A pattern, not a bet

What?

Three moments, a decade apart, same instinct. In 2015 I founded WatMed Media, a social media health broadcaster, because I could see where attention was moving; it became the springboard to five years as an invited health expert for Sky News and the BBC, through the COVID-19 crisis. In late November 2022, within a month of ChatGPT’s launch, I was using it as a research tool on the 370,000-word interview corpus of my MD: serious, rigorous work, while most people were still asking what it was. And in July 2026 I built this site, machine-readable in seven days, because the doctorate had shown me the potential of large language models, and their application to health communication. The weeks after went on craft and function, and working this way unlocked a creative flow state I had not expected. The same pattern ran through my industry roles: three digital platforms and a £75,000 innovation award won by competitive pitch at AstraZeneca, a 147 per cent increase in HCP engagement at Sanofi, GME:X reaching 50,000 healthcare professionals across four countries at Bayer.

So what?

Being early is not the achievement; anyone can open a new tool in its first month. The discipline is what you point it at. Each time, the test I set was work that had to stand up to scrutiny afterwards: a broadcaster’s output, an examined doctorate, a regulated industry’s content. Run the experiment before the opinion, and let the results make the argument. Done that way, early adoption is the opposite of novelty-chasing; done the other way, it is just being first to be wrong. What has surprised me most is how it compounds: the MD’s finding, that we live in a referrer society where people check, compare and increasingly ask an AI, became the reason this site exists, and the site became the working proof I now teach from.

Now what?

Two commitments. Keep giving each new wave a real job on day one, and judge it on the work, not the demo; the current experiment is this site itself, written for human and machine readers and measured against what models actually say. And keep turning private method into shared capability: I now spend real time at work teaching colleagues these tools and helping them find their own flow, because a pattern only counts as understanding when someone else can repeat it.