When I entered the project, the company's digital landscape was characterised by what I call "Digital Patchwork Syndrome." Over the past decade, software had been purchased on an ad-hoc basis: a CRM for the sales team, a separate ERP system for warehouse and finance, a third-party marketing tool — and a range of Excel sheets acting as the glue holding it all together.
The problem was fundamental: data was locked in silos. Management made decisions based on reports that were 30 days old, and employees spent up to 20% of their time manually moving data from one system to another. They wanted to implement AI, but the harsh truth was that their foundation wasn't ready for intelligence. AI without structured data is like a fast car without a road — it goes nowhere.
Phase 1: Diagnosing the digital debt. The digital strategy didn't begin with buying an AI licence. It began with a deep audit of three pillars: technology, data and people. We discovered the company was paying for 45 different SaaS licences. Half were used by only one person, and many had overlapping functions. This created "technical debt": every time a software was updated, there was a risk of breaking the manual connections employees had built in Excel.
The data audit was the most critical point. The company had plenty of data — but no data integrity. A customer like "Jensen Ltd." could be registered with three different addresses and two contact persons across the CRM and the finance system. If you asked an AI to predict Jensen's next purchase, the answer would be worthless because the input was flawed. We interviewed 15 key employees. The fear of AI was palpable — but the frustration with manual processes was even greater. People felt like data entry clerks rather than advisors.
Phase 2: Designing the AI roadmap. With the diagnosis in place, we designed a two-year strategy. The vision was clear: the company was to be transformed from a reactive trading company to a predictive technology firm. The strategy was built around the concept of "The Single Source of Truth." Instead of all systems talking to each other in a confusing network — spaghetti integration — all systems would feed into one central hub.
We deliberately chose a "Model-Agnostic" approach. This meant we didn't tie ourselves to one AI provider. We built an infrastructure where we could swap the AI model as technology evolved without having to change the core business logic. It's a principle that sounds technical — but it's fundamentally strategic: the AI model that's best today isn't necessarily the best in 18 months.
Phase 3: Cleaning before intelligence. The first six months were entirely about "digital hygiene." This is the phase many companies skip in their eagerness to play with the latest AI tools — but it's precisely where the battle is won or lost. We cancelled 15 redundant systems and used the savings of approximately 400,000 DKK annually to upgrade core systems to versions with open APIs. This is a cornerstone of modern digital strategy: if your system doesn't have an API, it's a dead end for your growth.
We used AI agents to sift through ten years of history and clean the customer database. The AI identified duplicates, corrected addresses and categorised customers based on behaviour rather than just zip codes. It was tedious work. But it was precisely the work that created the foundation the future intelligence would live on.
Phase 4: Implementing the AI engine. Once the foundation was clean, we implemented three specific AI layers. The first was an internal knowledge layer — an AI assistant that had "read" all the company's technical manuals, quote history and internal procedures. Previously, it took a new employee six months to become self-sufficient. Now they could ask their "Second Brain" and get precise answers on complex industrial plants in seconds.
The second layer was predictive sales. By linking historical purchase data with external market indicators — raw material prices, exchange rates, seasonality — we built a model that could tell salespeople which ten customers were most likely to order next week. This shifted the sales department from "calling in the dark" to having strategic conversations with clear context.
The third layer was autonomous operations. We automated the entire workflow from order to delivery: when an order came in, AI automatically checked stock status, reserved freight and generated a personalised order confirmation — all without human intervention unless an anomaly occurred. The system only escalated to a human when there was a genuine reason.
Phase 5: The human transformation. A digital strategy is incomplete without addressing culture. We introduced "AI Ambassadors" in every department — not as technical experts, but as user voices: people who had tried AI on concrete tasks and could share the experience with colleagues. We gave employees a say: "What part of your job do you hate most? Let's see if AI can do it for you." That changed the narrative. AI was no longer a threat. It was a personal assistant that removed operational noise.
After 18 months, the company was transformed. Revenue per employee up 25%. Manual errors in order management reduced by 95%. And when the supply chain experienced a shock in autumn 2025, the company adjusted prices and inventory in 24 hours — while competitors spent weeks analysing the situation. That's not a coincidental technology advantage. It's the consequence of having built the foundation correctly before activating the intelligence.