I use AI a lot. Ha, that felt like a confession, which surprises me because a year ago it was still quite judgey, judgey in the LinkedIn world for comms people to admit they were using AI. There are still the odd diehard purists, and I do think drafting from scratch is a muscle worth exercising, but surely we’re past confession territory by now.

AI helps me work through information, test my thinking, pull themes out of a pile of feedback and get to a first draft faster. But I’ve been thinking more about the bits it doesn’t do particularly well, because that's where I'm either having go-grey experiences or straight pulling my hair out at the endless loops I end up having with any given LLM of the day.

The problem, particularly in communications and engagement, is that what comes back can be really plausible.

You can give AI a project brief and ask it to draft a stakeholder plan. It will identify stakeholders, risks, channels and messages and, at first glance, it can look pretty good. But does it notice that the stakeholder who was making a lot of noise six months ago has suddenly gone quiet? Does it connect an apparently minor project issue with something else happening across the organisation? Does it know that a perfectly reasonable message is going to land badly because of what happened last time?

Occasionally, it does make those connections.

A lot of my work involves joining those dots, just being in the weeds in this work for this long seems to have developed some weird sixth sense in how my brain is wired. Something that looks relatively innocuous on its own can look quite different when you know what else is going on, what’s happened previously with a stakeholder, what another part of the organisation has said, or what might be brewing in the background.

We tend to call that judgement, and the bit I’m interested in is how we actually learn it.

I went looking to see what the research says and quickly discovered that research from even a year or two ago is starting to feel old in AI-time. More recent research is looking beyond whether AI simply makes us think less. There’s work exploring when we offload thinking to AI, when we use it to support our own reasoning, and when working with it starts to change the way we think about the work itself.

That got my attention, because I can see myself doing some of this in real time.

I use AI to interrogate information, test an assumption or see what I might have missed. Then I challenge what comes back, add context and make connections it hasn't made. Quite often I find myself looking at a perfectly plausible answer thinking, hmmm, no, something about that doesn't quite add up.

If you’re experienced enough to know something is missing, AI gives you something to push against. But what happens when you don't yet have that bank of experience? If the stakeholder plan looks comprehensive, the response sounds professional and the analysis seems sensible, what tells you to keep digging?

Experience has to count, but I’m not convinced experience on its own gives us good judgement. Reflection matters too. Looking back at a decision and asking what you noticed, what you missed, what else was going on and what options you had at that point is a big part of how I approach coaching.

So I think that’s the bit I want to explore over the next few weeks. As AI takes on more of the work we use to develop our craft, how do we make sure we're also developing the judgement to know when its perfectly plausible answer isn't good enough?

I’d like to know what’s helped you develop good judgement in your work. Email me and let me know.