AI integration
We bring in AI where it replaces identifiable repetitive work, not as a selling point. Depending on how sensitive your data is, models run either at a provider or entirely on your own infrastructure.
- Models can run on your own infrastructure
- You know where your data goes
- Measurable productivity gain
What we offer
Document assistants
Query your document base in plain language
- Conversations with your internal documents
- Sourced, verifiable answers
- Per-team knowledge bases
- Access rights kept separate
- Local deployment available
Transcription and summarisation
Automatic minutes for meetings and interviews
- Transcription of discussions
- Summaries and decision logs
- Speaker identification
- Export to your business tools
- Can run without any external service
Language processing
Automatic analysis and classification of text content
- Document classification
- Structured information extraction
- Sentiment analysis
- Automatic summarisation
- Translation
Process automation
Intelligent workflows built into your existing applications
- Automatic data entry and pre-filling
- Document validation
- Intelligent request routing
- Edge-case handling
- Integration with the tools already in place
What it changes for you
Data sovereignty
With self-hosted Ollama or Mistral, no data ever leaves your infrastructure.
Measurable gains
We target repetitive tasks whose cost can be quantified, so the benefit is observed rather than assumed.
Built into the business
AI slots into your existing applications instead of forcing yet another tool on your teams.
Controlled cost
The model is chosen for the actual need: a small local model often suffices where a metered API would be overkill.
Technologies we use
Mistral AI
European language models, capable and open
- European hosting
- Open models
- Controlled cost
Ollama
Running language models on your own infrastructure
- Local deployment
- No outbound data
- Full control
AnythingLLM
Conversational interface over your document bases
- Sourced answers
- Multiple sources
- Fine-grained access control
OpenAI
General-purpose models where raw capability outweighs data locality
- Broad comprehension
- Multilingual
- Stable API
How it runs
Step 1
Identifying the use case
We look for the repetitive task whose cost can be measured. If none stands out, we say so rather than adding AI for show.
Step 2
Prototype
A working prototype on your real data, to confirm the approach before any commitment.
Step 3
Integration
Production rollout inside your applications, with the model and hosting mode chosen according to data sensitivity.
Step 4
Measure and adjust
Tracking real-world results and tuning models and prompts as usage settles.
A project along these lines?
Describe what you need and we will tell you plainly whether it is within our reach.
Our other services
SaaS solutions
Business software hosted, maintained and reversible
Mobile development
Native apps, cross-platform apps and PWAs
Managed services and maintenance
Your servers administered, backed up and monitored