/
All articles

Organisation

AI doesn't fix flawed foundations – why automation and AI require the same structure

AI, automation and interoperability require the same orderly foundation. Language models guess when systems disagree.

By Jens Fahlén · Published · 2 min read

The analyst – the information monster with glasses at the computer

The myth that AI will clean up after us

Many hope AI will sort out messy data. That a language model will understand who the manager is, which unit applies and who should have which access, even when the systems disagree.

But a language model guesses from what it sees. It cannot check a formal decision about who the manager is. If three systems show three different managers, the answer is a guess, however good the model is.

AI doesn't fix flawed foundations. It just makes the flaws faster and harder to spot.

Three needs, one foundation

Interoperability requires systems to mean the same thing by the same concept. Automation requires clear rules and one source to trust. AI and agents need the same to give answers that can be used.

So AI is not a new need. It is the latest consumer in line. Twenty years ago it was databases, ten years ago integration platforms and APIs. The demands on the foundation have not changed.

Ontology in practice

The word ontology is often used in connection with AI. In practice it means three things this library describes: building blocks that say what things are, patterns that say how they connect, and business rules that say what applies.

One example is the rule that a manager is appointed through a manager relation and is responsible for operations, staff and budget at the same time. When that rule is in the structure, nobody – human or AI – has to guess.

No cleanup, no automation

Sorting out the foundation takes time. Processes need to be decided, sources appointed and consuming systems adapted, often HR and finance systems. It is a trial by fire.

But the work pays off three ways at once: systems can talk to each other, flows can be automated and AI gets something solid to stand on. Start with the foundation, not the tool.

Tags

  • #AI
  • #ontology
  • #automation
  • #interoperability

What did you think of this article?