Democratizing AI: How to Empower Your Organization

0:00What type of persona, what kind of values should that agent embody? Think of customer service. You have an agent and you want this to be a friendly, helpful customer service agent. You don't want it to be so helpful that it optimizes everything for your customer, right? That you suffer margin, that it gives too many and too much discount. So as technology teams are experimenting with agents, they need to ground these agents in this information, right? How do we want an employee at our company to behave and act? Welcome to the Talking AI Podcast, where we talk AI with both experts in the field and early adopters.

0:38I'm your host, Matt Page, and we're here to demystify AI for you so you can get some value from it. Let's talk some AI. I first saw today's guest on stage at a major AI conference commanding the attention of about a thousand people, and for good reason. He's one of the top voices helping enterprises turn AI hype into real business impact, And we hear a lot of people talk about that, but our guest today is actually doing that in a real way. Joining me today is Andres Wels, an internationally recognized AI leader, advisor, speaker, and he's helped Fortune 500 leaders unlock real business value from AI, accelerate the integration of AI across enterprise applications.

1:17And he's positioned SAP as an industry leader in AI. And if that wasn't enough, he's also the author of the AI Leadership Handbook, which we'll dive into later. Andres, welcome to the show. Matt, thank you so much for having me. It's awesome to be with you. Yes, this is going to be a fun one. You get a really unique perspective, I think, on AI and how enterprises are adopting it. And we're going to weave through this throughout the conversation, but you have these nine dimensions of AI adoption. But I think the key one to start with, and I think where a lot of organizations struggle with is AI literacy within the organization.

1:52How do you upskill your organization? and I've seen organizations that are doing it great. I've seen ones that are doing it very poorly. I've seen ones that are ignoring it. I've even seen the ones that are doing it great still have people that are kind of being left behind. But I'd be curious, what's your perspective there on maybe the right way to do it, the wrong way to do it? Who do you see doing it well? And are there just things that we just can't avoid? There's going to be pain in this process as we try to adopt AI. Look, I love that that's the first question of the show because I think it's so fundamental and so important that we bring our people along on this journey of more AI adoption, if you will.

2:32I mean, first of all, no business needs more AI. They need better leadership. They need better outcomes with the help of AI. But there, I think as leaders, as organizations, we have a responsibility to train our people, to show them how can you use these tools and how can you use them in a way that's different from how we have been interacting with software up until this point. So I see a number of large organizations embrace this, actually. They might use some enterprise versions of your co-pilots or assistants or chat GPTs, what have you. But in addition to making them available to their teams, they're not just, you know, dumping it on their desk and say, good luck.

3:12You have a tool tomorrow, your 10x more productive. But they also invest in training and upskilling. One of the best ways that I've found that this works is through a concept called social blended learning. That means there are different kinds of learning, different mediums, if you will, how you learn. It could be micro learnings, you know, short five minute snippets where you learn about a specific topic. It can be in cohorts learning together in an online session, in a virtual session, or even in person. And really also share the learning that you are making individually. So getting hands-on, trying out these tools, learning about what are the pros and cons, how do you use them in a way that you get good results and good output.

3:52And like that overall companies investing in this, they're seeing returns of you cannot several hours of time saving per employee because now they know how can I actually use these tools and have a jumpstart to using them. Yeah. And it's funny, like we do facilitate AI workshops for this very reason. You know, organizations want to upskill their company. Our main focus is building AI applications, but just tangentially to that, they need help getting their team up to speed, right? But the thing that I find most interesting and where we've kind of evolved, I'm curious to take on this, is as part of the training now, we're actually building.

4:28We're using the tools. Like, we're going from use case to, okay, we'll go try to build something with some of these tools. And what's cool about it is, you know, that used to be reserved to developers, engineers, things like that. But now really anybody in the organization can build something with AI, whether it's a very simple, you know, GPT or, you know, give them access to a lovable or bold and say, hey, go tell it to build something. And it's not going to be production ready, but it kind of, I see this in people's minds. It's this unlock, their eyes kind of brighten a little bit when they see this new like way of doing something.

5:04Quick break in the pod. If you're listening to this podcast, chances are you've been thinking about how to actually use AI inside your business. And that's exactly why we built the AI Opportunity Finder. It's a free tool that helps you uncover high impact, tailored AI use cases based on your business, your goals, your pain points, and your industry. No fluff, no generic use cases, just real ideas that fit your business and the ranked by ROI potential. It takes about three minutes to run and it's like having your own personal AI strategist for free. If you want to try it for free, check out the link in the show notes or go to hatchworks.com backslash AI dash opportunity dash finder.

5:43To me, that is super exciting because, you know, as much as that buzzword is overused, but it democratizes access to the technology. To your point, you no longer have to be a mathematician, statistician, trained PhD in some kind of optimization technologies and methods to use AI and to make it usable. So, you know, if you're in an AI or in an AI leadership role and your management asks you, what are we going to do with AI? How does it fit to our business strategy? And it could be one of the reasons why they're asking you, not just because they read the headlines, because they have the tools on their phone and they use them in their browser.

6:20It's a lot more accessible. So I think that helps on one hand also create this AI literacy and empower people to, you know, do self-study, self-experimentation, see what works for you. How are you using it? And I think that's exactly what we needed at this time in the tech industry, but also in any other industry, right? Helping people get more literate. How do we use these tools and get measurable value out of it? Yeah, it's funny. You get the good side of the tools, which literally I feel like every single day I become aware of five new tools, you know, but that's. It's like hell for an organization because you're like, you know, okay, well, which tools are approved?

7:02Which ones can you use? And of course, the point that triggered my thought there was people are using these regardless of if they're approved or not. I'm curious, like, is there a correct or recommended approach there in terms of how do you define the tools for the organization? Where do you allow leniency in terms of people playing on their own? I don't know. Maybe this dovetails. You talk a lot about this new role at the C-suite, the chief AI officer. Are they responsible for that? You know, what does that role manifest into in the future? Yeah. So look, it's one of the things that I write about and talk about in the AI leadership handbook, how chief AI officers should set up their role, right?

7:44It's a lot about enablement of the organization. It's a lot about understanding what is happening in the market. I mean, we're about four weeks into the new year, and it seems that every week there has been a big topic in the news, in the news cycles from artificial general intelligence to AI agents to newer, apparently cheaper, more efficient models. And there's just the last three weeks. So it gets dizzying a little bit, staying on top of all these things. But you need somebody, you need a central role in your organization that wears this hat of how do we actually incorporate AI into our business, into our processes?

8:21How do we bring our people along and educate them and enable them on how to do that? And the chief AI officer has emerged as one of those new roles that does have that remit. So multifaceted, right? On one hand, technology. What do we have? What do we need? How do we implement these things? How do we build applications? On the other hand, what are all the things that are happening in the industry that we need to be aware of? And, you know, sit through the news and show, is this relevant? Is it not? And then also bring your people along. And where I see organizations succeed in that model is if that chief AI officer is or reports to a very senior level executive, ideally to a president or a divisional president.

9:06Typically, if they are a VP reporting to the CIO or reporting to the CTO or chief data officer, they're usually more constrained. One, in terms of scope, right? The goals that are the IT goal, they're usually ROI driven, for example. Technology, it's more about what are the latest technologies we should be looking at? Not so much how do we work with the business and how can we drive measurable value. But in order to do that across the organization, ideally the chief AI officer works a lot. on eye level and also on a hierarchical eye level as a peer to the CFO, to the chief revenue officer and so on.

9:43Do you view this chief AI officer as an interim thing? And at some point in the future, AI is literally just proliferated throughout the whole world to where anywhere from your CFO to your HR leader, it's just a core competency they have? Or do you think it has staying power? Or, you know, look, I'm trying to figure out the timeline. It makes sense. Five, 10, 20, 50 years down the road. Is it still like a C-suite position or does it just kind of merge into everything? I believe it's a transitional role. And if you look at others over the course of the last 10, 15 years, right? We've had the chief data officer.

10:23We have the chief digital officer, the chief transformation officer. and with every new trend there seems to be a c-suite role and position that is tasked with leading this effort centrally and coordinating it across those different topics and aspects so i wouldn't be surprised if say over the next five years we see the hiring and relevance of chief ai officers increase again as other parts of the organization become more ai literate as technologists and business people ideally learn how to better communicate and collaborate because they've gone through so many projects in that organization right and as the organization also builds that muscle of how do we work with ai how do we incorporate it what can we use it for what shouldn't we use it for so i wouldn't be surprised if over the next couple years again the relevance of the chief ai officer decreases but i'm sure there will be a chief quantum officer or whatever the next technology with you.

11:22That's right. Yeah. And I think there is value. I think that's right. There's value in this conduit to get you to that point for the role in time. And it's funny how these things kind of go back and forth the whole, you know, I've been on this in my, with myself, like prompt engineering, be that being this cool new thing. And then it's like, oh, that's going to go away. But as I'm getting deeper into it, I'm like realizing how critical, how you structure your prompts from whether you're building software or working with AI on strategy or copy or these things, it is critical. So, you know, going through my own journey of like up and down with these different components and things.

12:00But yeah, go ahead. Yeah. See, the thing that I love about that is it's a lot more about it as humans and how we interact with the software and how well we communicate and articulate the goals that we want to pursue, then it is about this thing, this AI system doing something for us. If you put in a poorly structured prompt, you'll see it in the response. It's generic. It's not as specific as you want it to be. It's not as pithy, as punchy, as relevant. So we need to do a lot more thinking up front of what do we actually want? What is the relevant information I would give to somebody that has an understanding of language and can recite some concepts, but how do I guide this to narrow down so I get something that is really relevant and really valuable for me and the task that I want to do?

12:49I also believe that over time we'll see more low-code, no-code type tools. It'll be a lot more convenient as an average user to use these tools. Now, if you want to then dive a little deeper, build something yourself, customize something yourself, So then I think you will need to learn prompt engineering again, or it'll still be relevant for the foreseeable future as instructions of how do I want, for example, my agents to work? What are the goals that the personas, the things that it needs to adopt and how it should behave? So to your point, right? It depends where the scale tips and it also depends where you are and what you want to do then.

13:28Yeah. And you just saying that kind of triggered something in my head. I think at the end of the day, it's almost like a base principle of the ability to communicate. We've given it this moniker of prompt engineering, but that's what it is. Like you're communicating and it is very similar to interacting with a human. And I like to think of it in that way when I'm doing that. I find it to be more effective, but I'm curious, you know, we had NVIDIA's CEO, Jensen Wong, the other day say, IT will become the HR department of AI agents in the future. It's kind of a hot take. And I know a lot of IT people, especially on face value, are probably thinking, you know, what the hell are you talking about?

14:09You know, not super happy about that, but I don't know. What's your take here? Is there some validity to how he's referring to this? Any thoughts on it? So I was thinking about a similar comment towards the fall last year and published that in my newsletter, the AI memo. And I said, do companies need to have an agent resources department over time? And secondly, in this whole exploration of how do we work with AI? How do we work with these new agents that can take a goal and figure out subtasks and work on them independently and come back to you and say, hey, this is what I've done. Do you want to move forward?

14:48We're missing HR at the table. So over time in AI projects, we've added emphasis because we saw that, hey, there are biases in the data. We need to make sure that we mitigate it and reduce the biases. Last year, we saw more of the sustainability topics come in. Gen AI consumes a lot of energy and resources. How do we find ways that we mitigate that? And I believe that we will soon see and need to have HR experts in that same project team as well. Because what we do as technologists now is, in fact, redefine how work will be done in the future. Today, we have people that have roles, that have responsibilities, their tasks.

15:29It's very well defined. You know, the self-structure compensation is somewhat clear, at least within these roles. And there's an expected output. Tomorrow, we'll have agents. They will, you know, they will work on individual tasks that somebody does today. So how do you, on one hand, define how this agent also should interact? What type of persona, what kind of values should that agent embody? Think of customer service. You have an agent and you want this to be a friendly, helpful customer service agent. You don't want it to be so helpful that it optimizes everything for your customer, right? That you suffer margin, that it gives too many and too much discount.

16:13But you also don't want it to just push profits. and your customer satisfaction might suffer. Think about an agent in finance. There are various strict rules and guidelines. Think of accounting standards like IFRS that these agents will need to abide by. Code of conduct, all these kinds of things. So as technology teams are experimenting with agents, they need to ground these agents in this information, right? How do we want an employee at our company to behave and act? How do you want an agent to behave in that? Now, if you have 10 different teams working on this across the company, you might end up with 10 different interpretations of this code of conduct.

16:55And maybe three or four or five don't even include that code of conduct. So HR has figured that out, right? They've solved this problem long ago. That's why HR exists. So what can we learn? How we adopt these policies, how we want employees to behave. And I think that's a critical part where we will need to have HR at the table and where, in fact, IT and technology teams redefining what the future of work looks like. Yeah, HR leaders are rejoicing around the world right now. But, you know, you hit on something too. I think there's huge business model impact as well. Business model impact and then hit on the point of, well, we pay humans for this job.

17:38What does that look like for agents in the future? But if you think about it, you know, SaaS companies essentially exist to make white collar workers more efficient. If you get, if you boil down to like brass tacks, that's what 90 % of the SaaS products out there do. And now you have these agents that essentially, theoretically, you know, in the future could come in and just replace the worker altogether. So, and you see like major players like Salesforce, they've kind of shifted their entire strategy to agents. what they're calling a belief agent force. And then what does that do to pricing models?

18:15You have per seat kind of pricing as the standard out there, but I don't have a good answer for this. All I know is it's going to really disrupt things. And I think new players in the market that are starting with AI as a foundational component building with it have the potential to disrupt larger legacy players that, you know, you see are kind of doing the bolt-on or maybe the duct tape on of AI within their solutions. But what do you think in terms of like just SaaS in general? What happens to that in the future? I'm convinced we'll see a lot of change in tech, in that SaaS space, but also in many different industries.

18:58I really empathize with this service as a software instead of software as a service. So you build capabilities, right? You focus on the outcome that you can actually deliver for your customer. And you will eventually in the future do this with the help of your software, your clients that work on these tasks, that gather information, that do research, that analyze it, that maybe have a conversation back and forth. And a dialogue and a discussion and come out with the best result and present that either to one of your employees that signs off on it or eventually do this fully autonomously. And it's one of the things I've been thinking about quite a bit for the last few weeks is how do you actually price this?

19:42So I was fortunate to work on SAP's generative AI pricing model across all different lines of business from finance to sales, procurement, HR, different products. And it was pretty complex. Anybody that's ever worked in this pricing commercialization area knows it's not a simple topic. That's where I believe smaller players and startups will have a leg up because typically focus on one very specific problem that they solve on one niche. And you know what that problem is and you know what the outcome is that you're driving for your customer. So my hope and my expectation is that these startups are best positioned to look into outcome-based pricing models because they have a very defined buyer.

20:28It's not any C-suite. It's not a huge product portfolio that they need to look after, but something that's very narrowly defined. And then the question is, to me, you have human labor today, again, with job descriptions, with expectations, with proficiency, level skills, and a determined salary and a bonus in some cases. How does that translate to agents? Today, we pay people a salary for giving them more or less irrespective of the output. Certainly, you need to meet certain performance criteria and goals. But maybe Friday afternoons, you close up shop a little early. Monday mornings, you don't really feel it yet.

21:12Wednesday is your hump day. It's a lot to get through. Agents don't care about it, right? They're on 24-7. they should ideally have the same level of performance, of readiness, of excitement. So how do you monetize this? Because yes, there's some transactional cost, but the value that you generate on top of it is a lot higher than a couple cents or maybe a couple dollars for their infrastructure and software consumption. So that's where I think it'll get a lot more interesting. And ideally, if companies or starters are able to figure this out and the large ones as well, this should create a lot more value for organizations too because on one hand you can reduce your operational cost but at the same time generate at least as much value at least that's the intent right i mean the technology agents are still a relatively nascent kind of capability so we'll see over the course of this year i think as well it matures further we see more organizations experiment with it adopt with it adopted and get value from it too yeah and i think it's like If the audience is listening and you drive in your car, this is the key point.

22:22You got to lean into this change because you want to be helping push that change, not having the change happen to you. You'll just be a better position in your career. But you mentioned that service as software. That's the best thing I've heard all week. You win the prize for that. I love that flip of those two because this transformation, it also unlocks, I forget which report it was in. I think it maybe Sequoia Capital or something, but they talked about this similar pattern where, you know, software as a service is kind of constricted to this one area of what software does. But now you have this entire service market that it can now bleed into, especially with the advent of agents.

23:06And I think that's going to be a huge unlock for new business, you know, new potential, all kinds of new things. I think it's a really interesting perspective on that side. Yeah. And so it was a lot of, you know, to, to be created. Yeah. And I think one thing too, for the audience, you know, if you're plugged in in any way, shape or form, you've heard a ton about agents. We're talking about agents now. I will say if you want to go deeper on this, Google just put out a white paper pretty recently on agents. And it's just one of the best overviews I've seen and they break it into three kind of logical components.

23:46it's I think it's one of the best ways of thinking about it. You have three components. You effectively have the model, you have the tools and you have orchestration. The model is pretty obvious. It's like whatever LLM you're using, right? The big unlock though, is the tools that you're giving it access to and autonomy to use. So that could be literally it's using your Salesforce. Literally it's has the ability to code. It has the ability to log into your account accounts and execute tasks. That's a huge unlock. And then the orchestration layer is super interesting too. So this is essentially this ability to govern how the agent takes an information, performs some internal reasoning, and then uses that reasoning to inform its next action or decision.

24:31So in general, you have this loop that will continue essentially until an agent reaches its goal or a stopping point. And this is the big piece. So if you want to sound smart at your next dinner party or company event, talk about it this way. You're giving it, empowering it with tools. You have the model and you have this orchestration layer. And I think it's just a beautiful way of really dumbing down why agents have a ton of potential. I don't know, any thoughts on those components and kind of what that enables and how, you know, Google's talking about it there? So I think we're seeing more and more to your point that this first layer around larger language models is being commoditized.

25:15It doesn't really matter as much on one hand what provider you use. They all have similar quality, similar output. Over time, the expectation is that costs continue to draw. The part about what tools you give it access to, I think is a really important one, because as we empower these agents to do things on our behalf, to complete tasks on our behalf, on one hand, we want to make sure that it's valuable, right? That they have this to all the swath of information that our employees have access to, because the intent is that they can sit through it more quickly and they can, you know, synthesize information and analyze it and give us recommendations.

25:53Now, the other part is, you know, querying, reading information is one thing. If we expect it to write information, to, you know, commit information in that sense, if you think of code, we want to make sure, at least in the very beginning, that we trust that these agents work as intended. And so while it's nice to give it access to all your data and your financial systems and reports and your sales data, I think there's a level of trust we need to build as individuals, as users, that these things actually work as they should. And then orchestration, where we are starting with individual agents right now, certainly even departmental.

26:34So a team of agents that work on a marketing task, that create a marketing brief for a new campaign. There's one agent that analyzes your target persona and ideal customer profile. The next one looks at the pain points. The next one drafts some copy, for example, and have them work together. So from individual functional agents, we're already moving to departmental agents. And I believe that over time, we will move to interdepartmental agents. Marketing, ask finance, how much budget do I have? Can I get a little more? How much more can I spend, right? this quarter, these kinds of things to even intercompany type agents.

27:13So one agent from your customer's procurement department asks your sales agent for a quote. The sales agent again works with other agents inside your company to gather information, create the quote, send it back to your customer's procurement agent. That's where I think this orchestration comes in because you will need to have some layer of coordination, orchestration? How do these agents in my marketing department collaborate? How do they interact and interface with this finance agent? How does that team work with sales? So again, coming back full circle to how do we organize work in a company and the department that already does that today, right?