As one of my personal objectives this year, I organised and ran three prompting workshops for colleagues, led a small working group around the topic and helped put together an AI prompt library for the team. Along the way, I listened several podcasts, read two books (see below), experimented a lot myself and completed the AI-related part of the McKinsey My Digital Toolkit programme.
AI has been part of my work for a while now. At Coats, I was involved in introducing Microsoft Copilot, including piloting it with core users across different functional areas. Over the past year at Imperial Brands, I went a little deeper into one particular aspect of it: prompting.
AI is real, it is already here for a long time and it is not going away and if we want that relationship to be useful, we need to learn how to communicate with it: That is essentially what prompting is.
It sounds almost ridiculously simple: you ask the machine something and it answers. But anyone who has spent some time with ChatGPT, Copilot or another generative AI tool knows that the quality of that conversation can vary enormously. What you ask, how you ask it, what context you provide and how you continue the conversation all matter. Learning to prompt is therefore not really about learning a collection of clever commands. It is about learning how to cooperate with AI.
AI brings enormous opportunities, but I also think it creates very real risks for us as individuals. Once you start thinking about those, it becomes difficult to completely avoid the wider questions about organisations, society and perhaps even generations. I am not planning to solve those questions, although knowing myself, I might will. Let’s see..
So over the coming weeks, I want to use two or three posts to walk around the topic: what I learned about prompting, what seems to work, what doesn’t, key risks and opportunities, and what I think all this might mean for the way we work.
Consider this the introduction.
Coming soon: learning how to talk to the robot.
The books I mentioned above:
