In this conversation, Speaker A describes an experimental workflow where they treat ChatGPT as a virtual "council of experts". By prompting the model to adopt multiple personas—knowledge‑management specialist, learning‑and‑development professional, psychologist, organizational‑development advisor, over‑worked architect, CFO, skeptic, and others—they obtain diverse, discipline‑specific viewpoints on a problem. The dialogue mimics improv's "yes, and" rule: each persona builds on the previous input rather than contradicting it, creating a collaborative, generative brainstorming session. The speaker uses the council to explore blue‑sky ideas, test concepts before involving real stakeholders, and achieve a level of thoroughness that compensates for human blind spots. Christopher Parsons highlights the value of the skeptic persona for pushing back on enthusiasm, and both note the limitation that a single LLM cannot truly debate itself, offering only the knowledge it was trained on (often at a basic, 5th‑grade level). The approach is presented as a low‑cost, readily available tool for idea generation, perspective‑shifting, and preliminary validation.
Chapters
I kind of had the, one thing I like to do, sometimes it's amusing, is take something very literally and see where it goes. And so it's like, yeah, what if AI was my team? No, really, what if AI was literally my team? What would happen, right? And one of the things I thought of was this council of experts. Like, part of what I need in this work is advice, is just input and kind of that back and forth to chew on ideas and bounce things around. And so, yeah, I had this notion out on a run, came back and I just, you know, in a ChatGPT project, I said, you are a council of various experts, you know, containing these persona, personae. I was like, I want a knowledge management expert and I want a learning and development expert and a psychologist and an organizational development and blah, blah, blah. I think I put in like 8. And then the first query I gave this group was, hey, you're here to help me. Who else should we invite to join this council? And AI was like, well, you need, you know, blah, blah, blah, blah, blah, blah. And it's like 20-some-odd personas. It picked names for them and gave them each a little kind of role description. And it's, and so the way I use this, and it's kind of this funny thing, right? 'Cause it is, it's just ChatGPT. It isn't actually 20-something different personas. I had wanted them to kind of debate or disagree, but it can't, it can't disagree with itself. But what it does do is it puts on different hats. And so it lets me, inspect an idea from a bunch of different perspectives. And again, most of it's like ho-hum. And then every now and then you're like, oh, there's the goodie, right? That's the good stuff. And I use this all. So I have a, I have a personal ChatGPT account that I pay for, and then I have my work one, and I have one of, I have a sort of a similar council of experts in each one, slightly different, right? And I bring it, this isn't the place to bring like, what's the population of, you know, Canada. It's like, I'm trying to think about how to help project managers feel more community, you know, just bring it kind of a big wide open thing. And it, it's, it's wonderful. And I just, just go back and forth and sometimes it just digs up a nugget of something that's like, Yeah, I could have thought of that. I don't know if I would have just because human limitations.
So it'll come out from a lens, like a learning and development lens. Here's what I would do. But if I were an organizational psychologist, I would say like, this is the problem, or a community building expert or something like that.
The way it tends to work is kind of like, you know, there's that rule in improv of like, yes, and like whatever the person before you does, you never just go no, you know, you kind of roll with it. It feels a little bit like that. Each one's sort of like, I agree with what she said and I'll add, you know, kind of the organizational development lens or, you know, psychology. And there's one of the personas is the overworked architect and, you know, she generally jumps in with like, don't forget, you've got 5 minutes to prove to me that this isn't a waste of my time or I'm out, you know?
Right. The answer can't always be more. Right? Yeah.
The answer can't always be more.
You need the skeptic that's helping to push back on the enthusiasm of the other council.
100%. Yeah. Yeah. And it's, it's just been a rich, a rich resource to just kick stuff around that's very blue sky thinking, you know, generative. I mean, truly generative thinking. I use it, use it all the time.
It's kind of in the neighborhood of the simulator you were kind of trying to build. Like you can get reps with these different I know it's a little bit different, but like there's this kind of how can I run this simulation first before I go put it in front of real people to sharpen my thinking?
Totally. And it, you know, and it's sort of a way of maybe, maybe what it does is it enacts a certain thoroughness of thinking. Like I want to— well, two things, a certain thoroughness of thinking. So think about it as a KM person. Think about it as a busy architect. Think about it as a psychologist.
CFO. Right.
I'm actually, right?
Yeah.
I'm not a psychologist. I don't even know entry-level psychology, but I bet you ChatGPT does. Mm-hmm. Right. So, so it's kind of the thoroughness, but also just access to basic, you know, not amazing psychology, but like, you know, 101 level. Fine. That's better than what I've got on board right now.
Isn't the traditional guidance to write it like a 5th grade level? For example, right? Like it, and so I assume it's probably kind of in that realm with a lot of things because that's what it's basically trained on.
That, I mean, what else can it give but what it's trained on?
Yeah.
But it, but you know what, I, you know, it has read everything on the internet. I've read one one-millionth of what's on the internet. So it has a lot that it can show me.
