AI Integration
From mapping to concrete plan — without the hype.
TL;DR
We come into the business, map how you actually work, and identify where time is being lost. Then we match those time sinks to AI tools that fit your specific situation, and deliver an implementation plan with tool choices, use cases, pricing, timelines, and estimated time saved.
TL;DR
AI integration at WOLLUM means language models, RAG search, and automation are wired into the places work actually happens — not as a separate platform users have to learn. Typical delivery: one defined process in production in 4–8 weeks, priced NOK 80,000–250,000. We build on an abstraction layer so the model can be swapped when a better or cheaper one arrives.
Definition
AI integration is the process of wiring artificial intelligence directly into a business's existing systems and workflows — Microsoft 365, ERP, CRM, email, and document archives — so AI runs in the background or in the interface users already work in, without requiring new tools or changed workflows.
What we deliver
Discovery and scope
We pick one concrete process where the time saved is biggest, and set concrete targets for time saved, error rate, or response time before we start.
Model selection and architecture
GPT, Claude or Gemini — we pick based on performance, cost, and GDPR requirements for your use case. Built with an abstraction layer that lets us swap models without touching business logic.
Integration with existing systems
APIs into Microsoft 365, Tripletex, Visma, PowerOffice, Altinn, EHF, SharePoint, and line-of-business systems. Older systems without APIs via RPA.
Approval, logging, and fallback
Humans approve in the early stages. Every response is logged. When something deviates from expectation, it falls back to a manual queue — never into a void.
Operations and ongoing development
Handover, documentation, and training. You own the source code. A maintenance agreement is optional, not mandatory.
Timeline
Typical delivery: 4–8 weeks for one process
Discovery in week 1. Design and prototyping in weeks 2–3. Build and integration in weeks 3–6. Pilot in production with a limited user group in weeks 6–8. Larger integrations spanning multiple systems take 3–6 months.
Pricing
Pricing: NOK 80,000–250,000 for one defined process in production. NOK 250,000–600,000 for integrations connecting multiple systems or with their own logic. Running cost: typically NOK 500–5,000/month per process depending on volume.
What you walk away with
- One concrete process running in production with measurable time savings
- Integrated with the systems you already use — no new tools for employees
- Logging and monitoring in place from day one
- Safe fallback when the AI is uncertain
- Full source code and documentation that you own
FAQ
Do we have to replace our existing systems to use AI?
Rarely. The point of integration is that the systems stay. We connect AI to them via API or similar, and the workflow you have today stays unchanged.
What about GDPR and sensitive data?
We always set up in EU regions with data processing agreements in place. For particularly sensitive data we can use models that run locally or in Norwegian cloud. The GDPR assessment always happens during discovery.
What happens when the AI fails?
We always build with a safe fallback. If the model is uncertain, the case falls to a manual queue — never into a void. All activity is logged.
Do we own the solution once it is built?
Yes. You own the source code, documentation, and data. We offer maintenance agreements, but you can also run operations yourself or choose another vendor later.
Read more: in-depth article on this topic
We build what you need. Let's figure out what that is.
Send us two sentences about what's grinding. You'll hear back today, and we'll tell you whether this is something we can help with.