Annual workforce planning assumes history predicts the future. Last year's project took five people, so next year's takes five. AI breaks this: the same work might now need three people, or two, or completely different skills.

When historic assumptions fail, you need scenario simulation. But simulation needs rules. How do people get matched to work? Which skills for which projects? In most organisations that logic lives in people's heads, not in systems, and you can't simulate what you haven't codified.

Organisations are running AI experiments everywhere: recruiting pilots, project knowledge bases, AI SDRs, performance analytics. Each delivers local value. But experiments in isolation don't build organisational intelligence. Without shared data underneath them, pilots fragment rather than compound.

The recruiting AI doesn't know what the learning AI discovered. Performance insights don't inform talent decisions. Real organisational intelligence needs experiments to feed a common base of shared identity, connected data and coordinated findings. Most companies confuse activity, meaning lots of AI pilots, with progress, meaning intelligence that compounds.

Most HR tech platform vendors promise a path to the skills-based organisation. People leaders want it too: better mobility, smarter allocation, lower hiring costs. But most organisations can't answer basic questions. Where does skills data live? Who owns the taxonomy? How do you keep it current? What counts as a "skill"? These aren't implementation details. They're prerequisites.

Until organisations do the groundwork for skills data, not just skills tags, the skills-based organisation stays theoretical. The platforms work, and so do the ontologies. The problem is that companies aren't ready for what the platforms ask of them, including employees owning their own data. The question isn't whether to pursue it, but whether you're prepared to do the groundwork.

Leaders like to talk about outcome orientation, but their data architecture still reflects functional silos. When recruiting, learning and performance systems don't share basic data standards, outcome-based teams can't operate. The blocker isn't org structure. It's whether systems can answer simple questions. Are we talking about the same person? Is this data current? Can we trust what we're seeing?

Most organisations discover their data makes outcome orientation impossible. Systems use different IDs for the same people. Data is weeks or months old. The same skill shows three different assessments. The future belongs to enterprises with less software and better data underneath it.

Getting AI to work isn't an IT problem, an HR problem or a strategy problem. It's all three at once. Yet most organisations give ownership to one function and leave gaps nobody addresses. IT focuses on infrastructure and software. HR focuses on people. Strategy focuses on business outcomes.

The work in between, including organisational memory, codifying decisions and redesigning work, falls through the cracks. That isn't a failure of any one function. It's a coordination challenge that needs a new organisational muscle. The next decade belongs to organisations that learn to bridge these silos and treat AI deployment as a shared, interdisciplinary capability.