Here is the transcript with grammar and punctuation corrections:
Good morning, good evening, good afternoon, depending on where you're signing in from. Kumar Dattatreyan here with Agile Meridian. And I am here with my good friend, Glenn Marshall, and we're gonna tackle another topic near and dear to our friends over in XSCALE Land, and that topic is AI and agile alignment. And what is that? Agile alignment and AI, you might ask. I'm gonna turn that over to Glenn. I have an opinion what it is, but I'd love to hear from you, Glenn.
Okay. Putting me on the spot.
I do that well.
You do indeed. There's a... I think there's a lot of synergy between agile and AI. And they're a natural partnership. They can go together well. The iterative nature of agile, experimenting. That's exactly what we need right now with AI being in its infancy, changing so rapidly, yet already providing significant results. So I think this is a really natural fit, and I think that this is an important topic to talk about.
Yeah. You know, in preparing for this, and actually, Glenn and I have been chatting about this either asynchronously or in person, but what we're gonna talk about in this episode. And it's a bit of a struggle because this is such a new field. Yes, AI has been around for quite a few years, but AI that we know today, the generative AI, is fairly new. It's caught a lot of people, not by surprise, but it's really upended how a lot of us do our work. And so there's not a whole lot of research out there. I mean, there are certainly some case studies of how companies are embedding AI into their workflows, mostly in manufacturing and the medical field, medical space, medical billing, things like that. And in those examples it's more, you know, how it's sort of an evolution of business intelligence where AI is being used to amplify that intelligence to allow the human, the human people that manage these processes, to make better decisions. But really what this show is meant to be is how can agile enhance the integration of AI tools into the way a team works. Right?
Can I just add something there, Kumar?
Yeah.
Another area where AI is absolutely making a difference is in personal productivity. ChatGPT will help you create emails. We were talking before the call that you can record a call, of course, with somebody's permission. AI will provide a transcript, will provide a summary. So those kinds of text-based tasks are absolutely here today, and you can save time by using them. And if you're not, you're leaving some productivity on the floor, on the table.
Absolutely. In fact, I use it all the time. When I prepare for a podcast, I generally interview the podcast guest beforehand so I get to know them and get to know what their passions are and what they wanna talk about. And with their permission, I record the call, and then I feed the transcript into an AI tool. I'm starting to use more Claude now than ChatGPT, and I ask Claude to summarize all the things that we talked about and generate some questions, a scaffolding of questions that I can use as a starting point. It doesn't mean that I'm gonna ask all those questions, but it does save me so much time where, when I'm talking to this individual that I'm gonna interview later on a podcast, on a live show, I can focus just on the person and getting to know them and not have to worry about taking copious notes because I know I have this assistant that's there, you know, an invisible assistant that's augmenting my intelligence and my ability to ask the right questions so that eventually this podcast is gonna turn out better than it could with, again, with my full attention. I don't know. How are you using it, Glenn?
I don't use it quite as much as you. I'm kind of in a prescribed role. I'm coaching at a client. So a lot of what I do is coaching. But I think you said something very powerful, Kumar. People love to talk about artificial intelligence, but really it's more about augmented intelligence.
I think so too. You know, when it comes to AI and agile alignment, the way I see it is AI is so new. It's being integrated into lots of different workflows in lots of different companies. Most of the companies that are using AI now are the big firms. And the smaller firms are not there yet. Right? And there are some software firms that are serving those smaller businesses to help them integrate AI into their workflows. But where agile comes in, and this is where, Glenn, you were talking about earlier. Right? The whole concept of agile is to take an experimental approach. Can you talk a little bit more about that? How can an agile approach enhance AI integration?
Absolutely. One of the things that I recommended to the teams that I'm coaching is that they go and set aside a portion of their capacity, obviously, in consultation with the product owner or leadership, and dedicate a percentage of their capacity to continuous improvement or technical debt paydown. I always have a qualifier. If there's something truly pressing going on, the leadership team can yank that back. But please only do that on a temporary basis and not for two iterations in a row. So you're able to carve out a piece of your capacity. I recommend at least 10%, but some teams have gone as high as 25%. And the idea here is look at the most pressing problems that you have and come up with a solution for them. An outstanding approach, in my experience, has been this notion of an experiment. It's not right or wrong. In the case of AI, you would conduct a survey of the marketplace. It doesn't have to be comprehensive. Spend a time-boxed amount of time, find the top 3 or 4, present them to the team, have them pick the one that looks most promising, then create a proof of concept, present that to the team. If it works and people feel there's value, then you incorporate it. If not, you do it again. Simple.
Yeah. The whole experimenting, experimentation approach is really important and agile teams, whether you're developing code or you're developing marketing plans or you're developing educational, you know, learning material, content. There are a lot of tools out there that are coming out that help... Speaking of software development, what is your impression of these tools and how do you think they will impact software development going forward?
I'd like to be creative in answering that. We don't have to answer that. What is your pain point today? Is it CI/CD? An awful lot of organizations spend a ridiculous amount of time going from past QA, going through production. If that's where your bottleneck is, focus on that, find exactly the spot that is a problem. And then within that spot, look for an AI tool as long as they can add value. Go for it. Trust that they will be obsolete in 6 months, but you'll have saved a significant amount of time in 6 months. And just keep doing it again. There are always areas for improvement. This is the critical thing - never be complacent. It's always improving. And with AI, it's actually accelerating. Although that was a trend even before AI came to the scene. So find your pain point and look for AI tools. There's surely something. Make sure that they will be helpful. I think right now, when you're looking at large production quality code bases, they're not quite there yet, but absolutely for writing prototype, proof of concepts, looking for certain classes of bugs and problems and optimizations, they absolutely are here. Try them on your code base and find out.
Yeah. Actually, there are some tools out there that claim to be full end-to-end lifecycle code developers that are completely AI-driven. So it was something like AlphaCode that generates code and beats their human competition, if you will, 45 percent of the time. What does that mean? I don't know what that means. It just means that they're getting better. But they're not ready to take over by any means. And I'm gonna go back on the term that I used earlier. It's not really my term. It's something that Peter Merrild came up with from XSCALE. He calls it augmented intelligence. And so if developers use tools like AlphaCode and Copilot and ChatGPT and so on to generate code or generate tests or whatever it may be, it's the human and the AI working together to develop something maybe quicker, faster, of higher quality than them doing it alone. It's sort of like when I develop some piece of content that I post online, I use the help of AI to help me generate the first framework of that content, and then I change it. So I don't have to write the whole thing from scratch. Right? Or some piece of marketing or the questions that I come up with, with the help of AI for my podcasts. What is your thought there?
I think that there's an important caveat to add to the 45%. That was 45% of developers in some contexts, not all contexts.
Yeah.
But definitely there's value that can be added. And if we're not using it, we're not as efficient as the person next door who is using it. So I think we have a professional duty to go and explore and be on top of these things. But as you were saying, it changes so fast. It's mind-boggling, but using this simple experiment model, it is a way of keeping on top of it without doing a big stop-start exercise.
Yeah. One of the things that I think we could do on this show, Glenn, is actually experimenting with some of these tools and going through a full life cycle, development life cycle, and develop something. And document all the steps that we took to develop it and the tools that we used in the process of developing it. And I don't know about you, but I haven't coded anything for probably 20 years. Right? It's been a long time since I was a developer myself. So the thought of using these tools is actually a little exciting because I've always wanted to get back into development, but I've been too busy with other projects. But this would actually help me to do that and develop something and see what is the life cycle like? And how can you use a more experimental agile approach to develop something and develop something that works. What are your thoughts?
I'm in a similar situation, and I've kind of regretted not getting my hands dirty, so to speak. So, yeah, I'm in exactly the same situation as you are. I did software development for a long time and, yeah, and I do miss it, frankly.
Yeah. There's something elegant about, you know, it's not really working with your hands. Your hands on keyboards, I suppose, but you're building something. Building something that solves a problem for someone. So, you know, we're not gonna build anything that solves any, you know, any major problem. We're just building something to demonstrate or to ourselves how these tools work. There's also another thing going on that we need to be careful of. It seems like the focus is on just making it better, but there's a fundamental problem. You know, the cute phrase for it is "hallucination." The tools just make up stuff, and that's clearly a priority one bug. But the various vendors are so intent on improving, getting faster, that they're leaving that one on the table for now. That's kind of shocking to me. Surely, you could get AI to check itself, but they haven't done that yet. So that shows how new they are. And also, it shows how you have to be very careful when you're using AI. You have to check it. As Kumar says, augmented intelligence, not replacing human intelligence, certainly not yet.
Absolutely. So, you know, if I was to give advice to anyone, I think the first thing is, as you mentioned, that iterative approach, that experimental approach. Pick something small, do some prototypes, figure out what tools you can incorporate into your workflow, whatever the workflow is, whether you're developing software or you're developing marketing or you're using it to enhance your business in some way, and gradually introduce these tools into your workflow. Right? And, you know, if it is software, CI/CD would be a great way to incorporate AI to help that process. There are lots of tools out there that incorporate automated code reviews, automated testing, deployments, and things like that. They've been around for a long time and now AI is there to enhance them. I'm gonna stop there and see if you have any comments on any of those.
One thing we do have to be careful of is confidentiality. And, particularly, if you're working for a large organization, you bring in a tool like this, it will, and I'm noticing this with one of the clients I work with, has to go through security, has to go through legal, and there's some significant potential exposure if you're putting in your production code base. So you do need to be careful there and get some approvals, in a nutshell.
Yeah. Absolutely. Yeah. With any tool, right, you'd have to get those approvals ahead of time. The other things that agile is very good with, out of the box, if you will, is this notion of collaboration and feedback. Right? The whole notion of agile teams is that the teams themselves are collaborating with each other, the members of the team, every day, all the time, and with their stakeholders, with their product counterparts. And they're getting feedback all the time because they're delivering something of value that the stakeholders can review. And AI just enhances that ability to deliver something of value that can be reviewed and feedback can be obtained from whatever was produced. Any thoughts there?
The... another critical notion in agile is the notion of teamwork. Well, think of AI as another member of the team. Simple. You know, they're not right, they're not wrong. They have skills. They have strengths and weaknesses like everybody. Just incorporate them into the team. I do wanna go back to the experiment point, very briefly. There's no such thing as a failed experiment assuming you show up and you're conscientious about, you know, trying to gather data and assess what's going on. There's no such thing as a failed experiment. An experiment that doesn't give you the result that you want is simply an opportunity, rather, an option that you've eliminated. Don't do that again. Do something else. So you can't fail an experiment.
There is a notion of a failed experiment, and that's one that you don't do.
That would be...
I would agree. Or you start and you don't finish. That's a failure. Right?
Other... no other things that can be really valuable when incorporating AI into your workflows is this notion in agile, if not just agile. You know, if you're using whatever method you're using to develop products, hopefully you're using a more user-centric, customer-centric viewpoint to develop products that people really want and need. And AI can help enhance a user experience. You know, just like I use AI all the time to enhance what I do, you know, with the things that I... people that I work with. AI assistants can help user experience professionals and developers and testers and so on to enhance what they do. Thoughts there?
I think there's a lot of potential for AI to help with testing, generating test cases, analyzing the code, looking for certain common patterns, anomalous patterns. I definitely think that's gonna be an area where AI is quite strong fairly quickly.
Yeah. And then finally, you know, the hallmark of agile, agile teams really is their ability to be flexible and adaptable and be able to respond to changes in requirements and their ability to deliver on those changes, right? To be able to absorb the changes that come with increased knowledge of what users really want. And this flexibility is perfectly paired, really goes perfectly with AI tools because this is all about experimenting with these tools and how they work and how they can enhance the productivity of the teams and enhance the quality of the products that they produce. And so, really, AI tools and the incorporation of these tools into the workflow of a team, regardless of what they produce, and agile, they really go hand in hand. What are your thoughts?
A small spin on that. Agile is about responding to changes in the marketplace. Well, here, we have changes in the tooling. Apply agile to changes in the tooling. Apply agile to your software development process itself. Apply continuous improvement to your process itself. The agile of being agile.
Yeah. I love it. Well, I don't know what else or how we can talk about this. Probably for the next 10 or 15 minutes, but we are pushing 20 minutes already on our podcast. And so we're gonna end it here. We will be back in a future podcast to show the results of our experiments that we're gonna commit to doing, learning how these tools work and actually using them to create something. We're gonna try to do that with AI and see how powerful these tools are and what our workflow is in producing it.
Yep.
Any parting thoughts before we leave?
Yeah. It won't be trivial, but it won't be getting a earning a PhD either.
Exactly. Alright. With that, thanks for watching, and stay tuned for that episode when it comes out. Bye bye.
