"As you become a founder, there are a couple of things that are fundamentally very new. Definitely things that I feel are counter intuitive to people who have had longer careers outside of running a startup. In a large company, you will not tolerate a petulant genius. But in a startup you would monetize or you would operate based on that, or a person who is ancredible sales person who refuses to talk to anyone else in the office. I think if you come from any form of more senior large company role, you create a crystal clear message at some point, this is what I'm doing. I'm going to repeat the same message for 18 months. It is in the contract."
- Galvin Widjaja
Galvin is the CEO and Co-founder of Lauretta.io, a profitable Techstars company that builds bleeding edge military grade computer vision technologies that are now available to commercial buyers with no AI capability of their own.
Prior to this, Galvin wore two hats as a data strategist in DBS, building RegTech products, while also running Es Teler 77 Singapore, a chain of quick service restaurants serving nostalgically authentic Indonesian food to our migrant population. He was also CFO of a New York based Uber-for-everything startup as well as first local employee of a boutique management consulting firm that has since been acquired by PWC.
In his free time, you will often find Galvin creating new recipes in a bid in increase his children's tolerance to spicy food.
Jeremy Au (00:00):
Hey Galvin. Good to have you on board.
Galvin Widjaja (00:03):Hi Jeremy. It's really great to be here.
Jeremy Au (00:06):
I think what's really interesting is that I'm super excited to bring you on board because you are taking an interesting take on AI, on Vision and on surveillance and protection. Those are really interesting fields because it's going to generate a to of value for people across the world. Also it's a question of much debate around the world as well. Let's have a chat about that. For those who don't know you yet how would you describe your professional journey?
Galvin Widjaja (00:40):
I started as process consultant. I graduated from SMU with a degree in quantitative finance. Quantitative finance is like pattern recognition. But at the same time I didn't really want to be a finance person. I didn't think that it was that interesting to look at the numbers alone. To be very honest it was just to get off my own head. To be a finance person you go into your own head. I think I was deep enough in that. I needed to leave.
What I did was I moved into management consulting. I did that for seven years in a company that has recently been acquired by PWC before being pushed over by a ex-client into DBS. Then I went from process architect into data architecture. I did that for a number of years. I spent a stint in Boston. My wife was at Harvard. I was at MIT. I was hanging out with the system dynamics crowd which was also people doing pattern recognition in the process world but really geeky versions of them.
That's when I got my first taste of the startup world. I was CFO of a small startup doing Uber for X. 2015 there was many Uber for X startups all of which are now dead. That was a really good experience in going from zero to raise $6 million and then return all the money within six months. Then it was time for me to try something on my own. In 2017 two things was happening. One was that I had been unceremoniously given a restaurant chain which was giving me a heck of a culture shock.
Very different types of people, a very human work that I didn't need to experience when I was working with everybody who was like very highly self-actualized at the consulting and banking world. Then I was doing that and I decided to quit my job at DBS and go full on into my startup.
I build a startup there called Lauretta.io. It was initially built as a friendly manager that was supposed to help the restaurant staff. Then I realized step one, first thing I learned that the restaurant staff don't need help. B, they won't listen to this whole system. Because the data is so complicated that I needed to know what was happening in the building before it made sense what they were doing. Then it reached a point where it simply became the same cost to automate the data for the entire building as it was for my shop. Then I was like, just sell it to the shopping mall. That's how the business started.
Jeremy Au (03:29):
What a experience that you have there. What's interesting, of course, is that you've done a bunch of professional career journey as a consultant, both external and internal consulting. What was it like learning to be a consultant at that point of time? Do you think you learned some good stuff from them?
Galvin Widjaja (03:49):
No. Management consulting is the science of synthesizing a very little amount of information into a very large amount of insight. It is the diametric opposite of AI where you synthesize a large amount of information into a very little amount of insight. It is almost luck plus good salesmanship that you get your insight push to the claim. I think I learned a lot through it but at the same time it was really the upbringing question. Demanding parents but I also have indecisive parents. I decided to do all of them. Do the banking route and the consulting route and the accounting route and the company route. Just one after the other.
Jeremy Au (04:44):
Let's get into that. You're talking to another management consultant and you already knew that. My heart broke. It's like all the Bain and McKinsey. Did I not learn anything to be a founder? Ain't I qualified to be an AI leader? That's a very a sharp set of words but I love what you said. I mean, let's dive deep into that. You said there is a very little amount of information that you're building into a larger set of insights as consulting. What do you mean by that?
Galvin Widjaja (05:21):
Nearly every top consulting firm in the world right now falls into two categories. One of them are consulting firms that is 80% operations, 20% strategy. Then there's the flip side, which is primarily strategy. The first thing you should realize about the high strategy companies is that they are specifically designed to operate on their domain knowledge first before coming in.
Let me walk you through a real example, something that would happen. I will go into a very senior client. Knowing 99% of the things they're doing very well but I can point out the 1%, the blind spot. The part of the Johari's window that you don't know that you don't know. I'll go in there, I'll sit down. Then I will in a flourish show this one thing based on this heuristic that I know that they would probably have missed.
That I will look immensely intelligent knowing full well that in the 99% of the things that they do well, I have much less information than they do. Yet at that point in time, they would trust me more than that person who has been operating as a project manager in that project for 10 years. That is sad but it's really how it works. That is how you do it.
Jeremy Au (06:47):
Well, we got to go deeper into this one. Suddenly I feel like a few defenses, do I have to defend management consulting and the honor? Well, what I agree with you on is that clients, if they have the capability to consult a management consultant and even consider spending millions of dollars on a project is they have to be doing something right.
They have already created products, an economic engine of getting value from those products and continuing that cycle of production. They already have a team that's already built out to do that stuff. You're right. I think I agree with you that most companies are 99%, their workflow is embedded and ready to go. We're not fundamentally trying to change that. I do agree with you about a 1% of the things that they don't know and trying and figure things out.
I would make the fun counter claim, the apologist for management consultants is that those 1% times out of times when the executive is trying to make some large capital allocation decisions in terms of structure, restructuring or market turnaround, cutbacks location. Even though there is this 1% of things they don't know, but that 1% of the time is really at times when they really need someone to help them. It could be a difference between a great decision and a good one or a good one and a bad one.
I'm not saying the consultants get it right all the time as people in the industry know as well. But at least it's better to have more heads in the room than less. Especially folks who are coming in with a bunch of benchmarks, coming in with a bunch of taught leadership. I'm saying all these buzzwords that I hate myself or something like that. But I'm saying there's some value with us, tough moments. What do you think of that Gavin?
Galvin Widjaja (08:46):
I guess it's interesting that you say that. Different companies have very different weightages of change versus operate. With that change, you will find that the team for consulting sits in a very particular area. You look through the MLC, it is a top the Fortune 500. What you would find is that in the places, you will see that there are three types.
They're companies that have no consultants. Very rare. The majority fall into the two smaller types of, they are strong and steady companies that are using consultants to change. There are companies with very strong core businesses which they're using consultants to support the support functions. A good example of that is that if you were going to Exxon Mobil, you would not consult them on oil explorations. You would ask them to consult you on oil explorations.
But when it comes to things like their finances, their accounting, the way they do the books, there'll definitely be better people out there for it. That makes a lot of sense. The other thing that you mentioned which has things like high campus decisions is also true. But in my experience, it is 50 50.
This is a point where we can't name names. But there are many jobs that both of us would probably have been in where you know full well that the conclusion of the research that you do will be in favor of the person that brought you in the first place. It is true. That support may be needed. Sometimes you need to have a Vice President to split the difference in an even Senate but that's just what needs to be done sometimes.
Jeremy Au (10:49):
We're going there. Well, I would say there're two parts of it. One is in order for someone to make the call that you're going to spend a couple million dollars to have a bunch of freelancers or consultants or mercenaries to basically parachute in and fight a war for you. Quite clearly there's a problem. I mean, obviously there's a bunch of peacetime people who are using consultants to think about strategic review, et cetera. That makes sense to use this bunch of too.
It'll help you think about the future but I think if it's in wartime the problem is clear. I think you're also fair to say that if you have the mandate to bring in a couple of million dollars and blow that on a bunch of teams to ride in as calvary to support, you got to be able to point them at the right direction.
I mean the CEO and a sponsor team should already have a good sense. These are the problems et cetera. Obviously the high level problem is quite simple. Where you brought in the Polish and a Swiss mercenaries and a, you point them at the enemy and just, go dead away. The answer is there in that sense but how does being executed actually and how does actually being translated? There's a huge amount of work that actually still needs to happen.
Maybe the answer is we still have to right size i.e do a work force reduction i.e lay off folks. There's a target of 20% for costs but it's really difficult. You can imagine for an established organization to figure out independent/as impartial as possible with as little claims of bias as possible to run that process in a very fast or even cohesive way. In that case maybe it's just easier to hire McKinsey and ask them to do the hard work of saying, we need to lower costs by 20%. But whose name goes in there, McKinsey essentially running that process at some level.
Galvin Widjaja (13:09):
I have an interesting story around this thing that its pro consulting. It makes sense in this direction. In this particular project that we were on, there were two coal mines and they merged. They had dug close enough that you could take the two housings of the mines and just put it together and have one big housing for the two mines.
At that point in time, the obvious question was where do you cut things? But there's absolutely no reason why I thought that they had to be cut. There was no performance impact. None of the numbers can be matched. You cannot compare the numbers of these two mines. One has lot of coal, one has less. You can't change anything. The two biggest things that impact the successful mine, there's only two values.
One of them is the thickness of the coal and second is the weather that year. There's nothing you can do about it. There was at a point where they brought in consultants. Essentially the consultant from day one, they knew that there were a set of a very large number like 25 cutting exercises where there is no right answer.
To a person within the company, there was no right answer. To a consultant funnily enough, based on how the organization is built, have no wrong answer so long as one is cut. It would have worked out. The very dynamic of the large organization forces the consulting role to exist within that field. There's nothing wrong with that. Obviously it won't disappear. It's a question of which parts of it are good and which parts are a bit strange.
Jeremy Au (15:07):
Yeah, I totally agree about that. I mean, I think that's where it boils down to it. I think the big difference is that the strength of the consulting company in relation to the client is the process orientation of the consultant. Which is that to a company who was doing a lay-off or market turnaround, they hardly do that. Maybe they do that once every economic cycle. Once every five years.
Whereas to a consultant, he's zooming in, just like this partner and his team of consultants and they specialize in, I hate the phrase, but right-sizing companies. It was like Up in the Air, the movie. I mean, that's a bit actually doing the head count reduction versus the planning of it. But to someone else it's Tuesday for them.
It's like me and my wedding. It's like for me the wedding was the most important day of my life and I've never done it before. To my wedding planner it's like Tuesday. It's like this is another wedding. At the end of the day I just followed the process that she set up for me and it worked out. I could have recreated the wedding planning process from scratch. It would have probably been cheaper in terms of cash. But it would've been a giant pain in the butt. From emotional energy, actually getting married at the end of the day process.
I think that's how maybe consulting agencies are like the wedding planners for the companies. Like in terms of the process orientation. They're not going to walk you down the hall. They're not going to figure out the ring or the flowers, but they can lay out the stuff.
I really agree with you. To some extent because of that process orientation, the company knows there's a problem and the rough aperture of the solution is there. Then you're bringing in a consultant to run the process, to get to dare and actually boil down the level to a level tree answers to actually do that. It was like, we agree with the layoff 20%, but who exactly are the names? What functions to preserve? What functions to cut more deeply? I'm using this very negative example, but it's because it's a question and problem that no company really wants to do. Everybody wants to do the exercise called how do we deploy a $1 billion capital bonanza. There you don't need a consultant to spend a billion dollars.
But it's quite hard to do the opposite. Which is how do we cut a billion dollars some more? But I do think there's something that you did find a contrast? Well, you contrasted that to being a founder and being in an AI space. You talked about how the process is almost opposite which is you're looking for deep insights that you're trying to execute on as a founder. Did you feel like there was something I felt I had to unlearn to some extent? I had to unlearn being process oriented as a consultant and focus on really being results and action oriented as a founder. Did you feel you had to go through the same transition?
Galvin Widjaja (18:26):
100%. As you become a founder, there are a couple of things that are fundamentally very new. Definitely things that I feel are counter intuitive to people who have had longer careers outside of running a startup. I found two things. One are the type of people and how you work with them are fundamentally different. In a large company, you will not tolerate a petulant genius.
But in a startup you would monetize or you would operate based on that, or a person who is an credible sales person who refuses to talk to anyone else in the office. There are a lot of these in the startup world that are very useful and they will be remain in the startup world. These things I found very interesting.
The other thing I found was that if you built your product, you are constantly in the message building process which is something I was always learning. Something I really disliked for a long time. I think if you come from any form of more senior large company role, you create a crystal clear message at some point, this is what I'm doing. I'm going to repeat the same message for 18 months. It is in the contract.
But as a company is not three months in you're, I'm the best X and two months later, I'm actually not. But I have pivoted to something slightly similar. Similar enough that I don't have to tell all my other clients that I have changed my focus but different enough that the message for the new client has to be different. You start to have this very choppy message which for ages I thought I'm just a super bad sales person. I don't know how to do this and I realized like, no, this is it. This is how it is.
Jeremy Au (20:38):
I mean, I think it's so true. I think that's such a really good insight. Because you're right. As a consultant you had a three months project. Then you have your steer cos that you're doing every week or every two weeks. There's very deliberate zero defect. Perfect mini presentations to pre-wire for them. Then you end it off with a well-received presentation to everybody already who knows in advance so they're not too scared and spooked by the answer and it's cascaded down. It is a very intentional plan. There is a strong cadence but it's a planned process. But you're right.
As a founder you're just building a company step-by-step. It's a nightmare in that sense to believe that it's a one-off presentation. It's a one-off message to deliver that makes sense because it doesn't. Because you're like one month you have five clients and then next month you have four clients because you're all fundraising. Then the next month you have the six because you work your ass off and you stopped fundraising and then close the client back and you want the person back. Then you got another client along the way. There's some craziness, say like building a plane while its flying. That's what they say.
Galvin Widjaja (22:05):The Reid Hoffman. Jumping off a cliff and building a plane.
Jeremy Au (22:11):
How do we heighten this analogy more? It's like jumping off from the space station and building oxygen mask for yourself. It's like a single asteroid and finding a team of oil miners to drill a nuke into the middle of the asteroid.
There you are. You're making this transition into being a founder. You're learning to be a founder. Tell us a little bit more about why this AI dynamic? You building out this restaurant obviously. Then you're taking over the family business. I remember you and I hanging out and talking about how you're suddenly running a fried chicken chain.
Galvin Widjaja (23:02): Just in front of your office.
Jeremy Au (23:07):Exactly. I enjoyed the fried chicken I got to say. Suddenly you're there and then you're like, okay, AI. How does that happen?
Galvin Widjaja (23:18):
There a couple of things that I've learned about this. I think the first and foremost thing I that I had to learn was to accept the process. It's not as linear as I wanted it to be obviously. But beyond that, I think the first thing to think about is the question of how process oriented we are.
I was a true pure play process flow expert. There was a point in time where I would be flown into a place to draw the process flow. That is my thing. Imagine when I ran a restaurant chain it would open an hour late. The staff had relationship issues. I had to deal with police, investigating officers and gangsters at different points in time. Everything. It's a vastly different experience. Yet throughout that whole time, I initially thought I could build a form of automation or process across the top.
That's how we started. Then I realized that I'm now in a world where 99% of the things are not measured. Just to give you a bit of a color into that. Five years after I started that process, that was when there was an open innovation project for identifying the mall food waste that was provided by IMDA. That was five years later. That project wasn't resolved.
All your stock taking is rubbish if people are throwing away your stuff. Half of your ERP now it's wasted. Then the other half is paper, is physical pen on a receipt. People give me the receipt, but the receipt is blank and it's written in pen, the amount of money that they paid. What is the point of the ERP system?
You can see that all of these gaps were in that place. I decided to go down to the fundamentals. The fundamentals was, I try to three things. One of them was intensely tracking each of the staff. That was a total failure. As soon as I realized that wasn't working, we turned that off. The second thing I did was we used cameras to capture the information of customers coming in. Now I realized it was
Then the third thing I did at that point in time in 2013 was that I looked at the staff and I was still a person who had never run a business in my life. I decided to give everybody a raise even though the company was not making any money. That also worked really well. It worked so well that we were the last F&B shop to close down in had a 100 percent closed down rate during that time. Because it had been under renovation for more than a year.
We outlasted the entire food court. First the food court went bankrupt and then the food court owners themselves went bankrupt too. It was really bad. But I mean, loyalty was component one that worked out. Tracking people did not work out. Not how many customers there were about. Who they were? Why they were coming? Was such an interesting thing that I decided that's something I can really build up. That's when I began down the AI route.
The weird thing about that is, that's when it converged to the kind of business that I would have started if I had none of these limitations and things in place. My actual personality is I'm not a huge people person. I am a things and tech person. When I got into that field with a decent enough mathematical background I realized that, we can take this really far. We can take this as far as it has been taken very quickly. That's what we decided. That's why I decided we can do it this way. That's the point where you're to choose your direction. You choose a direction and then you just run with it.
Jeremy Au (27:59):
There we are. We are dealing with you. Since then you've been successful. You've deployed your systems and camera vision EI surveillance and protection in industrials for military and in commercial. What has been the drivers of your success so far?
Galvin Widjaja (28:18):
I think that one fundamental thing that has driven our success most of the way and that is that our business is designed to have the same shape as the evolution of AI. What I mean by that is that if you are for example, a media business, the most important thing for your business is that it scales naturally because media is always its own bottleneck. It's the more you have the better it is.
If you are however an AI business, your business's most important thing is to fulfill the most important gap of AI. The most important gap of AI today, well in the research world it'll be the TRL 5 to 7 space which is essentially the working prototype to I will buy this product space. That particular gap, if you essentially build your platform around that gap, you don't actually need to build the 1, 2, 3, 4, or the 7, 8 gap.
That's where we decided to build out stuff and that has been very successful. How does that work on real life? That means that the core AI algorithm that we have has changed from moderately good, industry standard all the way up to let's get picked up by the U.S. Army to see whether this is better than what we're having good. The actual interface for the app, it's the same. You can go to settings and change it. The app can constantly build platform type capabilities and the AI can keep evolving.
Maybe two years ago you would ask us how many people are at X and you will ask us something like when should you clean the bathroom and tomorrow or the today you could ask us for this person that placed his bag down on the floor, how likely is it that the person was doing that as a terrorist versus as a tired person? The nuance of the question has increased significantly but the interface is the same interface.
Jeremy Au (30:37):
That's amazing. I think that's really something that's under appreciated. It's like the design architecture to allow the known improvements of AI to be able to come in on a regular basis. Because we know the algorithms are getting better and better. We know that the training data is becoming larger and larger in terms of the set size. We know that sudden advances that are being pre-wired to us by the journals. All that stuff. I think it's super expensive to re architecture the whole company every time one of those things happen versus like, hey, okay, is it new release? Okay. We just naturally embed that performance improvement into our system and workflow.
Galvin Widjaja (31:23):
I think it's actually even more obvious in the AI. We are specifically in the computer vision space. It's even more obvious in our space because you would be able to predict the longevity of your competitors. We are in a super red ocean space. That is the one problem with AI today. It is very red ocean, very competitive. There are a lot of like for like that many of the customers are not educated enough to know the difference between a good and a very good product.
You would know that there are people who are bigger but will fade off at some point. There is a bit of Darwinism in it. When they die, they create food for the masses. Because every one of us does education. As we educate we increase the size of the pie, then you die and you leave the pie for others. That's kind of the evolution of the ESPs these days.
Jeremy Au (32:21):
Because we think of surveillance, obviously surveillance of industrials, surveillance for military purposes is pretty straightforward. But also like in the commercial sense as well where there's obviously the privacy dynamics expectations of consumers. How is it like? I mean, obviously you had to make a decision to be GDPR compliant. Then you had everything to do. How do you feel about that? Because I do know of competitors who have chosen to not be GDPR compliant.
Galvin Widjaja (33:12):
Almost all of our Asian competitors are not GDP. The first thing is there is actually a higher standard than GDPR that we encountered before we ever touched the European space. That is when we did our first project in Australia. One of the things that you do not do in Australia is you do not impact the rights of workers. Because essentially Australia's government is the Labor Party. It's a union.
Actually we found that to be by far the hardest. I think this is something that maybe I'm willing to share this which I think is a very important insight. In a computer vision space especially when it comes to privacy, if you are the kind of company that doesn't really care about these things, you will truly only stay in Asia. Because it's really only Asia right now in South Asia and Africa that has very low privacy laws.
Even when we were talking about the U.S. military privacy came into play. Because the contractors in the U.S. base are like the wives of the Colonel. You cannot facial recognition the wife of the Colonel. That is the worst thing you can do. We made this decision right from the beginning from an ethical standpoint.
The moment we reached a point where we knew that our technology was meaningfully ahead of the industrial standard. It was time for us to take that narrative and say, if you are a first in a space, you get to carve out what the role looks like. We made this decision to not save any personally identifiable information. By doing that, components like tracking people in a shopping mall became a lot harder than facial recognition.
On the other hand, it also meant that we could deploy pretty much anywhere at the same time. We ended up at both sites. To do that we had to essentially go to four or five of our potential clients and tell them that we can't do it if facial recognition is required. An interesting one is that, there was one of the Singapore government bodies that was requesting facial recognition on every floor of a building. Any requirement that came out from that particular industry over the next two years, we couldn't apply for. Simply because that was one of those minimum requirements. You need to track everyone in the building. Once we chose not to do that, then we built that title for ourselves and it somehow managed to work out.
We essentially were able to close one door and open others for ourselves. It makes a lot of sense. We also decided to take the next step. Next step after that is to crystallize our point of view into policy. Once we took that stand. Then the next thing to do was to talk to people in, for example, the ethics committees in Singapore and discuss these things. Take the very large gray area that sits on top of the PDPA in Singapore and turn it into a bit more of a black and white. In that way, you get to choose how much privacy we will be giving in the future.
Jeremy Au (36:48):
That's super true. I really respect you for that. Because at the end of the day choosing to be GDPR compliant is not only obviously a business decision, but also an ethical decision, but also a strategic position. In this case like you said, you had to let go of some business as a result in order to do the right thing, the legal thing at a minimum.
It opens up a whole bunch of opportunities that you couldn't access before. I think that's where the crux of it is. It's like you're meaningfully better but not obviously 10X better in terms of performance. But you are 10X better in terms of privacy compliance. Obeying the law of the land especially in the context of EU citizens who required privacy protections on a global basis.
Honestly that's the trend where the world is going. I mean even China is starting to move towards more regulation of facial recognition technologies rather than less. Even though they currently have a headstart on that. I think it's really the right call. I think obviously on a personal and ethical level but also I think the right call from a strategic level. It must've been one hell of a thing to do from a commercial level because you had to let go of some clients in order to pursue other ones.
What's it like to be part of a field where the policy is still being written? It's still very much relies on the industry to contribute a point of view. Because if you and I were building a startup in let's just say stuff that is a little bit more known like construction, et cetera, where we're very much abiding with current codes but trying to figure out efficiencies with it. But when we cut out about like crypto and we're talking about AI and computer vision, we're talking about drones. Very much the policy is still being built. The rules don't even exist. Maybe it's like a guideline or industry policy, body policy which is not the same as law. What is that like? Do you have to do? How do you think about it?
Galvin Widjaja (39:03):
The honest truth is that I absolutely love it. One of the things in my past is when I was being a data architect in DBS I was actually tasked towards regulation technology. In regulation technology, actually a big part of that is the fact that regulation technology on the surface is very obvious, which is there are laws. We need to do them. Connect the do with the law. In between the interesting where we can take you to where everything is missing. There is no real connection between the law and the person doing it on the ground. There is this invisible layer of interpretation in the middle. During that time I found out very early that I could gather the people who interpret the laws. That unit could sit in the middle at the council of men.
You may not be the ones who create the law nor are you the one who do it. But yet you're able to craft both sides of the narrative. Because you see, you allow the people on the ground to see how the law is written and you allow the people in the ivory tower to see how their law comes out and you actually control the narrative.
As a startup in this space, we are actually very much the same way. There are always guidelines. The guidelines are usually a combination of very strict but fairly antiquated laws but also a combination of the sentiment on the ground. Any of us know, and Jeremy I know that you spend a fair bit of time doing polling as a bit of a side job. You know that the sentiment on the ground is very inconsistently communicated up.
It is the way that a poll is written, creates the outcome that you see at the top of it. Even though the sentiment on the ground is the same. Likewise as a startup in this space, the way that you write your narrative allows you to change the sentiment on the ground. I give you example, in Singapore more than half of the country prefers their privacy to their operations. It's one of the lowest in the world. It's like 60% prefer their privacy versus more services. But if you communicate this in the same way, like we are the Google analytics of the physical world. They are like, I've given you my privacy in the online world. I can do the same thing in the physical world. They are actually very different because in online world it is contained within a particular area. Lazada does not wash cars.
It's very specific. But in the physical world, everything you do in between is far more indicative of who you are than what you do at the point of execution. If you take that narrative, you actually continue to allow people to take more, to give up more of their privacy inevitably without them knowing. If you ever take the other route of these others things I am not capturing, then people will immediately realize like, so everybody else is going to read that. Then the narrative becomes very different.
By doing that, you actually get to guide the narrative into the future. You can also play the strategic card that way. Because you can essentially take the hidden values. Like the brave browser of the worlds. That you can take the hidden benefit that people are monetizing and you convert that into your unique selling point.
Jeremy Au (42:41):
That was really interesting. Because I think you said two things that can be under appreciated. Which is, one is that privacy in a digital domain is what most people think about privacy and people are currently extending the same metaphor for privacy in the physical domain.
What you're saying is actually the privacy in the physical domain is way more valuable than privacy in a digital domain. Because I'm going to Lazada to buy more cat food versus in a private world, you actually are getting a very deep profound what the person is doing. That's both the value of the physical data but also the value to the business but also the value to the consumer and as a privacy.
I think the second thing that you're talking about that is pretty true is, we need talk about controlling the narrative. You're not necessarily controlling the narrative to destroy privacy. You're trying to control that narrative and craft it so that you can preserve more privacy. A little bit more like Robin Hood than the sheriff I guess and this example.
I guess the tricky part off course that's here is you're building out this CUPE division, you're building this out in the GDPR compliant way, does it feel like a bit unfair that all these like Asian tech firms that, I'm aware of a few of them as well, it's just like their sucking up data and they don't give a shit about compliance of GDPR and assume that there are no EU citizens and it a few division somehow. Be able as a result to sell a stronger commercial case two of the shopping malls did the security cameras outside homes and public spaces. Do you feel it's unfair?
Galvin Widjaja (44:30):I used to think it was unfair. But I realized that a big part of that is the inconsistency of my own mental model. Here's a bit of a guilty thing that I can share. But I think it is true for most B2B startups.
Very often when you are pitching to the client something, you haven't built it yet. You know that that guy is going to take so long to sign the contract. By the time the contract begins, it will be there. Even if it's six months later which in a lifetime of startup, it's like four 17 sprints or something like that. It's like, it'll be there by the time you sign the contract. The way that I think about things like the way that these companies work today is I think we have now reached a point of understanding of AI to know that you don't need to do that.
That it is far more likely that companies that rely on personal data will find that a handicap in the future. I think we are now at a point where it is starting to be clear that's not working out. In a way I feel vindicated. We didn't really know that from the beginning. But we decided it is worth taking an ethical standpoint, but now it's actually coming up as exactly like that.
A simple one is this. We do tracking. Let me just lay it out there. We do track a person in a shopping mall but we never use your face as a face identifier. Because of that two things happened. One open secret in the software world is that your biometric data is an actual line of code which means that I could copy and paste your facial recognition from one camera, paste it to another one and it will still recognize you. It is actually a key to your identity that you cannot change.
That's pretty frightening. But what we are doing is that we take everything about you, including your clothing and your outfit and then we track the as your entity. That person can be tracked for the next 24 hours. At the end of the 24 hours, two things will happen. One is that we will deliberately delete that information. Second, that information is meaningless anyway because you will probably change your clothes. You better change. I'm not talking about you Jeremy. Not into another black shirt or so. But essentially these two things will happen.
We thought this was hard. It was not nearly as accurate initially, but two things immediately became obvious once we started to do it and implement it. One is that we can actually track you from the back of your head which then goes to the second point, because we were on CCTV cameras and facial recognition actually has an optimal degree of 30 degrees up and down for your face. This non-PII, Personally Identifiable Information method works very well on CCTV cameras and facial recognition actually doesn't at all. If you will imagine in a way that worked out for us.
Jeremy Au (48:12):
That's amazing that you were able to switch your approach but also serendipitously discover that actually the approach does actually give you performance advantages for different use cases which is amazing. Gavin coming up to the end of the show here. But could you share with us a time when you were brave. Tell us of a time when you had a challenge or obstacle that you had overcome and be brave.
Galvin Widjaja (48:43):
That's interesting. Let me tell you a recent story. We had two members of our team who were both very valuable to the team and the relationship between them became complicated. That was something that was interesting. It became complicated in both ways. They became very close and then not so close. Somehow it reached a point where one of them was accusing another one of something serious enough that the police got called in. This happened not that long ago.
One of the thing that we had to do at that moment was, or one of the things that I felt I had to do at that moment was to find a way for both people to feel heard. Even when it was very clear that we are about to move into a point where one or both of them will leave the company and it will end very badly. During the experience I really spent a lot of time, essentially two or three weeks at a company just became blank.
Then we essentially found out what's going on there. During that time where he rebuilt the trust of every individual with each of these individuals. Because once this process begins the process of distrust spreads within the company. Both of them obviously don't trust each other but then everybody else questioned their relationship with these two people as well. That creates a very complicated dynamic. Eventually we reached a point where they actually both managed to resolve their differences. It never fully resolved. Eventually one left, one remained it. It cleared out but actually we are still all on talking terms. That involved amazing things. I can't really talk about too much. We were in the interrogation room. I remember we had a conversation about interrogation rooms recently.
I was in fact in one of them also a couple of months ago to solve this police case. During that time, I think the most important thing to realize is that there are three routes to take. One of them is there is a publicly right route that matches with things like democratic mentalities and like LGBT, that kind of thing. That's one component of that.
There is a business route and then there's a route that takes the human element first. Usually when you take the human element whether you get the first two results is completely a black box. You have no idea because you give the control back to them. You get to communicate, you get to say whatever you want. But if you are lucky and if the environment is right everything works out.
Jeremy Au (51:53):
That's a tough one. I can totally see how that would have been a tough situation. Thanks for being so open to share about how you manage that process. I think to wrap things up, thank you so much for coming on the show. I really appreciate it. Three parts that were really good. I think the first part off course was you opening up with saying that management consulting has no value. That shaping up to actually becoming some level of consensus about which situations a consultant is helpful and which situations the consultants is less helpful. This of course is brought to you by two former consultants.
Then the second thing off course is, thanks for sharing, obviously your experience in building a company. Thinking through the dynamics about what it means to build AI and found a business from scratch. Third off course was your deep thinking about the industry practices and really being a champion for privacy and GDPR in comparison to the industry. I've really had a hats off to you for helping protect us everyday citizens as we go about in our normal days.
Galvin Widjaja (53:14):The battle continues for sure.
Jeremy Au (53:19):Well, thank you Gavin for fighting the good fight for us.
Galvin Widjaja (53:22):
Thank you for having me here. It was really fun.