HAL Transcript: Harry Ghuman on the AI Delegation Frontier

This is the full transcript of the HAL: Who Signed Off On This? episode with Harry Ghuman, CEO of Alpha Decisions. Listen on Libsyn. The show is also on Spotify and YouTube. For the highlights, read What Is the AI Delegation Frontier? Who Gave the AI Its Authority.

The transcript follows the published episode. It is lightly edited for readability: repeated words, false starts and short acknowledgements are removed, and names that the transcription software misheard are corrected. Timestamps mark where each turn begins in the episode.

Transcript

Chris Sowa (00:00): Hello, this is Chris Sowa. I’m your host of HAL: Who Signed Off on This? This series is about governance, and today we are we’re gonna be talking to one of the leaders that has really been involved with all kinds of governance at large enterprise companies and work with large customers. I’m fortunate to have my old boss here, who’s also the CEO of Alpha Decisions, Harry Ghuman. At Oracle. He was there for over 16 years. He’s also was a senior vice president at Palo Alto Networks as well. He brings a unique perspective not only from working with large enterprise clients but also from a security perspective. Harry, thank you very much for being on the show. We’re excited to have you on board. If you could maybe just take a couple minutes and talk a little bit about yourself.

Harry Ghuman (00:52): Some of the things that you’ve seen in your journey Chris, let’s start with why it’s very exciting to be in this line of work right now. Because, as you said, we’ve been doing some sort of enterprise transformation for 30 years, and we’ve developed the frameworks and the methods that have been actually quite durable we looked at the different type of software and applications that were operation research type of things that were optimizing parts of networks and supply chains. Internet changed a lot of things substantially for everybody. And there was a promise and the companies really needed to adapt themselves then the mobile technology came and suddenly there’s a lot of power in your hands. There may be a new technology, but what the business is doing and what the capabilities that a business requires, you to figure out that what things are going to change for the business and where they are going to create the value. We would always look at a company first outside in, we would create a hypothesis and say, what is the business the company is in and where are the potential sources of value? And more than ever, we always looked at where are the bottlenecks? What is it that they are not being able to do that if we could actually solve some of those bottlenecks, there would be value. There are other very interesting principles that we learned at that time that not once you created the value, that it didn’t really mean that it is going to get realized because you can implement new capabilities, but that doesn’t mean that the value is going to walk out of, like the cost is going to walk out of that door. There are a lot of principles that we learned from them. So later on, I applied the same skills at Palo Alto Networks where the issue was slightly different because we are telling the customers in the different industries that what type of value is at risk and what is it, the different parts of the business, how would you protect them? And by protecting those businesses, where would they be able to get more value? And as the AI came along, I got like really excited because there’s a lot of people out there excited about experimentation with AI. Whether the customers are the technology providers, they’re talking about scaling back because they’re not seeing the value. I think this is where we can actually contribute a lot in the conversation because we’ve seen it before and we know how to properly set it up and where the value can be measured.

Chris Sowa (03:58): Great. And for those of you in the audience that haven’t had an opportunity to read some of Harry Ghuman’s articles, they are available on his LinkedIn profile. And Harry, maybe just give a shout out to your website where they can find this one. Okay. LinkedIn I think that this is an opportunity for people that want to dig into the into detail about decision making and cutting-edge thinking about AI. In some of your writings you argue that strategy turns into performance through decision quality, not through technology or data. So I guess the question is, where in a company do the decisions that create advantage actually live, right? And why does that matter more now that AI can take part of them?

Harry Ghuman (04:46): An enterprise is like a bundle of decisions. And every department, whether you are in marketing, you’re in finance, you are in accounting, at the end of the day, the manager is like supposed to make some decision. Every worker needs to make some decision. And so you look around yourself because nothing can move forward until you can make a decision. So decisions and capabilities, they sort of come together and then you create a business out of it. And so it’s the same conversation is actually even becoming more important now because yes, AI can give a lot of things for like research and things like that, but then used in business, it is actually capable of making a lot of decisions. And that’s where. A lot of the benefit is and a lot of the risk is.

Chris Sowa (05:46): Absolutely. One of the things you’ve described you know a CEO with five hundred pilots, forty seven vendors and no movement and operating margin. And as you mentioned, this is kind of goes back to the this question of how do you use the technology. What is your sense based on all your years of experience this time again, why aren’t people getting banged for their buck for all this work?

Harry Ghuman (06:12): Let me first clarify this example. It is a composite example. Just to keep the client confidentiality, I’ve actually taken two examples and actually joined them together but think about it, this is very common scenario. Everybody’s actually facing this. AI actually took the action, but it didn’t give itself the legitimate authority. Somebody else gave it that authority. And that’s, think, the title of the show today, that who gave it the authority. So you can actually look at that, hey, why was the agent able to do it? And maybe there were like some design criteria where it could give itself the authority or it was just a hidden authority or you thought that it wouldn’t do it. Mean this cycle is actually changing every day. So if you keep reading about the example the agents are becoming smarter and smarter in terms of like assuming more authority. And that’s where a lot of the controls really need to come in. And that we have to control.

Chris Sowa (07:28): And to that point, you know, th the guys that are implementing this technology, a lot of them know. What does it take to get them t to force things and actually do the right thing it just seems to me that a lot of people are pretending that there’s no risk. But the consequences are huge, right? How did you how do you how did you go about this at Palo Alto? Is it is a similar kind of thought process useful here?

Harry Ghuman (07:54): Things have progressed quite a bit in the whole agent and risk and everything else. Again, there’s a security company, like when you’re looking at the security, that’s one topic. But then there’s the authority as well. And so you have to actually not just think about it from a cybersecurity standpoint. A lot of it is the governance standpoint as well that what is the agent authorized to do and where are the controls? So those things have mushroomed quite a bit in the last even like several weeks in terms of the agent being able to go. Different places. So it’s part of the enterprise design as opposed to just like thinking about one part, just like you assume that yes, there are people that are agents that both have identities. Some of the identities like you actually give authority to one or the other. And once you think about all of the main business processes and make sure that you consider all the possibilities that could happen and you put effective controls around.

Chris Sowa (09:12): Harry is bringing up two I think key things. One is around the value, right? The value, you know, making sure that the projects were put in place are on the right target to begin with. And then also the level of governance and controls required to do that task safely, right? And I think that’s the part that frequently gets left out of the discussion. When people are excited by a new technology, they’re like, hey, how fast can you do this? I’m getting pressure from the board, Harry, because I’m I need to show progress in AI and I’m not doing it. Right? So similar to a threat scenario, right? Somebody wants to protect against a threat. But the threat isn’t imminent until it’s until it’s in their house. You have designed methodologies and explained to people how to quantify risk before it happens.

Harry Ghuman (10:12): Risk is everywhere, threats are everywhere. First, you have to understand what is it that you’re trying to protect. And that’s where a lot of the past thinking about in the cybersecurity has been that, these are the types of things that we want to protect the crown jewels of the company and then the employee IDs, and then you create the mechanism by which you can control that. And when you’re actually thinking about the AI, I mean, it’s everywhere, first you have to worry about the people. And now, agents can be so many more than the people in the company. And it’s sort of hard to keep track of what all of these agents are doing. And that’s what we need to do. But we still need to start from what is the business process and what are the main business processes that the client has and who in there is authorized to do what. And if you start from that side, then yes, you can figure out that what authority do the people have, what authority do the agents have, and which type of authority is it? I mean, is it like permanent authority or is it like for a few actions? And so there’s a lot of terminology that is also coming out, but the fundamental thing about looking at which process is it’s just like issue a $10 refund or create a $100,000 refund. I mean, it’s just like, at some point you have to think about what would be the next control.

Chris Sowa (12:06): If you’re talking with a CEO and there’s two different directions people kinda lead. One is they check the box. I’ve got Microsoft Copilot I rolled out to my teams, right? So they can organize their inbox. Not surprisingly when somebody tallies up the numbers at the end of the year, they didn’t really generate that much value. On the other hand, it’s a little scary to know you know your most important supply chain management process can add the most value to AI. What’s your thought process about this? Do we s do we start small and just accept that we’re not gonna get the value? Or do we, devise a process where we go after those core processes that are critical?

Harry Ghuman (12:55): Yes, and yes, because what I’d say is that companies should continue to experiment. If you don’t experiment, you cannot really progress to the more consequential decisions that you can actually apply AI to and the more consequential processes. Every company knows what their most consequential processes are, where the money comes in from and where the costs are.

Chris Sowa (13:21): You’re hitting on something really important here, that is putting the right measurement systems in place. What are the key outcomes, right? I mean, how you know, w is this any different from the way that you’ve set up management reporting systems in the past? If you were rolling out a AI set of projects?

Harry Ghuman (13:40): I think the principles are very durable. What we did, we started with the outcome that this is the outcome we are looking for. And then we divided those outcome among all of the different key people that we had. And then we drill down into how would we get to those outcome and consequently I had all the controls around that are we doing all the right things like project plans and the quality of the work that we were producing. So outcomes didn’t really happen just on their own because everybody was very clear on what outcome they’re accountable for and what is the method because you don’t want that outcome at any cost. Mean, there’s a lot of things that you have to be responsible anytime you actually look at creating an outcome, you have a lot of responsibilities as well. You have to do things ethically, you have to do things correctly. Then just that kind of system actually has been very durable for a long time. We should not forget that it is not going to be relevant in the AI age. AI is a tool.

Chris Sowa (15:06): You draw a line, and I know you’re a very quantified focused person when it comes to results. I know from experience, that you don’t just measure activity, measure financial return. With that as a backdrop, I mean, just for people that really are thinking about this for the first time. And as you’re advising a CEO and a leadership team what to measure and what would convince you that the AI program is real rather than what I’ve framed as is AI theater, right? Or measurements would you want to see?

Harry Ghuman (15:53): I think I’d still look at the business knows where is the value, where are the value levers for a business. And first we have to start from the value lever that what could we change that would actually result in value. If we can’t even imagine that part that if we change these five things and what five things would you change? Which are the constraints. Mean, you would have noticed that like every other article that I’m writing has a lot about the constraint moving. That’s a very important point of thinking that today say, if by chance I had this capability, then I could do X, Y, Z more sales or growth So you’re describing the bottle neck. That you can’t solve right now. You would attack that bottleneck. But now what would happen is if you attack that one bottleneck without thinking about the entire system, then something else will become the bottleneck. And this is what is going on in a lot of organizations that are implementing AI, why they get stuck. Because you solve something on productivity in one spot, but then the judgment is a bottleneck. Or somewhere else it gets stuck. Either you can continue to chase one bottleneck down at a time, or you can sit down and take a systematic approach and predict that what is the whole process, which would be the subsequent bottlenecks, that if I relieve this bottleneck, would there be another bottleneck? Maybe one of the bottlenecks is that you’re going to solve all these problems, but you’re already producing what the customer needs. The demand could be a bottleneck. There is no demand. There could be other materials could become a bottleneck because there is not enough raw material. You could not have enough like transportation and logistics because if you produce anything more, will it get to an economic outcome? That’s a systems thinking you should be able to plot it right in the beginning. No I we will be chasing one bottle after the other. That’s not a lot today.

Chris Sowa (18:14): The constraints framework is very helpful as way of thinking about it and how the goalposts bottlenecks shift around is something that I don’t think I’ve heard a lot of people talking about. That’s a good framework for us to use. When we’re talking about key decisions as it relates to AI. Are there some processes you I want to keep a person in the loop even when even if the machine is right maybe that is correct.

Harry Ghuman (18:41): But it’s not the one answer for everything. Because the AI can produce these decisions like thousands of times faster than a person can actually do anything about them. That if you say human in the loop and everything is going to be fine, but the agent is producing 10,000 decision in an hour. And what is the human going to do? And if the different types of agents, what is it that you can delegate to the agent? So thinking about it as a delegation frontier, then what would you delegate? And what is it absolutely you would not delegate? And if you actually think about the most consequential decisions, that’s where you want the human to be there, but it’s not the answer for everything because you would not be able to keep up with the one.

Chris Sowa (19:38): I think that there’s gonna be business decisions and maybe even moral ones based on what you’re telling me that we need to wrestle with, right? I mean, when it c comes down to machines and war and things like that, what percentage is acceptable for launching a missile that could take out, thousands of people? When would a human need to be involved? From a big war strategy perspective, that might be acceptable, but maybe not from a moral perspective.

Harry Ghuman (20:05): I would not leave it up to the end. If you worked in a company or at a bank and you’re going to commit the company to doing something extremely consequential to their survival, I would not put that in the hands of an AI. And here’s the issue, right? You have to think about the judgment. Because human have this judgment. And that they’ve developed over taking those decisions many times. And many of those judgment so far, I’ve not seen that actually can be built into the AI or you can actually trust AI to do some things. Now that creates a very interesting question. A of the business cases earlier on when in this current generation of AI and before as well, technology transformation, people think I am not going, I’m just going to let go of the junior people. I’m not going to get entry level people. And you’re in the financial services industry. And this was the thinking we will just get the entry level people. And that’s where we will cut people even like in the last week or two, there’s a lot of conversation around this. The point is that if you don’t get the new employees, the fresh employees, how are you going to get the people five years from now who have judgment? So you would want the humans to be able to do it. And it’s just, you don’t just have senior people. Mean, senior people were junior people at some point, partners were interns.

Chris Sowa (21:57): So that’s very helpful. Unfortunately we’ll have to have you back on. We get we’re getting to the end of the program, but I want to give you an opportunity close it out for us, What’s the right way to engage with you in Alpha Decisions?

Harry Ghuman (22:12): So engagement is not a constraint here. Can actually very easily go to my website, contact me. Mean, just like alpha decision.com, my email is out there and everything. But before you engage, you should actually read some of my thinking that in terms of whether it really matches your situation. Maybe 10, 20 % of the time your problem will just get solved by reading some of those things and help you with your thinking. But beyond that, everything really starts obviously with a conversation and then an assessment that what are your goals and what are you trying to do? We would actually start by building the hypothesis that if we do these few things, the value would be found in certain areas. And then you actually start building the program and work with the architecture teams and then be able to actually put it together.

Chris Sowa (23:18): Great. I want to thank you very much, Harry, for being on the program today. This is what Hal who signed who signed off on this is all about. It’s really thinking through the key decisions. And the key points that you made around systems thinking and constraints and value, you I think this has been very enlightening for our audience. And I look forward to having you back on as well. Thank you very much.

Harry Ghuman (23:42): There’s a lot more where this came from. So we’ll see you soon, Chris.

Chris Sowa (23:47): All right, thank you,

About the show

HAL: Who Signed Off On This? is a show of short conversations with the people who have to sign for AI: how they decide what to approve, where they insist a human still looks, and what they would need to see before they trusted a machine with more. HAL is the computer from 2001: A Space Odyssey. The show is about the people who make sure that never happens on their watch.

About the host

Chris Sowa is a founder of RedactSure and a former CEO of an AI company; he started it years before ChatGPT existed. He previously led AI at Accenture, served as Global VP of Strategy & Innovation at Schneider Electric, was CCO of Sovos, and spent more than a decade at Oracle, with earlier roles at SAP and IBM.