Getting Things Done in the Age of AI: What to Try and What to Avoid

Todd Brown (00:05)
Hello everyone and welcome to another episode of the Change Your Game with GTD podcast. My name is Todd Brown and I'm here as always with Robert Peake.

Robert Peake (00:14)
Hey Todd

Todd Brown (00:14)
Our goal in this podcast series is to give you hopefully the benefit of our experience, both as long time practitioners of this methodology that we call getting things done or GTD, as well as giving you the, some wisdom that comes from the work that we've done with thousands of clients over the years in helping them to realize the promise of GTD, which is to get more of the right things done and less time with less stress. And I'm.

thrilled to say that we've got a guest with us today, Scott Joslin I'm going to turn over to Scott in just a minute to introduce himself. But Scott, thank you so much for being a part of this. You are a longtime friend of the firm. You've been at many of our events and you're a familiar face to some of the folks who've been fans of the podcast over the years. Anyway, I will turn it over to you and let you.

give a little bit of a background on yourself and your GTD journey, if you don't mind.

Scott (01:12)
Yeah,

no, it's great to see you guys in this format. Been practicing GTD for about 15 years now. Helped me as I was progressing through my career as a data leader in some of the world's largest media companies, which tend to all want to come together at some point now. for the last couple of years, I've been running my own data consultancy, working with

really technically smart data and technology leaders who find themselves stuck. So I help them get unstuck and it's great to be here.

Todd Brown (01:49)
Yeah, great to have you. And just to give a little bit of a frame for the work that we're going to do today or the conversation we're going to have, ⁓ loyal listener Tim has sent in a suggestion which has to do with AI. ⁓ And in particular, what Tim has asked us to consider is the question, given everything that we know about AI and its application to the work that we do, what are two things that we would recommend to people?

in terms of their interaction with AI. And what is one thing that we would recommend that people avoid when it comes to AI? So ⁓ hopefully a bit of a framework to either refocus you if you're already a bit deep into the AI world or to get you started maybe if you haven't dipped a toe in yet. So with that in mind, ⁓ Scott, can I turn over to you? We were talking a little bit earlier before we hit record about some of the things that we might suggest people try. What's top of your list?

Scott (02:49)
Yeah, I think it's worth saying upfront ⁓ because I think a lot of the foundational pieces of GTD ⁓ rests ⁓ on this whole idea of trust, a trusted system, trusted inboxes. And I think it's fair to say that a lot of the work that we can do with AI agents can't be trusted. And so given that, think

at least for me, top of mind, is always to look at AI as something that sits outside of my GTD system. It's there to help it, to assist, but it doesn't take over any of those core pieces from capture all the way down to the reviews and the work, right? It's important and I'm seeing lots of benefits from it, but it's also important to keep it outside of that core system.

Todd Brown (03:46)
I think that's really interesting because I think that's one of the questions that I'm hearing a lot from people as I make my way in the world is, can AI really tell me what to do?

at any given moment, right? Can I really trust it with my decisions about minute to minute what to do? And that's an interesting, because of the way that a lot of the AI tools are marketed, think that's, a lot of people are just thinking, well, I'm not necessary in this equation. I'll just turn to my AI and let it decide what I'm going to do next. anyway, when we come onto the things to avoid, that's maybe one of the things to consider is where to draw the lines. But in terms of,

in terms of things to try, what's on your mind?

Scott (04:32)
Yeah, so I think the first one is I'm now in the position where I can rethink capture. previously it was, comes into my head, I'm in a meeting, here's a task, the next action, I capture it, we go onto the next one. A lot of the advances in AI and some of the apps available is like, can now be walking in the park, random idea comes up.

I have a complication on my watch, it captures it, and then when I get back to my desk and I'm cleaning up that and processing that inbox, it's there. And so an example is what I call a radar list, where it's just a random thing to help ⁓ drive some of the values in friendships and relationships. And it could be something, there's one that comes to mind as a recent wedding.

that was happening in Italy, which got mentioned. So I tagged it in the inbox radar and I have my AI agent go through, help me organize that. And it actually automates that, puts that as a task in my task manager with the appropriate tag as an agenda item and with the date. So in this case, it was in June. And so when that comes up, it then shows up on my to-do list and then I'm able to then action that.

I think in addition to that though, there are other times when these radar elements start to come up. So with the increased amount of meetings that are being transcribed and being sent out, a lot of the small talk kind of at the beginning and at the end that sort of sandwich these calls, there's a rich amount of data there that potentially could fit into ⁓ that radar, which because I'd be focused on, ⁓

meeting with a client or doing some feedback, I'm not in that head space to be thinking about that. But the agent then, as it starts to do that review, in addition to pulling out some of those harder next actions or waiting fors, it actually does the same thing and puts that onto that radar list with the appropriate tags, follow-up times, right? So it's this idea of sort of opening up that aperture of like expanding how I'm thinking about capture.

Robert Peake (06:59)
I think that's great. Yeah. And with that kind of, I call it sort of pre-processing, right? Not the official hard work in GTD, although, I mean, it makes it sound bad. It's the work you have to do anyway in real life of thinking through what matters to you, doesn't, and what you're committed to. That work that we all will always still have because by the way, AI has no commitments. Because it has no intentionality.

It wants nothing to be different in the world unlike us humans. ⁓ That level of pre-processing, yeah, it sounds like it's really opened you up to kind of a whole other level of being willing to have stuff come in with the confidence that, you know, actually I can kind of pre-parse this data stream in a way that it's not gonna end up being cognitive load when it comes time to decide what to do or not. It's gonna be more of a rich, a sense of a rich field of possibility. Is that fair to say?

Scott (07:57)
Yeah, and I think it doesn't necessarily follow the mindset that I'm in in the moment. Right? It opens that up where if I'm in that space of talking with a client and we're talking about deliverables, a wedding is just noise.

Robert Peake (08:04)
Mm-hmm, right.

Right, right, yeah, yeah. So use it in a way to expand your filters, yeah.

Scott (08:18)
But three months later, it becomes a talking

point.

Todd Brown (08:22)
It's really interesting that, you know, as you're talking about it, what I'm recognizing is that in a sense, what we're saying is we're making, we're making capture, we're making capture more efficient, right? And as a result of that, we feel like we can deal with more volume, which at the end of the day, that's, that's an awful lot what GTD is about, you know, even pre AI, right? That was an awful lot what GTD was about for a lot of people was, Hey, I want to increase, I want to increase, um, you know, efficiency so that I can get through.

more of the things that I know need to get done. And this is just kind of put that whole idea on steroids. But I love the idea ⁓ of, you mentioned the complication on your watch, which brings up an app, I'm assuming some sort of a capture app on your watch, which allows you to just talk into it and say, that occurred to me and off you go. Great stuff. ⁓ What about, ⁓

With an eye toward the second thing to try, we're talking earlier about this. And this is really interesting. I mean don't want to, no spoilers here, but this was really where I think we've started to touch on the beginning of.

helping the AI to understand your perspective a little bit more about what's important to you. So let me turn it over to you, because you made, ⁓ I think you'll make a better case for this than I will.

Scott (09:49)
Yeah, so it kind of gets into the clarify stage. And the example, as I sort of look into what I've been trying to do, is around, what do I do with all of this information that is coming in at me? And so one of them is RSS feeds. And I have them coming in from all sorts of different topics, really aligned to my areas of focus.

So there'll be stuff on leadership, there'll be stuff on AI and technology and data, sports, music, all sorts of different things coming in. Big mess. And part of that diagnostic is, what am I really doing with all of this stuff rather than just collecting? And so one of them is around, you know, helping me.

sharpen my ax in terms of how I'm working with client, upping my game on the latest research and things like that. Lots of writing and creating that comes out of it. And I, like many people suffer from blank page syndrome. And what I found happening was, say in my RSS feed, there was 20 articles, that would then become 20 next actions. And those 20 next actions would include opening up the reader, scrolling through.

having a quick sort of synopsis written down of what I thought was going on and then potentially, and then it's now time for lunch and moving forward. And what it was doing is it was really helping me avoid, quite frankly, the real work. And so what I've been able to do is then create this classification model of what I think is important. So what am I, and I do a score.

And if it meets that score, it then shows up in the system. And then I'm left with that option of like, am I gonna do the real work or I'm not gonna do the real work? It's sort of removed a lot of those off ramps that would typically go with the clarifying slash organizing piece. And so again, it speaks to the increased efficiency and productivity, but it also starts to get into some of those higher horizons of like, okay, well what...

What's important? What do I want to be doing here? What's that impact? How is this going to move these bigger blocks forward rather than that sort of dopamine hit of just hitting the complete button on all these tasks of just reading an article.

Robert Peake (12:29)
The bigger work, deeper work, work around your different areas of focus. Try facilitating some of that by allowing yourself to bring in lot of different kinds of content and then give it what I would call some kind of lens. You said you actually have a proper rubric, right? You're actually kind of scoring is this useful for me or not to this area of focus, which AI can do. But at least an angle in, right? An angle in that's a meaningful angle for you.

Scott (12:54)
Mm-hmm.

Robert Peake (12:59)
against this big sea of possible content. Yeah, that's great.

Todd Brown (13:03)
And what really resonated for me as you were talking about it was that what you're primarily doing here is managing what we would call read and review information, right? That's the kind of the topic, right? So everybody's got in their lives certain things that they want to keep on top of in terms of topics, et cetera. And what we put a label on that generally in GTD and a lot of people call it sort of your read and review list or your read and review kind of inventory.

this is something that pops up, know, again, pre-AI and post-AI, this has popped up for years in the coaching work that I've done with clients, right? And they end up with, and people generally, generally over, over capture, right? They, they overestimate how much they're actually gonna read and review. And they're not ruthless about having, having said,

and maybe print it out, right? This article, okay, I'm definitely gonna read this. They're not ruthless enough about saying next week or two weeks later, well, that was a cool idea at the time, but at the moment, the game's moved on, right? I really don't need to do that anymore. And what I love about what you've described is it's a way to help people.

to manage, you know, if we agree that keeping on top of things generally, and again, I love the fact that you've talked about everything from data science to music, um if keeping on top of things generally is of interest, I love the fact that this is a way to bring some of the capabilities that come in with AI to help us to manage all that better, because I'm sure there's hunger out there, and I'm sure a lot of people listening today will go, I really wish I could manage that better.

Robert Peake (14:45)
Absolutely. Absolutely. Great stuff.

Scott (14:47)
Yeah,

and there's different classifications or lenses for the different areas of focus or what I would call different roles. And each are managed differently.

Robert Peake (14:57)
And so just having that, staking that down, right? Knowing what your areas of focus are from a GTD perspective, externalizing it in a useful way, radically, it sounds like, improves your ability to use what AI does well, which is gather a lot of content, synthesize it, bring it to you in meaningful ways, because you've defined what meaningful is to you. And it's gonna be different for each person in different job roles and different life circumstances. But if you don't have it out there,

It's not context that AI can use at all. Yeah, it's great. It's great.

Scott (15:31)
And it still sits outside the system in that sense, right?

Robert Peake (15:36)
Right, right, you're still the gatekeeper, absolutely, of what you commit to, as it should be, and back to that thing of trust and not sort of ceding your trust or your what we call now calling cognitive sovereignty over to the system ⁓ to make those kind of calls for you for sure, yeah.

Todd Brown (15:55)
With that in mind, that's probably not a bad opportunity for us to pivot to the don't, the thing that we recommend that people not do. We gave a little bit of a preview to it earlier, and it does fall into the same category here of trust. think one of the things that we found a lot is that as people are getting familiar with AI, they make a, or they don't make a distinction between

the fact that AI is hugely knowledgeable. And Robert, I love the phrase you've used, which is having AI is a bit like having a friend who's memorized the internet, right? ⁓ Which is great and, you know, very. ⁓

Robert Peake (16:38)
Finally

someone with as wide a range of interest as me, you know, like the whole internet

Todd Brown (16:46)
But so it is knowledgeable, but the problem is that AI is not wise. It doesn't have perspective. It doesn't have experience. It can't provide that kind of ⁓ human perspective that we naturally do. And the problem, I think, for an awful lot of people is that those lines are very blurry. And they believe that AI is not just knowledgeable.

but also wise. And as a result of that, they over trust it. So anyway, Scott, I'd be interested to hear your thoughts. Does that make sense to you as well?

Scott (17:22)
Yeah, yeah, yeah. mean, all of the playbooks, the skills that I'm building, I mean, they have version number, they're tracked just to see. And part of my weekly review now includes a weekly review of the key agents that I'm kind of counting on. And it's just going into making and then doing those tweaks, right? I think the other piece of it is, you know, I think trust and arguments.

kind of go hand in hand at some point, right? There's this link and I think getting into an argument with it to try to fix it is not the way to do it, is just go back into the rubrics or the classifications or that first initial prompt that you had built in there and fix it there. And don't get into an argumentative long chat with one of these things.

Robert Peake (18:15)
No, and that reminds me of ⁓ one of the things you can do to improve the environment of trust with large language models is to actually use all of the same principles you would use to improve trust with a coworker. there's this great study and hat tip to maybe loyal listeners, Samir, for first bringing it to my attention. Anthropic did this study about emotion in relation to LLMs, right? And you guys are nodding, you know this one well. ⁓

But basically when you give a model a lot of don'ts, a lot of, you know, hard and fast unexplained rules about what would be bad and focus on what's bad, ⁓ you actually get a lot worse results. And so what they found to distill it down basically when you give a positively framed target to an LLM, it produces better results. When you give it as much context, including the rationale behind why you're giving it the guard rails, ⁓

And when you're giving it sort of calm stakes and giving it permission to take the appropriate time it needs to do the job, you improve the outcomes there. So absolutely what you said, don't argue with it, get flustered, et cetera. ⁓ All the good leadership behaviours, it's almost like cultivating psychological safety with your agent, right? It actually works the way it does ⁓ in a team environment. ⁓ And what I love about this is these are all of the

practices that not only make for good leaders, but make for ⁓ good individual effectiveness, a la GTD, right? That we put out those positive, you know, desired outcomes, those project statements that are positively framed, that we gather appropriate project support to frame the why or move up the horizons to understand some of the bigger reasons for what we're doing to motivate ourselves, and that we give appropriate attention to those things. A GTD-er, you know, will do a 10-minute thing in a sort of 10-minute way.

rather than this sort of frenzied multitasking approach. And weirdly enough, agents have a kind of frenzied multitasking approach that you can access by prompting them in ways that go, me the thing and hurry up. And the results are just as devolved, just as bad as when you do that to your own brain, which I think is so fascinating.

Scott (20:26)
⁓ Yeah.

Yeah, I have found

that ⁓ setting the boundary of ⁓ what information is being leveraged has been also very helpful. So the creation of knowledge bases and artifacts and keeping it focused on that, making sure that it has an understanding of the boundaries, it knows its role, and then it can start to execute on the task.

But if there's any ⁓ fuzziness or cloud around any of those things, that's when the drift I've seen starts to happen.

Robert Peake (21:13)
I'm curious your thoughts.

Scott (21:13)
And it's a good

signal to go back in and say, it's not the AI, it's me, and go back in and seeing where I need to make that adjustments so the results are more in line with what I'm looking for.

Robert Peake (21:27)
Yeah, almost an account, you the accountability factor does not, it does not get shifted over to AI at all. Yeah. Todd, I'm curious your thoughts on trust experiences or things you've seen also with clients in terms of, you know, having trust, losing trust, war stories, just, I think it's a rich ⁓ topic. We love your thoughts.

Todd Brown (21:49)
Yeah,

I mean, just very briefly, think that overall, you look, we're at the beginning of this, right? We're hearing from clients that management is saying, go learn everything you can about AI, right? And I think ⁓ there's a little bit of a tinge of panic in that, right? And for some organizations where senior management doesn't know really what's possible, what's, you know,

where they should focus their time and energy. And so the message is, we don't want to get left behind, go learn everything, right? And so one of the things I think we've learned over the last while is that GTD provides a wonderful framework. And Scott, you've given some examples today that I think make this really clear, provides a wonderful framework for helping you to figure out how can AI help?

Right? How can it help as I make my way through my day to try to get through my day in the most friction-free way possible and achieve as much as I can? So I think it's a... ⁓

Yeah, I mean, and I'm seeing that out there in the world, out there in the world all the time. we're, we, we as an organization and Robert, you know this, we're spending a lot of our time with clients figuring out, you know, what are your, you know, what are your tool sets? What are the givens? Where's the flexibility if there is any, and what does that mean in terms of how you can, you know, for yourself and for your team and for your organization, advise people about the best way to use this new, this new technology. And this is of course happening in an environment

where everything's evolving all the time. ⁓ it's a very exciting time to be doing the work. And you also need to be, as soon as you say, right, we're going to write a white paper and that'll be the last white paper on AI and GTD, we're nowhere near that, right? The world is just changing, it's just changing too fast. So final thoughts, guys, anything, any sort of single line takeaways before we draw this to a close?

Scott (23:51)
Markdown's your friend?

Robert Peake (23:54)
I never thought I'd be coding in Markdown, you know? The last language I expected was that, yeah.

Todd Brown (24:00)
You

guys need to explain what that is for the folks that don't know.

Robert Peake (24:04)
Yes, indeed, indeed. So ⁓ plain text is obviously the thing we all know and love. And in Word, you've seen it create bullet points or bold or what have you. Markdown is basically just a way of doing that in plain text. And it's very simple. It's the kind of things you might do on a text message, like put a couple of asterisks on either side of a word to emphasize it. Or when you're making a list, maybe indent and put a little hyphen or asterisk or whatever to...

simulate that bullet point. Markdown is that kind of thing. And actually when you, for example, copy a message out of an LLM and you paste it in, it looks a little funny because it's plain text. That's what markdown is. Not hard to learn. ⁓ And definitely I pretty much sort of take notes in that format now because it just makes it that little bit easier to paste it in as the kind of preferred lingua franca of LLM.

Todd Brown (24:58)
Great stuff for those of you that might not be familiar. Great. Scott, any final, I'm sorry, Robert, any final thoughts from you then about?

Robert Peake (25:06)
You know,

for me, think it is be practical and kind of don't get in over your head on that theme of trust. It really is that, you know, as you were saying, Todd, a lot of companies saying, learn everything you can everywhere all the time. And then also not necessarily providing a rich set of tools because they aren't sort of there yet in terms of security and governance. So I would say, you know, be reasonable, be moderate and be practical about this stuff. You don't need to go off and install

you know, open claw on your machine and try and do all kinds of crazy automations if that's not your level of comfort. Nor do you need to believe that you should be doing what we're seeing in the coding world, which really is a kind of revolution within your own knowledge work domain. It's not necessarily going to map. Be practical, be present, be nice to your LLMs as we're discovering, and just try and come from a spirit of exploration, experimentation.

and enthusiasm as much as you can because that's not only makes LLMs work better, it makes us work better too to be enthusiastic about this stuff.

Todd Brown (26:14)
Great, great, ⁓ great notion to end on. So thank you all for being with us for this episode. ⁓ As always, ⁓ please like and subscribe if you'd like to hear about more of our content. ⁓ And for me, thank you very much, Scott, for being our guests today. Much appreciated, Robert. Great to see you as always. And to those of you out there, thanks again for being with us, and we'll look forward to seeing you next time. Bye for now.

‍