AI Adoption: From Small Magic to Scaled Magic

Do you remember the first time you used a broadband connection? Based on my readership’s demographic, the likely answer is yes. Going from a world where you had to wait for webpages to resolve pixel by pixel, to one where they loaded instantly. The result of this reduced friction? A Jevons-paradoxical increase of browsing and time spent online, in turn leading to and enabling the development of better and richer experiences, driving increased time spent online etc.

Broadband reduced the friction of browsing the web and so helped unlock its full potential.

I have noticed something similar with AI. While it is tempting to focus on large-scale or showy use-cases, the real magic seems to happen at a much smaller scale. I have seen colleagues’ eyes light up when AI obviates or accelerates one of the myriad tasks they perform as part of their day-to-day role – and this “micro empowerment” becomes the hook for further exploration and eventually scaled application of AI.

Turns out others are seeing this too; last week the President of OpenAI, Greg Brockman, told me a similar story and even gave me a name for this pattern.

I had asked him a simple question: what distinguishes organisations that successfully adopt AI at scale? My thesis: it is less about individual AI teams and their leaders, and more about promoting ownership and practical usage across the C-suite and their functions. Adoption tends to drive more adoption.

In his response, he spoke about how the impact of frontier AI is so far-reaching people need to experience it for themselves. AI theory is not enough. He gave the example of how ChatGPT’s Slack integration allows him to quickly answer operational questions – answers that previously would have taken a few minutes to look up, now answered in mere seconds.

From minutes to seconds: broadband. He even gave me a name for it, the “empowerment loop.” Because once people realise these small things are possible, and they have more time, guess what happens – they start doing bigger and more ambitious stuff. Small magic leads to scaled magic.

Reflecting on the journey we have been on at Currys, this has been one of the hallmarks of our successes. Instead of pitching wholesale change, we started with well-defined pain points that colleagues could own, such as navigating internal approvals or wrestling with sales reporting. We got them in this empowerment loop. Because the loop is not just a loop – it is a flywheel.

What starts with Slack integration can end with 80% of your codebase being written by AI agents. Not only does the empowerment loop feed upon time saved, but as people build their confidence they attempt larger tasks, which frees up even more time, expands their skills, etc.

Greg could have given a big, shiny, ambitious example – but instead his answer focused on something small, practical, and essential. Perhaps the next AI hackathon should not start with a “big hairy audacious goal” but instead focus on the “humble annoying task”. Tackling a few of those HATs does more for your adoption ambitions than you might think – and will likely do more to set the flywheel in motion than an exciting BHAG.

– Ryan

Cover image by ChatGPT.