Building Digital Trust: The Human Side of AI with Manoj Mohan #legacybuilding #ai

Welcome to the Legacy All Stars podcast. We feature inspirational stories from people who dedicate themselves to helping others. Here is your host, Alan Pavloski.

Hello, everybody. Welcome back to the Legacy All Stars podcast, where we celebrate inspirational people making a difference in the lives of others. I'm super excited about this episode because for those of you who know me, I geek out on tech and I geek out on AI, and there are risks associated with that. So I want to make sure you all know how thoughtful this guest is. Today on Legacy All Stars, we're talking about digital trust, the human side of AI, data privacy and security. Our guest is Manoj Mohan, a veteran tech leader whose work at companies like Intuit, Meta and Apple has focused on making powerful systems safe, useful, and accountable. Manoj believes leadership starts with service, designing software with people in mind. And he's passionate about practical safeguards that protect families, customers and communities while still unlocking innovation. We'll explore the real pitfalls of enterprise AI adoption and how to build guardrails without slowing teams down. And the daily habits that keep Manoj energized for a long, healthy lifespan. Welcome to the show, Manoj. It's great to see you again.

Thank you for having me, Alan. Excited for this session and look forward to our conversation today.

Yeah, as always, we've spoke several times before about some of these subjects, so I'm really excited for our audience to hear from you.

Well, let's jump right in. You've been called a kind of digital guardian. What does that mean in practical terms? And where did that mindset start for you?

So, Alan, the term digital guardian sounds lofty, so I'll leave the term there. But I see it more from a practical angle. Everything that we build with technology is usually for automation or for some form of productivity improvement. Now, while we bring around that productivity improvement or automation, how do we keep the software product and the users safe and secure with all of the right cardrails built in? That's, in essence what I think about as the most critical component of everything I contribute in my work. So, you know, how did it all start? So for me, it started back at Apple. Apple has a ton of things that Apple does in the context of safeguarding every bit of customer privacy. So getting early inroads into that culture definitely helped me learn and educate myself. Then at Facebook, I think it was about preventing misinformation, preventing targeted clickbaits on senior citizens and things like that. So, again, super important problem statement for the community, for our society. And then at Intuit, it was about Protecting all of the super sensitive financial data, be it for companies, be it for end users like all of us. You know, for example, small business that relies on QuickBooks and let's say if they're using QuickBooks payroll, you know, they are trusting our software to deal with all of their money movement. So how can we, who build the software, who conceptualize the software, ensure that it is safe, secure and does that in a fast manner, adding value for the small business? Right? So that's how I see it. And you know, I think for me, the mindset, I think started with a simple question, right? If there is a product, and if I have to give that product to my family or to anybody in my family, would I feel safe and secure for my family members to use that product? If not, then the product not worthy of it, right? So that's the standard I have tried to carry throughout my career, throughout all of the enterprise software and platforms that we have built. So hopefully that answers the question that.

Definitely answers the question that really shows the ethics behind what your philosophy is all about. So I really appreciate that. I've been wanting to ask you this since our last conversation. What is a story from earlier in your career that shaped your values around protecting people and data? And are there any mentors that you still think about today?

Yeah, good question. So early on in my career, you know, just like every new engineer on the block, I was obsessed with being able to ship software as fast as possible, build features as fast as possible, and so on. So for the most part I would be focused on rolling out feature after feature. And once in one such instance, we were rolling out a system and something that I was not paying attention to would have potentially resulted in logging so some sensitive user centric data and that potentially could have translated for vulnerabilities in terms of breach and so on. So I didn't realize it, it was my ignorance, I was still focused on rolling out the feature quickly. And that's when a senior engineer caught me and told me, Manoj, if you're trying to move fast, and if you're trying to move so fast that you end up breaking people's trust, you are not innovating, you're not adding value, you're essentially gambling. So that I think landed hard. It was good feedback and I have kind of imbibed that since then. And that's how I've changed, to not undermine the importance and not to think about it reactively in all of the work that I have done ever since. And so for me, the way Now I think about it is like scale without trust is like speed without control. Right. It's pretty dangerous and we should never take that sort of approach. And all of the leaders that I admire have taught me that, you know, protecting people's data isn't a side task or a nice to have. It is the capability of how you can steer the car to be on the road with the right guardrails at all points in time. Right. So that's how I see it.

So listen, as a follow up question, can you give me an example of when you prioritize a user's or a teammate's well being over a short term metric? And what happened?

Okay, so there was an initiative that I was working on back at Apple where we were actually running behind schedule on that particular project. The work in itself was about implementing all the privacy checks and the guardrails for one of our critical Apple care systems that we were scheduled to launch for our users and we were behind on schedule. So the easy call at that point in time would have been to just go ahead with making the launch date and then continue working through implementing all of those safety checks and balances behind the scenes. But however, instead I choose to take a different approach. I guided the team to do the right thing, even if it meant I was having a tough conversation with my bosses or all of my stakeholders as to why we are not ready for the launch. So we did all of that extra diligence and you know, sure enough we caught a handful of bugs that could have exposed some of the sensitive customer data. In the short term, I think we did end up missing the launch date. We had to extend it by a couple of weeks. But for the longer term, we did end up protecting the brand. You know, I think that's the fine balance. Right. I think, in my opinion, I think that servant leadership, I think is remembering that the real scorecard is, isn't the one that's immediately ahead of you. It's actually the trust that compounds over the years that you're trying to serve for your stakeholders. So that's how I see it.

Yeah, yeah, most definitely. Thank you for sharing that. Because there's just so many short term thinkers in business nowadays and I think having people that are thoughtful about the due diligence that goes into making sure that you're putting those guardrails in place before a premature launch is just so important. So thank you for that. For leaders racing to adopt AI, what are three pitfalls that you're seeing and one habit that reliably Prevents them.

Okay, so I think this is the favorite question, because I get asked this question many, many times. The way I would frame it is I see similar kind of pitfalls come up over and over again. Maybe one of them, I think, is a lot of companies prematurely try to treat AI like a magic wand, expecting that you're going to see instant ROI just because you started leveraging AI even without fixing any of the underlying data footprint or the data quality issues that need to be tackled, that magic is almost never going to happen, number one. Number two, I would say, is the governance part of it, Right. If companies think of rolling out features without the right guardrails, be it assessing for bias, assessing for privacy, guardrails, assessing for compliance and things like that. Right. So if you skip on these governance aspects, it's almost like you're waiting on certain problems to explode at any given moment, and that's not how you pursue it. And then the next one I would say, is a lot of companies try chasing the demos instead of building resilient platform capabilities. So, you know, again, something that is not going to scale for the longer term. So I would say these are the typical ones that I see. Now for the second part of your question. The habit I think, that prevents all of this, I would say, is, you know, I use a fairly simple approach, which is one I think we should slow down and, you know, frame the problem clearly before you start developing anything. Every successful AI launch started not with algorithms, not with talking about LLMs. It actually started with, you know, with a more important question, which is, what is the specific pain point that we are solving for the customer and how do we earn their trust 100% of the time while doing that particular solve? That discipline of staying focused on the customer user Persona and avoiding the shiny object syndrome has enabled us to build, responsibly, build adding value for the customer and things like that.

Yeah, for sure. I want to talk to you about privacy now, because privacy is now a major concern for most people. What frameworks do you like for minimizing data collection and exposure while keeping insights strong?

So frameworks for privacy, I would say some basic principles that I rely on are, I would say minimization, anonymization, compartmentalization, and things like that. What do I mean by that? I think when you're building a product or when you're building a user experience, don't collect what you absolutely don't need from the user, because every single piece of information that you collect from your customer or your user is the onus is on you to manage it responsibly for the lifespan of the product, for the lifespan of the customer. So that's one anonymization I think is essentially around stripping all of those sensitive identifiers so that the data is still, you know, let's say if you're exposed to some kind of vulnerability or breach, even then that data should be unusable, should be undecodable for the person who's doing that breach. So that's the second part. And then, you know, the third part, the compartmentalization is essentially keeping all of your sensitive data in tightly controlled zones. Even if you have a private cloud, even within that private cloud, even within your organization, the access to that sensitive data has to be even fewer and limited to only folks who absolutely need it. Yeah, so that's how I see it. It's also a discipline muscle for an organization that you will have to inculcate within the organization. I think that's another important aspect of it.

Yeah, yeah, that's great. Fraud and abuse evolve fast. So what signals or patterns matter most in 2025? And what's changed since the early machine learning days?

Okay, great question. So in the early machine learning era, I would say the fraud models were mostly built on static rules. Like, you could look for failed logins, repeated attempts, unusual geographic locations that you typically don't see, mismatched devices. All of these were kind of static rules, in my opinion, that did not change overnight. So they were kind of static. And your machine learning model could kind of adapt to identifying these patterns by kind of studying through data. But now I think fraud looks very, very different, because now, just like all of us, I think attackers are also using a ton of automation, a ton of synthetic identities, and sometimes AI generated content. So all of that, I think now translates to the ability for machine learning or AI to detect behavioral and relational signals, for example, how accounts interact, or whether transaction velocity for a particular credit card or an account, is it very arbitrary and things like that, and whether these patterns kind of repeat across networks and things like that. So I think the shift has been from being able to detect the bad actor to more of being able to spot the suspicious ecosystem from the bigger footprint of users and things like that. So, you know, I think every technological evolution, I think fraudsters are also kind of advancing their capabilities, and we have to keep our systems abreast with all of those capabilities to be detectable.

Yeah, yeah, for sure. So this is a different kind of question. If you had to choose better data quality or bigger models why? And really can one coexist without the other?

Okay, okay. So if I have to choose between, let's say better data quality or bigger models, I would vote more for better data quality every single time. That's how I see it. Let me explain. Back at Facebook, when we were building content recommendation and personalization ML capabilities, we experimented with a ton of different large models and so on. But the end discovery for us quickly was that if the data footprint that we are feeding to the models had glitches like duplicates or missing values or let's say, or there is bias baked into that data set, then no matter how big the model that we try to leverage, you're basically pouring dirty water into a fancy coffee machine, which means the coffee that you're going to get is never going to taste the best. Right. On the other hand, if we invest in platform capabilities in terms of cleansing all of those data quality glitches or standardizing the underlying data content, then even with the modest models, they perform incredibly well.

In that sense. I think I would vote for improved quality as compared to, you know, the bigger models.

Yeah, yeah. And of course, if you get the data quality and you have the bigger model, you have a better result as well. Right. So you and I have talked about this before and I'm really excited about this because I'm into it. You know, we've talked about health, lifespan and leadership. What do you measure and what daily practices keep your energy and clarity high?

A ton of things is a short answer, but let me add some color to it. You know, I would break it down into a couple of different buckets. So one I would say is the morning routine or the mental hygiene to start your day. So I typically try to start my day early in the morning. I start with meditation. I don't use social media, I don't rush into work related emails, any of it. When I start my day and you know, I hydrate myself with water quite a bit and I journal a few thoughts and then decide on what are the important priorities that I can solve or add value in that given day. So that ability to start my day with some calm prevents the chaos through the day. So that's the balance that I find. The second one I think is what you feed into your body is probably the most important thing. So the regimen around diet I do follow intermittent fasting. So intermittent fasting combined with finishing a dinner early has been absolutely transformational for me. And I typically break my fast in the morning with a protein smoothie that's packed with greens and berries and nuts and things like that. So that also gives me a steady inflow of energy minus the sugar, minus the heavy carbs, which can make you feel more drowsy over time. Right. So in that sense, I think it's fueling my brain as much as my body with the diet regimen. And then the third one I would emphasize is fitness. Right. Any point in time when I'm feeling overwhelmed, I try to go and do some workout or take a nice walk with the nature, and all of that kind of calms me down and brings me to a more composed state, and that helps with being more energetic and having a fresh new thought process. So the way I see it is, you know, all of these buckets, I think, be it the morning routine, be it the diet regimen, be it the fitness, all of that helped me maintain my composure, maintain better decision making under stress, and also make myself available as a leader for my team and my stakeholders and all of them.

Yeah, yeah, Great, great input. I love that. Talk to me about a moment you changed your mind about a major product or policy and what evidence moved you about that?

So I'll talk about something that happened at, you know, Meta. Back at Meta, we were trying to roll out AI models that detect misinformation. And initially, like I said earlier, my instinct was also that we would use bigger, more sophisticated models that would potentially solve the problem faster in a more accurate manner. But then, in line with what I was sharing earlier, the evidence started to trickle in and the data showed that the accuracy wasn't improving as much as we thought. And more importantly, some of our users, the user engagement was dropping, probably because they were losing trust with a higher number of false positives and things like that. So with all of those learnings, I kind of changed my perspective. Instead of doubling down or staying razor focused on the latest and the greatest model, we started investing in better data quality and better explainability. We worked on labeling pipelines, we added better context, we gave the users more transparency into why we were flagging something, citing the right reasons for it. And that actually, the results from that was fascinating because our accuracy improved, our user confidence improved, and that translated to better user engagement and so on. So the learning for me there was, I think when it comes to product or policy decisions, the evidence that truly matters isn't about the model performance. It's more so about the human feedback loop and how well are you able to measure that. Human feedback loop is where I think the proof is in the pudding. Right. So changing my mind There reminded me that the digital protection or the privacy guards that we build and we continuously measure, all of that enables the right user experience for all of our end users.

Fantastic. Yeah. Again, it's aligned with your previous answer about data quality. So I love that. If you were advising a community nonprofit today, what three low lift steps would you take to protect donor and family data?

So essentially easy wins in terms of protecting data footprint for a nonprofit is what you're asking, right?

Yep.

Okay. So simple ones I would say is use strong passwords, enable two factor authentication for all of your financial accounts, your donor systems and things like that. Because you know, you do not want any kind of casual break ins, especially when you're a nonprofit and trying to optimize for cash flow and things like that. Two, I would say, I think something I mentioned earlier, which is do not try to store every piece of data forever. Keep data only in the context of what's needed for whatever duration you need. Or even better, you know, keep all of that data in some sort of low cost storage that is not easily accessible and things like that. So that I think is good. And then I think also train and educate and empower your organization and the people within the organization. Because if you think about it, a lot of breaches don't necessarily come from hackers. There is a good chunk of them that happens because somebody clicked on a bad link within your organization and clicking on the bad link created some vulnerabilities, created some exposure. So you want to prevent those things because that's very well within your control. You can prevent all of that. So these would be the easy wins in my opinion.

Yeah, yeah. And again, in our previous conversations, everything's always been about ethics and about guardrails and about legacy. And this is the legacy question I want to ask you again. This is what this show is all about. How do you want to be seen in the future? What do you want your impact to be? So my question is, when your team or family Talk about impact 20 years from now, what do you hope they say and what are you doing to earn this right now?

Okay, I think that's a tough question.

Let me.

So if I think, you know, 20, 30 years ahead, the way I think is I don't want people to talk about the tech stack that I helped build or things of that nature. I would rather have them think along the lines of, okay, he took up a problem statement and he left it in a slightly better or a much improved state than where he had inherited it from. Right. It could be systems. It could be teams, it could be family, it could be, you know, it could be any of these things. It is, for me, it's about how I have added value or created value. And more than likely, it's incremental value that you create because that way you're getting that feedback that, okay, you're on the right track creating value, and so on. So that's how I see it. You know, in terms of work, it's mostly about how do I inculcate the right processes so that the organization can scale? How do I inculcate the next wave of leaders who can enable taking a bigger, broader responsibility so that we are collectively able to deliver more as an organization for communities? I think it's about how can I deliver value based on my expertise, based on my leadership principles. It could be startups, it could be, you know, in the space of data protection, or it could be for enterprise platforms and so on. And at home, I think it's about how I double down on my health and make my presence for my kids, for my wife, for my extended family, and create a conducive, healthy environment where everybody is able to pursue their own dreams and thrive through that journey. That's, in a sense, what I think about the imprint that I would like to, you know, contribute for everyone that I'm working closely with.

Yeah, yeah. You know, I think if anybody can say they left the world a better place than they found it, you know, that's enough for me. But I love the way you broke it down into different categories and you're really focused on making sure you leave a lasting impact on every aspect of your life. Well, Manoj, sadly, we've come to the end of the show. I really appreciate you coming. I want to thank my special guest, Manoj Mohan, for sharing his journey and his insights today. Thank you so much for being here. I'm Alan Pavloski signing off for Legacy All Stars.