Royce Sin spent a decade at HSBC automating things nobody asked him to automate. He didn't ask for permission. He just did it, showed people the results, and let the time savings speak for itself. That instinct, to question why things are done a certain way and then actually do something about it, is what eventually led him into the AI space.
In this episode, Peter and Dave sit down with Royce Sin to talk about what it actually takes for AI to stick inside an organization. Spoiler: it's not about the tools.
We get into the tension between flexibility and reliability, why most people are being set up to fail with AI, and what it means to think like a manager when you're not one. Royce also shares his MIND framework, a practical way to think about AI adoption that he developed through hands-on work across enterprise and startup environments.
There's also a good conversation about the trades, no-UI as an ideal, and why the most dangerous move in transformation is knocking down fences you don't fully understand.
This week's takeaways:
- Think of AI as a new type of employee. Set it up for success the same way you'd set up your staff. Design roles and processes to match what it's actually good at.
- Not every rule is a hard rule. Before treating a constraint as a blocker, understand what's behind it. Some fences are load-bearing. Some aren't. Know the difference before you act.
- Don't just bring in AI. Know what outcome you're after. If you can't tell whether it's working, you don't have a tool problem, you have a clarity problem.
Have a thought on any of this? Reach us at feedback@definitelymaybeagile.com
New episodes released every Thursday to challenge your thinking and inspire action.
Listen and subscribe:
Peter [0:04]: Welcome to Definitely Maybe Agile, the podcast where Peter Maddison and Dave Sharrock discuss the complexities of adopting new ways of working at scale.
Royce [0:12]: Hello, my name is Royce. I started at HSBC with a bunch of us and did that for a good 10 years. Very heavily on the technology side, the planning side. I did release management, I did a lot of innovation. I was always against the grain, always looking at how to do things better, more faster, more efficient, and better adoption. So I led a lot of transformation projects when I was there, which was all the fun. Then I moved on to PayPal and then a fintech startup. I wanted to go smaller. I came from enterprise and wanted to grow smaller because it's faster, more innovation, more opportunities to do things. And most recently, last year with AI, I just had to take the opportunity and start my own business and spend a lot of time on AI. Because if I had a job, I could only spend a fraction of the time, and there would be a lot of resistance, security issues, data issues. So that's kind of the career history. My background is a lot in operations, some sales, some development, and now I'm looking at how to use AI in the trade space.
Royce's Path Through Big Tech
Dave [1:30]: Royce, you and I bumped into one another at the bank, at HSBC. And you actually did a lot of speeding things up, accelerating things through there.
Peter [1:42]: Yeah, and I love the "against the grain" because you said that very early as you were introducing yourself. You implied that the rest of the organization had no interest in improving themselves, which seems a little odd.
Royce [2:00]: I wouldn't say it's no interest, but if you've worked a job for so long, you can only see what's in front of you. It's more tunnel vision. I was only recently diagnosed with ADHD, which is another quirk, but my friends always told me I had it because the things we did were just so tedious. It was very process oriented. It's a bank, so everything has its checks and balances. And I always asked, why do you have to do it this way? "Because we always did it this way." I started digging in to understand, okay, what's the end result we're trying to get to? If it's just this, can't we just simplify? I had my hand slapped a lot, but I just did it anyway behind the scenes and told them, "Hey, I've been saving a lot of time doing this." That's kind of how you get attention. You just do it anyway. Better to beg forgiveness.
Dave [2:54]: Yeah, absolutely. Can you talk a little bit about the against-the-grain thing you're most proud of, what stood out?
Royce [3:05]: At the bank, that's probably what started me on a path of being thrown into a lot of transformation projects. I started as a co-op, an intern, and our job was essentially to update a dashboard in Excel. That was manual. We'd go talk to project managers around the world, get their updates, and put it in. I was dying from it. It was so tedious. By the end of my term, I just automated the whole thing. I wrote macros, learned macros back then, changed the process, asked people to start putting things in the shared drive instead of emailing back and forth. At the end of it, I had automated my job and the other co-op's job. They actually hired me after that. We were in a hiring freeze, post-2008, but I had done enough that they wanted to keep me on. Because of that, every project related to transformation came my way. I always broke down what we were trying to do and why we were doing things a certain way, and built from there. Most of my career was that. I was always trying to innovate in the background, asking how I could make this simpler, how I could collect more data. That led me to building a team in India to do data management and systems. The highlight was probably the innovation competitions we had at the bank. Canada, being a smaller country, won three times in a row. We beat out all the other countries. That was down to thinking differently and questioning how we did things.
Dave [5:02]: There's a natural leap from what you're describing in terms of big organizations. I always think of big organizations and spreadsheets, and it's always a surprise but not a surprise at how many critical steps in a software delivery lifecycle are still managed through a spreadsheet somewhere, buried with some poor soul collating data and producing a report from which decisions are made. As we move toward AI, these are exactly the invisible steps people forget about, which are completely crushing the ability to take advantage of AI.
Automating Tedious Work Without Permission
Royce [6:00]: Yeah. And I personally like spreadsheets, even though I automate them for the right reasons. Spreadsheets are easy to adopt. Anyone can use them. They're flexible for any use case. I've also used a lot of systems where you always have to fight the system to do your job. If you have to fight your system to do your job, that's a bad system. But it's always a balance between how do you make a product you can scale and how do you give people control and flexibility. This is where I think AI is going to be very powerful. You can have AI spreadsheets, as many as you want, as messy as you want, and AI will make sense of it. You won't lose track of things. People can work the way they want instead of being jammed into a process designed by someone who doesn't understand their day-to-day. I've been part of pushing down tools as well. It's not the greatest experience on any end. But I think AI is going to create new opportunities to allow people to do their best work, aligned to how they actually work, while also standardizing and consolidating. Translating from my brain into something Dave understands, something Peter understands, something management understands. That's where a lot of the excitement is for me.
Peter [7:08]: There's always friction between those two different worlds. You need flexibility to go fast, but at the same time you need consistency because you want economies of scale, or you need data to be reliable. You can't always rely on a non-deterministic system to interpret data and give you back the same information. I had one last week where I said, here's an image of a table, recreate this. It got the columns and rows correct and then just made up a bunch of content to go into the table. So at scale, if we need reliability, how do we manage these two worlds as they come together?
Royce [7:58]: Yeah, I think that's very true. The dream is everyone works the way they work, but there have to be some guardrails. People are still figuring that out. A lot of what I'm looking at is, okay, I want people to just send voice notes. For myself, I hate updating CRMs, I hate updating project management systems. I just want to talk and say, hey, I did this today, then have it update automatically. That's possible now. But as you said, it's not fully free form. What I input is unstructured, but there are requirements around it. You must mention which project, you must have your update, certain minimum things. There are easier templates or structure we can give people that are easier to adopt. From there, AI can infer the rest, saying, you're working on this project, it's probably this one, you mentioned this name, it's probably this person, grab the rest of the data and put in the updates.
Peter [9:00]: Yeah, that imperfect match between different pieces of information. Here's my interpretation, and here's my confidence level that I got it right.
Royce [9:09]: Exactly. And question back to both of you, how have you seen this, and how have you used it in the organizations you're working with or your own?
Dave [9:22]: I'll let you have a stab at that. So as you were describing that, one of the key things is coming at it with a structured versus unstructured data view of how to pull things through. Working with the way I actually work, so I can verbally communicate what I'm thinking, and then trying to turn that unstructured data into structured data. It's a bit like the classic problem with user stories in the sense that if it's only one-way communication, you're going to end up with structured data which we think is correct but hasn't been validated. What we're seeing more success with is that validation round. You start with unstructured data, throw it at whatever you're using to convert it to structured data, and as Peter said, it comes back and says, I know this, this, and this, but here are three questions. Can you just clarify my thinking about project A or project B? That little bit of clarification is a lot like refinement conversations in user stories. I think I know what your unstructured data translates to, but let's validate it. That double-loop learning piece is totally different from systems where you're forced to enter data in a structured way, or you enter it and hope you got to the right structure. That loop is really a critical one that is beginning to be built into the process steps we're talking about.
Why Spreadsheets Still Run Everything
Peter [10:59]: It does create an interesting piece, though. The three of us have enough of a technology background to understand that conceptually. But someone coming to this totally fresh doesn't necessarily think that way. Why would I want to do that? What's the benefit to me? One thing I've been playing with as a thought exercise is this concept around the different roles people fill in an organization. You've got technologists who do the really complicated stuff, and you've got people who are using AI to create internal websites or write the Excel macro that's actually being used to make business decisions, versus the actual system that we think is being used to make those decisions. That's the problem you were describing before. There's a risk view here as well. If someone's writing a little piece of code they're just using for their own purposes to take data out of a PDF and upload it into an Excel sheet, that's one thing. If they're sharing it with teammates so they can do the same thing, that's another. But at the point they want to give it to the whole department, from a technology perspective, we no longer want them just running that off their laptop, because if their laptop goes down, that capability disappears. There's a piece around helping them through that journey as they move into higher-risk delivery within the technology system. AI starts to allow the crossing of those boundaries in different ways, which I think becomes interesting.
Royce [12:52]: Yeah, definitely. AI has so many variables you can't fully control. If it works for one person, as soon as you add another person, it's a whole new set of variables because they use it differently. I do agree that the skill set required to effectively deploy AI solutions is different from what most people are used to, and that skill is lacking, especially in small organizations that have never seen what systems look like at scale. They're just running their business. I've encountered a lot of individuals who still think, oh, AI is a tool. As long as I bring it in and people start using it, that's a win. Sure, you can use it as a glorified spell checker and writing tool, but you're not actually going to reach the real transformation unless you understand your operations and how your organization works. And, again, what's your end goal? What are you trying to achieve?
Dave [13:58]: One of the distractions you often get in these conversations is this focus on standard operating procedures. One of the things you described well is that everybody interacting with an AI tool or product is going to create different variables.
AI Flexibility Versus Reliability And Guardrails
Dave [14:25]: The magic is not standard operating procedures. You don't need non-deterministic systems to follow them. They're standard, therefore deterministic. But we can make something much more powerful because it can respond to Royce's needs differently than Peter's, differently than Dave's, all within the constraints of what we're trying to achieve. That variation isn't about value stream mapping and SDLCs and standard operating procedures. It's about really understanding what the operations are and where the variables are that we can explore.
Royce [15:01]: Yeah, and I came up with an acronym at one of my meetups: MIND. M is everyone is a manager. Everyone needs to think like a manager. You don't manage a tool, you manage outcomes. I believe if you want to be successful with AI, you can't just tell people what to do. You have to teach them how to think like managers. This is the end goal we're all aligned to. That's normally what organizations do as people move up and become managers, directors, VPs. Everyone at those levels is less and less structured. They don't have an SOP for their jobs. They just have, okay, that's the goal we're trying to hit, I'll figure out how to get there. This is where there's a gap. You're giving lower-level employees an AI tool but not teaching them how to think like management, how to align themselves with strategy. Just telling them to use it and do their work faster. They want operating procedures because you've trained them to be order-takers. That's not how AI works best. If you actually teach them how to think like managers, how to find their own way to get results, here's the tool that helps you do that. Then there are going to be a lot more successful use cases. And it helps adapt to the variation from one person to another. As long as everyone is running toward the same direction and knows the guardrails, then go at it.
Dave [16:37]: I was about to ask about the I, N, and D.
Royce [16:41]: Right, okay. So MIND. M is everyone's a manager, that's the mindset for training people how to use AI. Second is integration. People often ask me, what tool should I use? I always advise against subscribing to 15 different tools because gurus online will tell you that's how you become an AI expert. The power of AI is having all your data in one place. People tend to silo everything. So just integrate all your data and use one ecosystem. I hardly ever deviate from Google. I use Google Workspace and try to use all of their AI tools because they only get better. I was tempted to get one of those fancy tools that summarizes your emails and does all these things. Then Google launched the same thing and it was already included. Why would I want to manage another subscription and more security overhead? I see a lot of small business owners where those who do adopt AI end up with something very complex to manage. They subscribe left and right, then a whole bunch of things get abandoned. So I is just integration. Use one ecosystem.
N is new data. This is a new way of thinking about your operations. We don't often collect all the data that's available to us. Meetings, for example. We can record them now, we have summaries. But what about in-person conversations? Do you record those? That's an opportunity to gather information too. I can put a phone on the table, turn it on, and start recording. It's just a way of thinking about how your existing process today can create new data you can use in your operations. Data sets that were never collected before because if you were doing it manually it would be too tedious, but once automated at scale they become very valuable. Call information, for example. What do people say on calls? What are they interested in? Suddenly you're seeing upsell opportunities in support calls that you never thought about. Designing your operations so they generate data has to be the new mindset.
The last one, D, is division by design. This is where people put AI on top of a human process. This is how I do it as a human, now I'm going to have AI do it. But that does a disservice to AI because now you're limiting AI to work like a human. Whereas if you look at a process and ask whether it can actually be redesigned to take advantage of AI's speed, AI's context, the fact that it can read a lot more information, then redesigning processes to work better with AI is the opportunity. These are just things I've learned over the last year and a half of diving into this, plus experience with transformation and seeing what works and what doesn't.
Peter [20:18]: I think it's a good acronym and a good perspective. I agree with the manager piece at the start. That's what agile transformation was trying to do for the last 20 years, create autonomy within teams so they could actually act and operate autonomously, which also required the systems they were working with to allow that. AI is starting to accelerate that journey. The division by design piece is one where people do struggle a lot. That mindset shift of how do I rethink the approach to this problem, especially when we started the conversation talking about "that's how we've always done things." It's doubly a problem when you now have a tool set that enables you to completely rethink everything. That puts you in a very different spot.
Dave [21:30]: It's an innovation problem. A lot of people automatically close off from that. People tell me all the time they're not creative or not good at innovation, even though they actually have loads of creative and innovative experience. They just don't think about it that way.
The MIND Framework For AI Adoption
Dave [21:45]: And to your point, the way we did it yesterday becomes the way we need to do it now, so we're going to automate that. And you've just missed the opportunity. It isn't just optimization. It's not just about automating existing processes. It's really rethinking, as Royce said, the new data. What can you learn from data that is relevant to the business, that's available, but maybe hasn't been captured or processed before. There are also more and more interesting conversations about what exactly are we experts in that AI systems currently can't cover, whether it's emotional intelligence, reading body language. There might be something there eventually, but it's not commercial right now.
Royce [22:44]: Yeah. And I do have some sympathy for people in these situations because the hardest part of transformation projects is that you still have to do your day job. You still have to keep the business running. And then you have to step back and look at the process from the outside and do something you've never done before. Working on the business instead of in it. That's not something people are normally taught. Coming in from the outside helps because you can ask and probe those questions. But then the challenge is you don't have the context, you don't fully understand what they're trying to do. So people who are in operations, who understand how things are done and why, are going to be very important. AI will supercharge them because they're the ones who, if you can pull them out of the operations and give them the tools, can focus entirely on thinking about what you're trying to achieve, what you can eliminate, and what AI can take over instead.
Dave [23:54]: Royce, can you talk a little about what you're looking for? You hinted at working in trades or closely with trades. You've been doing some exploration in AI and process automation. What are you actually seeing?
Royce [24:12]: I spent a year just exploring, talking with different types of business owners, different industries. I actually started in trades a long time ago, but the wait lists for programs were over two years so I did something else. A friend I grew up with has an HVAC company, so I'm helping him out in that space. My goal is to build a SaaS product that doesn't get killed the next time OpenAI or Google releases something new. So I'm looking for those opportunities in processes and operations, and specifically in the physical world. The digital side, any desk job, those spaces are fairly saturated. Everyone has tools. But when it comes to the real, physical world, there's still a lot of opportunity. There are still a lot of conversations not being captured, still things being missed. I'm looking at how we can use AI to simplify these processes, or give traditional spaces new ways of working they've never thought about before, that give them the ability to capture data and opens new doors to opportunities. So trades is an interesting space.
Peter [25:44]: For sure. There's a lot going on there, and it tends to be lots of small individual businesses as well. So how do you help them? They tend to not be as tech savvy in general.
Royce [25:58]: Yeah, that's a challenge. They're not tech savvy, they're fragmented, they don't like learning new things. Selling to them has always been hard. It's hard to reach them, hard to get adoption because they have to learn a new UI, build a new process, and they don't want to do that. They just want to do their work. Even before this, I've learned from people at the bank to look at how you bolt on processes so it feels invisible. You're not changing their day-to-day. If you can actually do that for a lot of these traditional spaces, where they're creating and capturing data without even knowing it, that's the best experience. The best UI is no UI at all. Whatever is already their day-to-day, that's what we should target. I think AI makes that possible now.
Dave [26:56]: How do you deal with the governance and security side of things? If I step into trades and think about no UI as the best interface, conversational, there's a ton of information that can be picked up that should be protected. How do you solve that?
AI In Trades And The No-UI Ideal
Royce [27:21]: I think there's a lot of research that needs to be done, and it depends on how you want to do it. You could say, just slap a camera on and record everything. Lots of problems with that. But I believe for certain conversations, you can record them as long as one party agrees. Don't quote me on that, it's something I read. For myself right now, I'm looking for safer opportunities and getting everyone to buy in. But a lot of those problems come at scale. When I was at the bank, I saw a lot of innovation killed by the thought of, oh, we have all of these rules, therefore I'm not going to try at all. But if you never try, you never really understand which rules are the hard ones, which are negotiable, which will change by the time you actually get there. A lot of rules are changing now. Open banking was a no-go and then suddenly everyone's open to it. I would just start anyway. There's a hundred-and-one problems to solve. Okay, we know what those security problems are. There's a whole bunch of unknowns. Go solve those ones first, then negotiate on the security ones. Because there are always ways around it. Put things on Canadian servers, get people to sign agreements. There are ways. And as soon as you start creating value for all parties involved, they may want that value enough to say, okay, I'm fine with you taking this information. That's how I'd look at it.
Peter [29:12]: There are plenty of opportunities, and it's about not letting the current constraints of the system get in the way. That reminds me of the concept of Chesterton's fence. The idea that you don't tear down something you don't fully understand. If you come across a fence in the middle of a field with no apparent reason for it being there, there might be a very good reason. Until you understand why it's there, don't knock it down just because you can. That can cause problems too.
Royce [29:57]: Yeah. Or Elon Musk removes it first and then goes, oh, I actually needed that, let me put it back.
Peter [30:02]: Well, yeah, maybe that wasn't such a good idea.
Royce [30:04]: Yeah. I think it's about risk and reward. If you feel like there's not much risk here, try it out, turn it off for a week, see what happens.
Dave [30:19]: And what you're talking about, Royce, is small scale. If you're dealing with a single provider or a smaller organization, that risk is reduced. But if you're dealing with something province-wide, that's a totally different conversation.
Royce [30:35]: Yeah, absolutely.
Peter [30:36]: Yeah, understanding risk is key.
Security, Governance, And Testing The Fences
Royce [30:40]: I guess because I'm in the startup world now, it's always about de-risking. There are a hundred-and-one problems, probably a hundred more that are unknown. If you already know the security problems, just table those for now and solve the other ones first. I've seen medical companies and other high-risk companies start in other countries because there are too many rules in the US or Canada. They figure it out there, get the tech right, get the usability right, and then start hardening it for the current market. There are a lot of ways to go around constraints. That's the beauty of innovation and startups, and also the challenge.
Dave [31:23]: Yeah, for sure. I feel like we want to head towards three points to summarize.
Peter [31:29]: Yes. So if we want to wrap this up with a tidy bow, we each get a point for our audience to take away. Royce, you go first. What point would you like our audience to take away from this conversation?
Royce [31:45]: I would say the one I find most effective is the everyone-is-a-manager mindset. Think of AI as a new type of employee. They're good at certain things and bad at others, so you should design your jobs and roles accordingly. You want to set AI up for success, just like you set up your own staff for success.
Peter [32:10]: Dave, what do you want to add?
Three Takeaways And How To Reach Us
Dave [32:13]: I was going to go with the manager one as well, but I'll find something else. I'll touch on what Peter was just mentioning at the end. We've often talked about risk, and referenced how everything comes back to risk in conversations about transformation. And again, we've just seen exactly the same thing. It isn't a case of following all the rules that are there, but really understanding what's behind them. From a risk perspective, testing some of those boundaries, seeing what you can work around or safely bypass, and knowing which fences have to stay.
Peter [32:54]: I think that's a good one too. For me, there was a lot of good stuff, a lot around innovation and how we approach different problems. The piece I'd take away is around ensuring we're not just bringing in tool sets for the sake of it. What are we trying to achieve? What is the outcome we're looking for, so we can tell whether we're actually achieving it? Because a lot of what we're seeing right now is, oh look, we got AI, great, but we're not seeing the outcomes we'd like. And it does tie back to something we've also touched on, which is how hard it is to step outside the system, look at it, and start making changes. Which of these bits do we really need and which can we easily change? It often takes someone else coming in to help you figure some of that out. Okay, with that, I'd like to thank you, Royce. Great conversation. And thank you, Dave, as always. Everyone can reach out at feedback@definitelymaybeagile.com and don't forget to hit subscribe.
Dave [34:01]: Thanks again, Royce.
Peter [34:03]: You've been listening to Definitely Maybe Agile, the podcast where your hosts Peter Maddison and Dave Sharrock focus on the art and science of digital, agile, and DevOps at scale.