# Fenja AI — Business Description

*Synthesised from the on-site copy of the customer presentation (entrance, timeline, platform overview, deepdive, roadmap, advisory board). Updated to the current customer-facing deck.*

---

## At a glance

**Fenja AI** is a Danish company **and** an AI platform — one organisation, one product, one name: *"trusted, sovereign AI built in Denmark, for Europe."* The platform is **client-managed** — it installs into the customer's own environment (on-prem or controlled cloud), runs on their hardware, and never sends data to a foreign provider.

It brings **language models, organisational knowledge, tools, and agents** together in one architecture so a customer doesn't just run an LLM — they get governed work done under their own control.

**Backing & provenance** (shown across the site):
- **Backed by Innofounder** (Innovationsfonden / Innovation Fund Denmark)
- **Part of AI Lab**, BioInnovation Institute
- **Part of The Regulatory AI-Sandbox** (Datatilsynet & Digitaliseringsstyrelsen)

## The strategic thesis

The pitch argues that **European digital sovereignty is now a security imperative, not a procurement preference**, via a 13-event timeline (Sep 2024 → Q1 2026):

1. **Concentration risk** — 3 US firms run ~70% of Europe's cloud and almost all its AI capacity (Draghi Report).
2. **The "kill switch" is real** — Microsoft disabled the ICC prosecutor's email under a US executive order (May 2025); testified to the French Senate it could not guarantee sovereignty even for EU-soil data, because of the US CLOUD Act (June 2025).
3. **Governments treat dependence as a vulnerability** — Copenhagen's Microsoft bill +72% in 5 years; Copenhagen and Aarhus announced Microsoft exits; the Jan 2026 Danish threat assessment listed the US alongside Russia and China for the first time.
4. **Trade weaponisation** — Jan 2026 tariffs on Denmark and seven EU nations tied to Greenland; the DSA/DMA/AI Act reframed by the US as trade barriers.
5. **Regulation won't close the gap** — *"sovereign capacity has to be built, not legislated."*

Framing question: *"When AI runs Europe, who runs the AI?"*

## Audience

Danish and European organisations that need full control of their own AI — **both public sector and highly-regulated private organisations** (protecting their data, IP, and regulated workflows from US-vendor dependency). The deck is **customer-facing** — an introduction shown to prospects — that also explains Project Bifrost in context.

## The platform — four layers

Positioned against two insufficient alternatives: renting AI from US firms (sovereignty risk) and DIY open-source LLM deployment (*"a model alone isn't enough"*).

1. **Foundation — language model.** A state-of-the-art open-source model, deployed on-prem. The starting point, but not yet Fenja.
2. **Foundation — knowledge.** What makes it Fenja. **Wiki** (organisational/departmental/personal knowledge, captured via the *Fenja AI Compiler* into the structured *Fenja Wiki* with citations) + **routines & memory** (stand-ups, recurring tasks, working memory) — *"a coworker who knows how things get done."*
3. **Tools — how Fenja acts.** Document retrieval (RAG), structured-data/SQL query, system actions (read/write across systems), custom tools.
4. **Agents — how Fenja scales.** Supervisor (orchestration), specialists (subagents), skills (portable capability), workflows (customer-composed, governed end-to-end).

*"Everything you need, with full control"* — one client-managed stack from foundation model up through governed multi-agent workflows.

## Capabilities ("Choose your Capability")

Four progressively-bundled tiers:

- **Fenja Core** (foundational) — essential LLM capabilities with **Fenja Semantic**, a safe custom chatbot that understands the organisation.
- **Fenja Dev** (developer toolset, + Core) — secure AI-supported development platform.
- **Fenja Analyze** (strategic intel, + Core) — agents that find, analyze, and present insights.
- **Fenja Agentic** (automation, + Core/Dev/Analyze) — the complete framework; governed agents collaborate on critical processes.

## Implementation roadmap

*"One foundation, many use cases."* — **Setup** (platform live in your environment: model, SSO, brand voice, baseline policies) → **Knowledge** (capture tacit expertise via AI-led interviews) → **Tools** (wire into warehouses, SharePoint, APIs) → **Agents** (governed end-to-end workflows). A continuous **Govern & scale** band runs alongside all stages (security, compliance, change management, training, advisory-council feedback).

## Project Bifrost — the engagement programme

A parallel initiative (not a product) that brings a curated group of Danish/European organisations into Fenja's product design — *"built with them, not just for them."* Members get three inseparable things: **Community** (a peer network on the sovereignty problem), **Advisory Council** (regular sessions shaping the roadmap), and **Pilot Projects** (early platform access at a reduced price, subsidised by the Innovation Fund). After joining: the team reaches out, the participant gets a project-portal invite, a first advisory-council meeting is scheduled, and pilot participation is discussed individually.

## Advisory Board

A real, named board — *"Bridging industry & sovereign AI"* — drawn from regulated Danish institutions and major enterprises:

- **Søren Friis** — IT Director, DSB
- **William Irving** — Chief Data & Analytics Officer, Norlys
- **Ulla Nygaard Eliassen** — Associate Improvement Project Director, Novo Nordisk
- **Anna Jessen** — Director, Process Excellence & Digitalization, Novo Nordisk
- **Mathies Laursen** — CDO, Nationalbanken
- **Torben Schütt** — Office Director, Center for Cyber and Digitalization, Forsvarsministeriet
- **Mads Nyborg** — Chief Consultant, Data & Analytics, Københavns Kommune
- **Håkon Daltveit** — Chief Consultant, Data & Analytics, Københavns Kommune

## In one paragraph

**Fenja AI is a Danish-built, client-managed AI platform — backed by Innofounder/Innovationsfonden, part of the BioInnovation Institute AI Lab and the Regulatory AI-Sandbox — for public-sector and highly-regulated private organisations that need full control of their own AI. It combines an on-prem open-source model, an organisational knowledge layer (Fenja Wiki + routines/memory), a tools layer, and a governed agent framework into one stack inside the customer's environment, sold as four bundled capabilities (Core, Dev, Analyze, Agentic) and delivered via a Setup → Knowledge → Tools → Agents roadmap. The customer-facing pitch is grounded in a concrete thesis — US cloud/AI dependency is now a security and trade-policy risk Europe must answer by building sovereign capacity — and is reinforced by Project Bifrost (community, advisory council, subsidised pilots) and a named advisory board from DSB, Norlys, Novo Nordisk, Nationalbanken, Forsvarsministeriet, and Københavns Kommune.**