From fragmented data to a comprehensive data strategy

Starting position

A large government-owned company manages the planning, construction and management of national infrastructure projects worth billions of euros. Over more than three decades, extensive data assets have accumulated — project plans, budgets, personnel resources, approval processes and geospatial information.

This data formed the backbone of daily operations: 80 per cent of employees confirmed in a subsequent survey that historical data plays an important role in their daily work. Yet the way this data was handled no longer met the demands of an increasingly data-driven world.

Challenge

The organisation faced a classic pattern of insufficient data readiness:

70 per cent of data processing was manual. There was no central data source — instead, redundancies, inconsistencies and media breaks across departments. A single source of truth did not exist. Data literacy among employees was uneven, automated data flows were almost entirely absent, and awareness around data handling was inadequate.

At the same time, valuable publicly funded data sat locked in organisational silos — unreachable for data-driven analysis, AI applications or provision to external stakeholders and innovative services.

External pressure compounded the problem: software vendors promised the moon, creating expectations that were impossible to assess realistically without a sound strategy.

Objectives

Rather than betting on individual technology solutions, a holistic approach was chosen. The data strategy rested on four pillars:

1. Organise data efficiently — Data governance, clear responsibilities, elimination of redundancies and creation of a central data foundation
2. Establish the technology base — Modern data architecture built on open source, replacing manual processes
3. Build internal expertise — Data champions as multipliers in business units, DevOps competencies
4. Embed a modern data culture — Data literacy at every level, understanding the value of data and responsible handling

Approach

Phase 1: Data analysis and strategy development

The first step was listening. In 48 personal interviews with employees from every department, a comprehensive picture of the current state was established. The trust-based setting of one-on-one conversations meant that respondents spoke openly: they described pain points in daily operations, named specific bottlenecks — but also contributed their own ideas on how AI could help them in their work.

This survey was not a tick-box exercise — it became the foundation of the entire strategy. It produced not only recommendations for action, but also concrete approaches for prototypical AI applications based on actual need rather than theoretical assumptions.

In parallel, the existing data landscape was analysed: data sources, formats, flows, quality and access rights. The result was a 120-page strategy document — a complete blueprint for digital transformation with measures, timelines, role concepts and prioritised fields of action.

Phase 2: AI prototype for workforce planning

From the interviews, one use case emerged that combined high practical value with feasibility: AI-supported workforce planning.

The organisation managed numerous large-scale projects simultaneously, some running for more than ten years. Workforce planning had been conducted once a year and purely on a project basis. In a market with scarce specialists, this meant: staffing needs were recognised too late, resources could not be secured or optioned in time.

The AI prototype represented a paradigm shift — from project-based to demand-driven planning. The machine learning model analysed historical project trajectories, planning changes and resource allocations, achieving a forecast accuracy of 90 per cent.

Given the data situation — few but complex long-term projects rather than masses of consumer data points — this was the best achievable result. The prototype demonstrated clearly: forward-looking workforce planning is possible, changes in demand can be detected early, and planning can be adjusted dynamically rather than once a year.

Phase 3: Refinement and architecture

In the third phase, the data strategy was refined on the basis of insights from the AI prototype. Architecture decisions were concretised, governance structures sharpened and the roadmap for organisation-wide implementation updated.

Results

The project delivered three central outcomes:

Success factors and lessons learned

The representative employee survey proved to be the single most important success factor. The personal, trust-based setting ensured that not only problems were named, but constructive ideas were contributed. Combining needs assessment and AI ideation in one step was highly efficient.

The availability of publicly funded data was a structural advantage: the data foundation already existed — it did not need to be created from scratch, it needed to be made accessible and usable.

Budget security for implementation must be addressed early in the project — ideally before project start. In the public sector, political priority shifts and budget cuts can stall even the best strategy. We delivered a fully implementable result; implementation did not fail due to the quality of the strategy, but due to a lack of follow-on funding.

External software vendors with unrealistic promises posed a permanent challenge. Independent, vendor-neutral advisory is particularly valuable in such environments — it enables organisations to assess promises realistically and base technology decisions on facts rather than sales pitches.