AI Foundations
Modern AI did not arise in isolation. It emerged within organisations that had already developed a long-standing imbalance: a preference for certainty, narrative, and communicative fluency over the slower, more demanding work of engineering clarity. The same forces that shaped software development in the 1970s quietly shaped the conceptual foundations of AI.
The Research–Industry Gap
AI research was exploratory, probabilistic, and often ambiguous. Industry, by contrast, wanted roadmaps, guarantees, and “AI strategies”. Into this mismatch stepped a familiar intermediary class — people who could translate uncertainty into confidence. In doing so, they widened the gap between how AI works and how AI was described.
Weak Conceptual Foundations
As a result, AI inherited three structural weaknesses: over‑promising, under‑explaining, and misframing. These were not technical limitations of the models. They were narrative limitations of the culture surrounding them.
The AI Hand‑Waver
The Human AI Consultant emerged as the successor to the Senior Hand‑Waving Engineer. Their job was not to understand AI — it was to narrate it.
The Result
AI works astonishingly well, but its conceptual foundations remain fragile. Not because the models are weak, but because the narratives around them are.
- AI “expert systems” of the 1980s were sold with certainty but failed due to brittle logic and poor data.
- The 2010s saw AI roadmaps and “transformation programmes” proliferate faster than working models.
- Many organisations adopted AI strategies before acquiring datasets or engineering capability.
- Public understanding of AI was shaped more by consultancy narratives than by research mechanisms.