It’s fascinating to see a post I wrote over four years ago—before LLMs were on the scene—still resonating today. Someone recently ‘liked’ it, which got me thinking: while the AI landscape has changed dramatically, the core challenges I highlighted remains just as relevant.

Many executive teams still assume that ‘doing AI’ is as simple as flipping a switch. The rapid rise of public AI tools like ChatGPT, Claude, and DeepSearch only reinforces this illusion. Yes, these models are impressive, but they mask the foundational work required to implement AI within an organisation. AI isn’t magic—it’s built on structured problem-solving, rigorous data preparation, and iteration. The hardest part isn’t the algorithm; it’s identifying viable use cases, ensuring the right data is available, and embedding AI into workflows in a way that delivers real value.

The companies winning with AI aren’t just deploying models—they’re rethinking their entire approach to data, decision-making, and execution. That remains the real challenge.