Case · Hempel · Marketing AI Hub

AI is everywhere. Advantage is not.

Hempel - a world-leading coatings supplier, founded in Denmark in 1915, with 7,000 colleagues and 24 factories worldwide - spent three years proving AI could lift productivity. It did. The advantage they expected did not follow. The Hempel Marketing AI Hub is the foundation we built together underneath the models, so the next three years compound instead of evaporate.

1915

Founded in Denmark

7,000+

Colleagues worldwide

24

Factories globally

4

Industries: Decorative · Infrastructure · Marine · Energy

The market reality

Everyone uses AI. Almost no one scales it.

The early movers in B2B marketing were right about one thing - AI lifts productivity. They were wrong about the second part: that lift does not become a moat, because every competitor gets it too. The real gap is structural, and it sits underneath the models.

88%

of companies use AI

McKinsey · State of AI 2024

7%

have managed to scale it

McKinsey · State of AI 2024

60%

lack the data foundation

Bain & Company 2024

Client perspective

“AI is only as strong as the foundation you build it on.”

Rasmus Vestergaard · Head of Marcom & Market Intelligence · Hempel

The journey

From experiments to a verified brain.

Three years of moving up the curve - from individual tools, to agents, to productivity wins, to the realisation that none of it adds up to advantage without a shared source of truth underneath.

2022

Baby steps

Hempel's early experiments with segmentation, generated imagery and the first wave of GPT-3.5 tools. AI felt like an unfair advantage waiting to be picked up.

2023

Agents and personas

Custom agents built with Birdie to handle research, briefing and copy tasks for Hempel's marketing team. Output went up. Speed went up. So did everyone else's.

2024

Productivity proof

Hempel's case-story output multiplied 3×. The global hempel.com replaced 14 manually maintained language sites with an AI-translated stack wired straight into the CMS.

2025

From content to system

Realising productivity alone is not transformation. Rebuilding around the full marketing G2M process with a verified brain - the Hempel Marketing AI Hub.

case-story output, with the same team

14 → 1

manual language sites replaced by one AI-translated .com wired into the CMS

EU

resident, self-hosted, reviewed by compliance 3×

The turn

Start with the problem. Use AI afterwards.

The first three years answered a different question: how do we use AI? The honest answer turned out to be uncomfortable - faster content, no advantage, same workload. So Hempel put the enthusiasm down and went back to the actual marketing problem.

What followed was a structural rebuild around Hempel's full go-to-market motion: aligning scope across markets, compressing brief-to-launch, removing the ceiling set by team size, and making every piece of output genuinely relevant to its segment - not just translated into the right language.

The brief

Four marketing problems AI should actually solve.

01

Reduce G2M complexity

Aligning scope, ambitions and insights across markets and segments around one shared source of truth for brand, products, segments and personas.

02

Reduce time-to-market

Compressing brief-to-launch cycles that used to be measured in months, with studios speeding up production today while the memory and graph quietly shorten every next campaign.

03

Increase scalability

Lifting the ceiling set by team size, by letting agents work from the same shared facts in parallel, which is what actually makes a 10x step-change possible.

04

Greater relevance

Applying persona, market and segment context to every piece of output, so the result feels like genuine local adaptation rather than word-for-word translation.

What we built

Not a prompt wrapper. The orchestration layer that makes frontier AI actually work.

We do not build the frontier models themselves. We build the layer between them and Hempel's marcom stack - across five architectural zones covering ingest, memory and graph, agents, studios and the surfaces Hempel's marketers use every day, all on standard open-source components and EU-resident infrastructure with a full audit trail.

The ingest pipeline turns files, meetings and signals into structured, citable facts, while the memory graph keeps learning continuously, so the studios that speed up campaign production today also cut the setup time for every campaign that follows.

The Hub is co-built with Hempel's marketing team in a weekly sprint rhythm - no quarterly release cycles to wait for, which is how ideas reach production in weeks rather than quarters.

Could your organisation use the same?