moderator  0:00  
Welcome to the scrum.org community Podcast, the podcast from the home of Scrum. In this podcast, agile experts, including professional scrum trainers and other industry thought leaders, share their stories and experiences. We also explore hot topics in our space with thought provoking, challenging, energetic discussions. We hope you enjoy this episode.

Eric Naiburg  0:25  
Hi and welcome today's podcast. My name is Eric naberg, and I'm the Chief Operating Officer here@scrum.org and I'm joined by Darryl Fernandez, who's an executive advisor, and he'll introduce himself in a moment. What we're doing today is we're talking about some questions that we received during our webinar, managing your AI teammate, turning AI from experiment to strategic partner. And throughout the webinar, we received a whole lot of questions, and we weren't able to get to all of them, so we thought, hey, why not? Let's talk about them here. Darryl, you want to introduce yourself real quick. Sure appreciate it.

Darrell Fernandes  1:03  
Eric, looking forward to this. Darryl Fernandez been around technology since the late 80s. Recently jumped in here with scrum.org to look at some AI capabilities and where AI could play in not only the role of the scrum team, but also in product development and in within scrum.org itself. So really looking forward to talking a little bit more about AI as a teammate and how we might think about AI as we go forward.

Eric Naiburg  1:34  
Great. Thanks, Darrell, and welcome back. Thank you again, Darryl for joining me today and we talk about some of the questions we received from our most recent webinar around AI, and hopefully some some answers to those questions as well. These are things we weren't able to get to into the webinar as we were talking about hiring AI as a member of the scrum team. Interesting topic, interesting discussion can always happen. And you know, it's probably the biggest thing you see in the news when it comes to AI today and and fear in executives. And you know, you're an executive, and you've been an executive to some pretty large companies as well. It's ethics, security and risk, right? And of course, that governance plays into that and other things. How do we balance, really, the this whole idea, right? We want to use this. We want to leverage a new team member, and this team member happens to be artificial intelligence and lives inside and out of the organization, just like we do as employees and as people. But how do I balance that risk of ethics and security?

Darrell Fernandes  2:54  
I think the first stop is to take a look back to what you've learned over the last 10 or 15 years in your cloud journey, right? There's so many parallels here between the the cloud journey and how we've progressed through the cloud journey, what we've learned about moving data into the cloud, what we've moved about learned about moving code into the cloud, what we learned about using third party software in the Cloud, all of those lessons should be brought together here before you even embark on how to use AI, because they're all relevant their AI is a cloud offer. AI is going to have access to a lot of the same things that depending on what role you're asking AI to fill on the team, a lot of the same things that a cloud solution would so you really want to start there. So the first thing to just protect is think through your cloud journey. Think through why you made, the decisions you made for your particular industry, for your particular organization, enterprise, and how you went to cloud. If you went to cloud, what were the constraints that you applied there and and I would, I would venture that 90% of those are going to apply to AI as well. So, so how you think about data privacy, how you think about securing your connections, how you think about ensuring third party kind of edge protections around the AI capability are all very similar. I don't think you want to bring AI into the team without security clearance. That's appropriate for the team, much like you wouldn't hire somebody who hadn't gone through the right background checks onto the team. It's very similar to that process, obviously more more stringent than bringing an individual in, because you have to think about seven by 24 protection, if you will, around that edge. And you can't rely on the AI engine to apply discretion like you can an employee. You. Um, so, so I think that's where you start sharing development data. I think if it's true, non proprietary software, I think that's that's fairly, you know, benign, if you're working on how to capture customer demographics, I don't think there's too many companies that capture customer different demographics, that differentiates them in a unique way. I think that's a fairly standard process, and it's fairly standard practice. Think there's other things, like how you may model future projections in healthcare and financial services and some of those organizations, types of enterprises. I think those have more potential proprietary structures to them, you might want to protect those a little differently. You might want to think about how you leverage AI in those areas. If you're thinking about business strategy and analysis and white space evaluation and marketing market evaluations, those may be more more interesting to protect from a from a AI integration perspective. So you definitely would like, I think, to have an enterprise agreement that has the right enterprise protections around it, so that that you have structure and assurance that your AI implementation is secure in the way that you expect it to be. Don't be afraid. Every organization that's worried about these things has a compliance officer. Talk to your compliance officer. Talk to your CISO. Have those conversations that may feel like it's a roadblock in the short term. I get it, but it's going to give you much more success in the long term, if you can do that, and you can use AI effectively without worrying that every query has to be thought through to make sure you're you're holding it to a place that's so generic it wouldn't be identifiable. Are you really getting the best answers if you have to do that right? So I think those are the techniques that I would think about in this space. Interestingly, much like we've talked about in prior episodes here, ask AI, ask AI what the risks are. Ask, ask your model what the biggest concerns are from a privacy perspective, and ask it if you've thought through something, list what you've thought through and ask it what you've missed. That's an amazing use case for AI is to build those edge cases you can you can think through the happy path. Plus, AI is great at digging deeper and seeing beyond what we traditionally see, and pushing the envelope of the edge cases to make sure we're being thorough with those kind of things.

Eric Naiburg  7:39  
So I think one of the things we need to consider then too, is what AI team member do we allow on the team or not? And kind of what I mean by that is, I think I was on a call today in a meeting where somebody just turned on the Zoom AI tool to record the meeting for them so they could transcribe it into notes. And I forget what they call it, but they're note taker, and my first instinct is that's going out into some generic zoom AI. It's now got my data. And if this meeting has personal information, or corporate strategy information, whatever in it, it may now be exposed, where somebody else can ask that tool a question and get answers about something I never wanted exposed. So do I need to lock my doors, right? You know, I worked, I worked at a company where we had a data center and I had to badge in. Do I need to be able to badge in. So does that employee need to be able to badge in so that we can ensure that our information is secure? AKA, if it's a generic tool, it's out you have to use our note taker that's built into the tool that we have in order to take notes in that meeting that

Darrell Fernandes  9:01  
we know is secure in our enterprises, exactly. Yeah, and I think that's a that's a great use case. And I think a lot of us are almost ignorance, not the right word. Is too strong a word, but we're benign, like we're we're passive about it, because we're so used to it already. You know, whether it's Gemini in a Google meet, whether it's the Zoom tool, like they all have them, we're very passive about, oh, somebody turned it on. They're taking notes, but, but if that's on in a meeting that we're in, we really should be thoughtful about the kinds of things we're discussing and whether they're proprietary or not, and whether we should be comfortable. Would we be comfortable having that same conversation loudly in the middle of a public restaurant in New York City, right where we don't know who's around us and if we are great, because that's essentially you're broadcasting that conversation, potentially through AI, in a really, in a really broad communication. Yeah. Right, assume

Eric Naiburg  10:00  
that, well, exactly. And I think right now, we're, I don't know, maybe naive is the right answer, or the rate the right thing, that we're that we are just naive. Maybe it's, we're not thinking about it holistically. It's just, oh well, you know, someone's just taking notes, and it's much more than that, because that's getting fed into a model that nothing said that they can't spit what it learned out to somebody else who has nothing to do with the conversation. It's, it's that, that child, you know, I think back to when my my kids, were young, and they're, they're sitting listening to us and grandma and grandpa talking, next thing you know, there's that you're you're talking to somebody else, and they're spewing something that was not really meant to leave that room.

Unknown Speaker  10:51  
Never happened to me. Eric,

Eric Naiburg  10:55  
I think this just this is a huge risk that we that we have to be prepared for well, and

Darrell Fernandes  11:00  
that that information is also going to be used the next time that tool is transcribing somebody else's conversation. So so somebody else, you've described the white space that you're going to go explore, and somebody else comes in and says, what are some interesting white spaces to explore in this space that's now part of the lexicon, that's part of the model. It may be a very small part of the model, but it's part of the model, and so it may actually be leading others down a similar path, right, right or wrong, but it's just a really interesting space. So I think you do have to be really thoughtful about are you too passive as you enter some of those capabilities.

Eric Naiburg  11:38  
So how do we start to expose the sources? Do we need to expose the sources that AI is learning from so that we can determine do, do we have a data leak? Are we getting bad data? You know, maybe, maybe you and I were just on on a conversation, and we had it turned on, and we were just really spitballing crazy, crazy ideas, which happens? Next thing you know, I don't know, two of our co workers are in a meeting or asking AI something. It starts taking those crazy ideas and saying, thinking these are good ideas, not knowing that we were just spitballing, right?

Darrell Fernandes  12:18  
I think asking AI to source a response is always a good thing. And I think I talked about it in a prior episode, arguing with AI about certain metrics, right? So, you know, asking AI why it came up with a response, how it came what data it used to come up with a response, validate the response, show me the work, if you will. It will. I mean, that's the that's the beauty is it will, and it will be able to source how it derived the response it it gave you, to a degree, but it's really important if, if you need it, that you are sourcing it. And I think, I think you have to do that to avoid the this, this, and I've heard it a couple Well, well, I got that from Ai, right? I asked AI, and that's what it told me. Like, AI can suggest to us, AI should not tell us, right? Like, so AI can lead us to do some more research in an area, AI can indicate a path to explore. I don't think we should. I don't think we're at a point where AI should tell us an answer. I think AI should expose opportunity for us. It should clarify things for us. But I think we have to be very careful that well, AI told me to AI told me to end the call with Eric, so I hung up. I mean,

Eric Naiburg  13:43  
yeah, at the same time, it gives us an opportunity to think more broadly, or to think outside of our bubble as well. I actually used these Gemini today or something, where we were having a discussion about whether or not people from the outside would understand terminology and language we would use, and it was all people internally to scrum.org, that were talking and my concern was the conversation we're having and the terms we're using, we get it because we're in the middle of it, The person who's on the outside looking in, are they really going to understand the difference or the subtlety? And actually, and Gemini use this term as well, the subtlety of the difference. And so I asked the question, and Gemini said, Well, what you're saying is not wrong, but the subtle difference could be huge to somebody who doesn't understand that, right? So now I'm able to take that. And I'm not saying, Oh, well, Gemini said. So what I did say, you know, I asked AI about this, and here's some of the feedback that we got from it. How do we. How do we think about it in this way?

Darrell Fernandes  15:03  
Well, and I think it's really important that that's how we use AI in this in this iteration of AI, and how we're bringing AI into our teams, much like if you brought a junior person into your team and said, Hey, should we go build this $4 million product? And they said, Yeah, that looks interesting. We wouldn't just start the project. We do other validation. We do other other exploration. We understand the opportunity. And I think you have to really be thoughtful about the responses you get from Ai, not only from this perspective, but from the the potential for bias, the potential for hallucinations. You've got to be thoughtful about AI. You got to poke at it a little bit to ensure that those other underlying challenges that we know still exist in the models aren't present in the interaction you're having at the moment, in the explorations that you're doing.

Eric Naiburg  16:03  
Well, yeah, and I think, yeah, we've got to constantly be questioning ourselves as well, and questioning how we're using it and what we're thinking whenever we talk about risk and we talk about security, the word ethics, kind of pops in and pops in and out, and we talk, we always talk a lot about the ethical use of AI, from the standpoint of, you know, pushing it too far and things like that. But there's, there's another one that keeps coming up, and it's funny, I think you and I were actually talking about this the other day, but that the power usage in the extreme power usage that's being used when we do AI and over, over the holidays this year, I saw lots and lots of people generating these AI holiday card pictures of people in these skin tight suits and so on. And you and I were talking about, how many trees did we kill, or how many, you know, how much oil did we have to drill just to generate all of that? So where do we need to consider this? Is it even? Are you seeing people considering this, especially because AI is so easily available today. You know, I don't think about, I know my wife, when she was playing around with those pictures, wasn't thinking about, Oh, how much did that cost somebody or what, what impact did that have? How do we start to think about that? How do we start to get that impression in our organizations.

Darrell Fernandes  17:41  
I think this is a really interesting space that has yet to kind of come to life. I worry about it as a as an ex technology leader, much as I worried about cloud. Now I'll go back to the cloud analogy, which we talked about the prior episode, or earlier in this episode, actually the cloud when we put bad code on optimized code in the cloud, it got pretty expensive, pretty fast. The piece about the cloud was the bill came immediately with AI. There's still a lot of this free usage model. You sign up and you can it's an all you can eat buffet. And that's actually a pretty, pretty appropriate analogy right now, because it isn't all you can eat buffet, the entire whoever's got access can can go and consume as much as they want, as inefficiently as they want, and there's no additional quote, unquote, cost to that yet, right? So one of the things we're trying to push on with the with the onboarding AI into the team is to be as efficient as we can, because while it's an all you can eat buffet today, at some point it's going to be a la carte. At some point, you're going to have to pay for usage, because the energy consumption, the hardware consumption that's being generated right now is not sustainable. As it continues to grow, that cost is going to have to be borne by someone. It's not clear how the advertising model that kind of supplants the search engine model will fit into an AI model. You don't want a sponsored AI response. You don't want AI to say, well, sponsored by Company X, here's what I would tell you, No, I don't want that. I want the right answer regardless. So the model that we know can can support search may or may not fit cleanly into an AI model. So that cost structure is going to have to be born somewhere and and the consumption is huge. So I've talked about it from a cost perspective, but from an ethics perspective, it's very similar. We want to be as efficient with this because we don't want to burn more energy than we need to. So how can you set the best context? How can you ask the best questions, the less back and forth you have with the. Model, the more efficient you're being in using the CPU consumption, the data consumption, the power consumption, that all goes behind generating that answer. If you don't need a picture, don't ask for a picture, right? Pictures generate a ton of consumption. So why ask for one if you don't need one? Be thoughtful about how you interact and encourage others to be thoughtful. You know, to me, that's that's really, really important. That's one of the things that actually led to an early rev of the paper, was this consumption is going to go exponential if it hasn't already. And I don't know that our energy models can keep up with it. We've already seen pricing concerns in certain markets and in the US around this consumption driving pricing up, and what communities are, how communities are kind of dealing with that. It's, it's, it's a problem we're going to have to

Eric Naiburg  20:53  
address, and I think it's something we have to educate our teams on as part of this hiring process, because the people on our team may or may not even understand or conceive this, just like I'm sure my wife wasn't thinking creating those pictures is maybe having an impact on the globe, our team doesn't understand that, or isn't thinking that, or members of them certainly aren't. And I think as part of the onboarding process, just like when we onboard new team members, we put them through things. We also put our team through things to understand how to work with this new vet, especially if you think of it as a vendor. Often, if I hire a new team member, they're just part of the team. You don't talk to the rest of the team about here's how you're going to treat this person, but hopefully they already know how to treat that person well. But if we hire a new vendor often, we do have that vendor relationship conversation, so maybe we need to think of it more as a vendor relationship when it comes to this piece of yeah, here's how you need, how you should or shouldn't, and here's why, here's the impact this is going to have. And no, it may not cost the company any money today, but it could in the future, and ethically, as an organization, we don't want to destroy the world. And if we overuse this just for the sake of overusing it, we're having an impact on that. That's right. So I think this has been great. And thank you, everybody who's listened. Thank you again, Daryl, for joining. And I guess we always say scrum on. Maybe we say AI on, or something like that. AI is always on. It's probably listening now. So everybody, have a great day. Thank you. Thanks, Eric. You you.

Transcribed by https://otter.ai