnosis to a measured result.
We work in short cycles. Every phase has concrete deliverables and a clear decision at the end — you can stop at any point.
Frequently asked questions
Will our data be used to train third-party models?
No, and we put that in writing. We use enterprise configurations where submitted data is not used for training, and where the case requires it we propose EU hosting or models running on controlled infrastructure.
What if the AI gets it wrong?
We assume it will, in some percentage of cases, and design the process around that: confidence thresholds, human review on exceptions and continuous accuracy measurement. That is why we always start with a pilot on real data.
Do we need AI, or just automation?
Often just automation — and we will say so. AI earns its place when the input is unstructured: free text, scanned documents, natural-language requests. For everything else, simple rules are cheaper and more reliable.
What does the EU AI Act change for us?
For most SMB use cases the obligations are about transparency and documentation, not heavy certification. We keep a record of which systems you use, for what purpose and with what oversight, which covers most of what is required.
Let us look at one of your processes.
Pick the process that hurts most. In a 30-minute conversation we will tell you what can be automated, the effort involved and the return you should expect.