Digital transformation and AI programmes, delivered.

I'm David Brandt, a programme leader based in Copenhagen. Over 20+ years I've been a developer, consultant, CEO and programme manager.

Today I help organisations deliver complex change, build an AI portfolio that pays off, and get people to use what's been built. Available for interim and advisory work through Kofoed Brandt ApS.

[Portrait]

Services

  • Programme delivery

    Leading large, cross-team transformation programmes from plan to go-live, including turning around ones that have stalled.

  • AI portfolio

    Setting up how AI ideas are collected, prioritised and taken from proof of concept to production.

  • AI adoption

    Change management and enablement, so AI tools become part of everyday work.

  • Interim leadership

    Interim head of a delivery, PMO or engineering team. I lead the people, stabilise the organisation and hand over to a permanent hire.

Selected work

Global logistics

[Programme delivery case]

[One or two sentences: the situation, what you did, and the result.]

[Industry]

[AI portfolio or adoption case]

[One or two sentences: the situation, what you did, and the result.]

Nordic banking, 2005–2008

AI advisory solutions

As CEO of Intellix, I built an investment advisor and a debt advisory tool for midsize businesses for a Nordic bank.

“[A one- or two-sentence testimonial from a former manager or client, used with permission.]”

[Name], [Title], [Company]

Side projects

Things I build in my own time to stay hands-on with AI. Each one is on GitHub with the code and a short write-up.

[Photo or demo GIF of the robot]

Robotics · voice and LLM agent

Agentic Lego Boost robot

A Lego Boost robot I talk to in Danish. Speech-to-text runs on the laptop's own neural engine, and an LLM agent turns requests into movement. The first version was steered by hand gestures using webcam pose tracking.

[Screenshot of SIFT]

Product · AI portfolio management

SIFT: AI portfolio system

An intake and prioritisation tool for AI ideas: collect proposals, triage them, and track the promising ones from idea to production. Built as if I were setting AI direction for a fictional company, and run at zero hosting cost.

[Results chart]

Experiment · model evaluation

Local LLMs vs. classic ML

A benchmark on the Banking77 intent dataset, run on a MacBook. An 8B local model scored 64% zero-shot; embeddings plus logistic regression scored 93.1%, matching a fine-tuned BERT. The biggest model isn't always the right tool.

About

I started out writing AI software in 2000 and later ran Intellix, selling AI decision support to Nordic banks and insurers. After that came years of consulting and .NET development alongside project leadership.

Today I lead technology programmes in global logistics, covering AI, SAP Transportation Management and event-driven platforms. I'm at home in the steering committee and in the architecture review, and a lot of my value is translating between the two.