In the last 24 hours, I’ve discovered both a validating limitation and a powerful liberating use of LLMs. It’s not all good or all bad—sometimes it’s both. I’d like to share these stories with you.
Coaching: the human advantage
One of my clients had been using a popular LLM to do a bit of self-coaching to work out what he wanted to do next. He’d become a bit fed up with his current role after a couple of years, and was looking for change. But because he’d become part of the furntiure, he couldn’t really remember what fired him up. So he went looking for answers through an LLM. Interestingly, the LLM explored what he didn’t like about roles in order to build a profile of what he should be looking for. A fine enough approach – and one I might have taken myself as a coach. But what he came out with (and what so often comes out of LLMs) is a very generic sounding list of ‘you sound like you might find project management satisfying’.
When I was coaching him through this, and listening deeply, I noticed that actually there was an underlying lack of feeling being expressed. He would talk about his out-of-work sporting endeavours with a smile on his face, energy in his body language, and spark in his eyes. He would talk about work as ‘quite enjoyable’, or ‘quite satisfying’ with that very rational list of things we all do to justify why something is ‘fine’ or ‘ok’. The LLM couldn’t experience the energy, so came up with something generic.
As two humans interacting, we tapped into the energy and started to map out a compelling map for his next decade.
Sam 1 : 0 LLM
Refining an OKR: the LLM advantage
The second example involves using an LLM to refine an OKR for a piece of new work. OKR stands for Objective and Key Result. I often frame these simply as:
Objective: what do you want to be better or different?
Key Results: what would be the result of that?
I had some alright starter thoughts and context which I put into the LLM. Although this time I’d been reminded that, of course, LLMs don’t ‘run out’ of ideas. I’d assumed that I would put a prompt in and it would generate the single ‘best’ answer. But it could write OKRs all day if I wanted it to.
So I asked it to come up with 10 versions of the OKR that I’d written. And without breaking a sweat – because it doesn’t sweat – but it probably is environmentally damaging – sorry – it came up with ten versions.
I could then pick the best Objective and Key Results for the work based on my understanding.
Spoiler – they weren’t all from one version, neither were they from the first version.
Next time you want to come up with an improvement using an LLM – ask for loads of them and pick the best ones.
Sam 1 : 1 LLM
A fair match, played to our strengths.
PS – I wrote this myself, not with an LLM. Although I did ask the LLM to come up with 20 titles. This is the one I chose.
PPS – I have permission from the coaching client to share his story.
Sam Grayston 