01
The answers exist, but nobody can find them
The knowledge sits in manuals, contracts and systems. Employees would rather ask a colleague than search — and the colleague does not always know either.
Knowledge search across your own sources
Generative AI for businesses
Generative AI that goes live, not just to pilot
We build generative AI connected to your own systems and data — RAG solutions, document flows and AI assistants. We start with a defined case and scale only once it measures up.
Direct with the developersFixed price with a clear scopeYou own the code and data
EU hosting
Your documents stay in the EU
Defined scope
A pilot before a full rollout
Your sources
Answers from your own documents
Handover-ready
Code, data and documentation
Where we create the most value - Three situations where generative AI makes a real difference
You do not need a finished requirements specification. What you do need is a recurring problem, a clear target group and a willingness to start small.
02
The same questions are answered over and over
Support and case handlers spend time on recurring questions instead of the cases that genuinely need a person.
AI assistant with escalation to a human
03
You want to test before you commit
There is an idea and a budget, but nobody wants to lock in a large investment in something that may not hold up in production.
A defined pilot with real users
Less suitable when simple, rule-based automation solves the problem better.
Case study · Herrljunga municipality
An AI chatbot that guides citizens through a complex application
A geothermal permit application requires several details and geographic understanding. We built a chat-based form that asks the next question based on the previous answer and validates the location directly against the ancient monuments register, the land survey authority and the municipal GIS — while the applicant is still in the application.
- Model
- GPT-4o
- Hosting
- Azure OpenAI in the EU
- Validation
- Geodata in real time
What you get - This is what we build with generative AI
One contact, one responsibility and one coherent delivery. We connect the use case, the data and the integration so the solution ends up where the work already happens.
01
AI assistants and knowledge search
We build assistants that answer customers and employees from your own sources — not from general internet content — and hand over to a person when the question calls for it.
- Support and self-service
- Search across manuals and contracts
- Escalation to case handlers
02
Document automation and data extraction
Unstructured text from email, forms and documents becomes structured data. Case handling gets faster and nothing falls through the cracks in follow-up.
- Classification and summarisation
- Extraction into the right fields
- Quality control and logging
03
Integration into your systems
The AI ends up where the work already happens — in business systems, finance systems, CRM and the support platform — instead of in yet another interface to log into.
- APIs and integration layers
- ERP and finance systems such as Fortnox, Visma and Business Central
- Permissions inherited from the system
- Documentation and handover
How an engagement works
A defined first step — without locking in the whole journey
The most common reason AI projects stall at the prototype stage is that the first step is too big. We start with a case small enough to abandon and real enough to give an honest answer.
Map your use caseDefine
We go through your processes and ask outright whether AI is the right tool here. Sometimes the answer is no — and knowing that early saves you time.
You get: a recommended use case, approach and price
Build and test
We connect the solution to your sources and build a version small enough to abandon and real enough to give an honest answer.
You get: a working solution tested by real users
Launch and follow up
We integrate the solution into your systems, measure which questions it handles badly and adjust. You should be able to update sources without calling us.
You get: a live solution, documentation and a maintenance plan
Technology you can own
No black box.
No unnecessary lock-in.
Your documents are stored on servers in the EU. We use OpenAI or similar services, but can set up self-hosted models when the data is sensitive. Data can be encrypted, and you always own your documents and conversation logs.
Hosted within the EUEncryptionSelf-hosted model for sensitive dataIntegration into your systems
Sectors - Generative AI where it makes a difference
The AI solutions we build are always grounded in a sector and its data. These three, for example.
Previous projects
We work with companies across industries. Read our previous projects and what outcomes the solutions delivered in practice.
Before you decide - Frequently asked questions about generative AI
The key things to sort out before you choose an AI partner, a first use case and a delivery model.
What does it cost to build an AI solution?
The cost depends on how many sources the solution has to read, which systems it integrates with, and how high the requirements on security and follow-up are. We define a first use case and present the team, timeline and price before you decide — a defined first step costs considerably less than a full rollout. Our software development pricing guide provides an initial benchmark.
How quickly can we see whether it actually works?
The timeline depends on scope, data quality and integration needs. We start with a defined use case tested with real users — not with internal demo testing. The goal is an honest answer on whether the solution holds up before you invest further.
How do you integrate generative AI into a business system?
Through the system API or database, with an integration layer between the AI and the business system. We read data from ERP, finance systems or CRM — for example Fortnox, Visma or Business Central — let the model work on it and write the result back to the right field. Permissions are inherited from the system, so nobody sees data they do not already have access to, and every call is logged for traceability.
Can you take over a pilot that has stalled?
Yes. We can step into an existing prototype, assess what holds up for production and prioritise what is missing — data quality, integrations, security and follow-up. Rarely does everything need rebuilding. Read how to take an AI pilot to production.
Who owns the solution and our data?
You own the code, the documentation, your documents and your conversation logs. We build with established technology and document key decisions, so your own team or another partner can take over and keep developing the solution. No lock-in with us.
Do you work at a fixed price or on a time-and-materials basis?
We are happy to work at a fixed price when the use case can be clearly defined — which it often can in a first step. For ongoing development we work with a transparent, prioritised backlog. After the first mapping we propose the model that gives you the most control.
How secure is our data when you use AI?
Your documents are stored securely on servers in the EU. We use OpenAI or similar services, but can also set up self-hosted models if you have sensitive data. Data can be encrypted and you always own your own documents and conversation logs. See what your AI policy should cover.
What happens if the AI answers incorrectly?
Generative AI can give incorrect answers. That is why we build in quality assurance with testing, logging, limits on what the model is allowed to answer, and the ability to escalate to a human when uncertain. We follow up after launch — otherwise quality degrades over time.
When is generative AI the wrong choice?
Generative AI does not suit every problem. If the flow is structured and predictable, rule-based automation is often faster and cheaper to build and maintain. If you want to predict a numeric outcome from historical data, conventional machine learning is a better fit. We will tell you when that is the case.
⚡ We answer within 24h!