Embedded where you work
AI lives inside your CRM, your inbox, your internal portal — not in a separate dashboard nobody opens.
Service · 03
AI integrations · Taastrup, DK
Practical AI, inside the tools you already use.
Summaries, classification, drafting, smart search — embedded where your team works, not in a separate ChatGPT tab.
Boring tech where it matters
I use the model that works, not the model that is on the press release. OpenAI, Anthropic, open‑source — judged on the output, not the brand.
Honest about limits
LLMs hallucinate. I build the guard rails — RAG, citation, human‑in‑the‑loop — so the AI never silently breaks something.
Your data, your choice
I will tell you exactly what is sent to which provider, what is logged, and how to switch providers later. No vendor lock‑in by accident.
What's included
- Discovery: which task LLMs would actually help with, and where they would hurt
- A short written proposal with scope, fixed price and deadline
- Integration into your existing systems (CRM, inbox, support tool, internal app)
- RAG or fine‑tuning when needed — not when fashionable
- Evaluations against your real data so we know quality before launch
- Cost monitoring and per‑user limits
- Documentation for non‑technical staff
- One free month of fixes after go‑live
A good fit for
- Consultancies drowning in incoming documents — classify, summarise, route
- Support teams with a known knowledge base — RAG‑powered answers
- Sales teams drafting the same kind of email 50 times a day
- Operators with searchable archives where keyword search is no longer enough
Not the right fit if
- You want a public chatbot that talks to strangers without oversight
- You expect AI to replace a senior employee one‑for‑one
- You need on‑premises‑only LLM hosting at enterprise scale
How we would work
- 01 A free 30‑minute call to identify a task LLMs are actually good at.
- 02 A one‑page plan with scope, fixed price, and an evaluation rubric.
- 03 Build a small pilot against your real data, with measured quality.
- 04 Roll out, monitor cost and quality, free fixes for the first month.
Common questions
Self‑hosted or hosted LLM?
Hosted (OpenAI, Anthropic, Google) for most cases — quality is highest and Danish handling is excellent. Self‑hosted (Llama, Mistral, etc.) when data sensitivity or volume justify the operational cost. I will walk you through the tradeoff for your case.
How do you handle hallucinations?
Retrieval‑augmented generation grounded in your real documents, mandatory citations, and human‑in‑the‑loop for any irreversible action. We also evaluate against a held‑out set of real cases before launch so we know the failure rate, not guess at it.
What does it cost to run?
Token cost varies by model and volume. I build in cost monitoring and per‑user caps so a runaway prompt cannot blow out your bill. For most small‑business use cases, expect API costs in the range of €20–€300/month — we will model it more precisely once we know the actual workload.
Will my data train someone else's model?
Not if we set it up correctly. I configure the provider settings so your inputs are not used for training, and document exactly what is logged and for how long. If you need stricter guarantees, that points us toward self‑hosting.
Contact
Tell me about your project.
A few sentences is plenty. I'll reply within a working day.