How can human and artificial intelligence collaborate in eyecare? Robby Dally shares his experiences on trusting AI to dramatically cut repetitive problem-solving time and radically reduce labor costs.
How can human and artificial intelligence collaborate in eyecare? Robby Dally shares his experiences on trusting AI to dramatically cut repetitive problem-solving time and radically reduce labor costs.
I think where you run into that is where you have like very knowledgeable, Business users, right. Who really understand their data. you know, the common trap you run into at first is you make a prediction and they're like, well, yeah, I could have told you that. Right? Like this, this and this. It's a very obvious thing. the flip side of that I would say is, look, if the machine learning doesn't get the obvious one. Right. Well, okay. Maybe there's a reason. It didn't because it wasn't so obvious, but actually that's kind of proof of validation
Ari Kaplan:Hi everyone. I am Ari Kaplan, the director. AI evangelism and strategy at data robot. I asked Robbie Dalley with al-khan labs to join us. he's director of global insights and analytics. So why don't you go ahead and introduce yourself and tell us a little bit more about you.
Robby Dally:Sure thanks. I'm excited to be here, so, Been in alcohol now, 13 and a half years kind of worked my way all throughout the organization. started out of school and it, I spent some time in market research, marketing forecasting, and then most recently I moved into the space of advanced analytics and, trying to get us to the world where we do machine learning and data science. That's been an interesting journey because it's been sort of popping up all over our icon and I'm think as anybody who's going through the journey to machine learning knows that it's not just about the technology. It's also about domain expertise, business expertise. And so, I, I preface kind of where we are now with my background a little bit, because I think that's helped me, Kind of move my way into data science, even though I'm not really a kind of a data scientist by trade, I'm a much more of a citizen data scientist.
Ari Kaplan:Yeah. Very cool. So I saw on the LinkedIn, you've been a brand manager three years. what's it like being on that side of the business?
Robby Dally:Yeah. I mean, honestly, that was, that was super different than any of my other jobs. Right. so I was managing one of our IOL lenses in the United States. So responsibility for marketing materials and working with our Salesforce. And really that was, like I said, it's very different than what I've done before. What I loved about it though, is I got a ton of experience, really like understanding our business down to a deep layer. And, uh, you know, I had a lot of background with data, but now, instead of just like seeing like rows on a spreadsheet that represent products, I'm like out there in surgery, you know, seeing cataract surgery. And now I'm like, Oh, okay. There's the Iowa going? Cool. Right now we're doing phaco. Most of the vacation. I see the tips that are used. Oh yeah. Those are in like custom pads. So it's just, it was really interesting because it helped me understand the business a lot more. I would also say probably not my forever job either though. Right. It's one of those that, like, I did it super thankful I did it, has been very valuable for me, but definitely not a dream job for me. And what I realized there is like, I love the analytics, so. I really tried to bring analytics into the job, how are we making decisions based off of data, not just like gut feeling or like what's some of the key surgeons we talked to are saying. And so, you know, for me that was, like applying that analytics. And then I used that to actually jump into my next job and, and that's really become kind of a recurrent theme for me is analytics, no matter what the job I'm in, whether it's directly analytics or. not much analytics, right. With marketing.
Ari Kaplan:That's one of the things I find fascinating as I been in the industry many decades, uh, 32 years now, I believe, hopefully I don't look that old, but, I used to work at Oracle corporation, and what I found is there's like this natural progression you go from intern. Then you become like an it person, either a data Wrangler, programmer, uh, and then they become, you know, at the time it was like the ideal throne being a database administrator. And now I kind of see people that are in it, trying to become a data scientist and then sometimes, similar skill sets. But in other times, There's not overlapping skills. so what do you see like that progression? Does that make sense?
Robby Dally:Yeah, it does. And Even for me, like, again, I've been at outcome 13 and a half years, like I said, which is I think kind of a more exception and less rule now. Right. Somebody being at a company that long, and I talked that up too. It's a great company, but I also look at it like when I started at Alcon, it was like, you said, very like, you go into vertical and you start going up. Right. And I'm like, I come into it, you know, more or less right out of college and like, All right. I'm going to be the CIO someday. Like that's my dream and I'll just work my way up. And, you know, I kind of realized along the way, like I liked several of the things I did in it, but there were other things I wanted to do. And so I made the jump over to market research as an analyst. And, you know, I realize not everybody out there is like a historian of outcomes. So we were under Nestle then, we were. A part of Novartis and now we're independent. Right? So when we moved into Novartis, one of the things I really liked about that was Novartis is very much into like switching verticals. You know, it's not always just get into like one area and work your way up. It was like, okay, you like this? Great. You need to be more rounded, go over here now and take a turn over here, go over here and do a turnover here. that change helped make it where I can make those jumps across. slowly along the way, I figured out really what I was passionate about, which was analytics. And that's kind of where my journey has led, man. And now I want to, I want to do analytics. I want to do machine learning and data science. I love it. It's just so much fun, right? Like this is the most fun I'm have in any of my jobs since I've been at Alcon. And I'm not just saying that because you're talking to me on a podcast. It's true. I love it. but I look at that and I'm like, okay, I see, you know, to your point, why somebody on our team wants to be a data scientist or maybe has aspirations beyond just a traditional it path. And then there's people who want traditional it path. And that's awesome too. So I think it just a forge you opportunities to do that now.
Ari Kaplan:some people like for the Oracle DBA is like their pay grade, like doubled their salary. And I think in data science, it is, different compensation and different demands. Especially in these trying times, but touched on, you know, one word I like, which is fun. So what excites you about it?
Robby Dally:yeah, I think it's, challenge and the problem solving. So. when I was in it, I did a lot of work with sequel. Right. I've kind of got in there and I really got into SQL love it. that's where I developed my data chops. And you know, for me, like getting, not a traditionally trained data scientist, but definitely somebody loves data and loves data wrangling. And, and so that's where I get a tool, like a data robot liberates, a person like me because I don't, I don't have the bandwidth. To go learn Python. Like I probably should. but I don't, and, I can use data robot instead and I can do all my workflows and Altryx prep my data and really that's where the strong data part comes into it. And I love, I love the data wrangling component, right. That's really fun. And it's super frustrating sometimes, you know, but also very rewarding when you crack it. And, with data robot, so much of it is just get your dataset. Correct. And then get in data robot and let data robot do its magic. So that's what I love, getting in there and messing with data. You know, if I look at my calendar and I have like a day full of meetings that I kind of sad, I'm like, Oh man, I don't, I don't get to go in and mess with, my data.
Ari Kaplan:Totally get that progression from being that data Wrangler. And being an intellectually curious person and going, Hey, I'm wrangling this data with Python or other tools. what does this data mean? How can I use it?
Robby Dally:And I think again, that's kind of like, the curiosity that comes along with it too. Right. So if I think about how I want to build my team, somebody could be a really slick programmer. But if they're not really like intellectually curious about it, I think for me, that's kind of a minus, right? So I want somebody who, who gets excited about it loves just kind of exploring the data, exploring the outcomes and the team might build. I'm super thankful that I have everybody I am because they're all really just, they love it, right. it's not just a job to them. they look at it as something that is really fun. They love exploring the data. it's definitely not like a nine to five. It's say, look, we're going to get the job done. And I'm not even mad about it. Cause I'm I'm not saying we've worked like 80 hour weeks or something, but like, you know, it's fun to explore and you know, there'll be nights when I'm on just like, God, I gotta jump on and go solve some Alteryx problem because I'm not really the person who wants to just be sort of sitting there directing everything and not involved in it. Like I believe. if I'm going to lead a team, I need to be able to at least do some of the work and understand what they're going through and, help troubleshoot. So it was, you know, we were doing a join together at the end. We missing 200,000 records. So let's go look at the workflow and figure out did we have a joint that broke and we lost records and look, anybody who's used Ultrix has felt that pain before. Or SQL even. Right. So, again, it's like the joy of curiosity and that's just, what's so much fun to me and I, I encourage people, you know, like find a job that you love and, confined joy and curiosity, and, a story to that. Right. there's school near my house, actually my Alma mater, TCU, and I go talk to their analytics program sometimes. And, My talk has changed a lot over the years, but now I just love it because not really come so full circled analytics and I'm doing machine learning and it's sort of cutting edge. And, um, I just really, I hit that point. I'm like, find something you love. And if you love analytics, that's not a bad choice. that's only growing right. And getting more and more useful in the whole data's the new oil is an overused cliche, but it's true though. Right?
Ari Kaplan:one of the things love about al-khan I know some of the listeners may not be familiar, you know, LASIK surgery, but it might be helpful to. You tell us a little bit more about al-khan and what offerings you have.
Robby Dally:Yeah, no, sorry. I should have started there actually, because, I think in reality, most people have probably been touched by an outcome product. Or had somebody in their family who was so two main areas of business. There's our surgical side, which is a cataract surgery, refractive, LASIK, or vitreoretinal. So like trauma to the eye, back of the eye type surgery, very serious. cataract is, of course they, you know, somebody who's had cataracts, they get their lens replaced. We're a leader in that industry. So there's our surgical side. And then there's our vision care side, which is contact lenses and then dry eye and ocular health. So. you know, we don't have glasses, of course. but we had just about everything else as it relates to the eye. And now, like we even have some over the counter products for dry eyes. Patanol for example, so if you have something going on with your eyes and you're not blessed with like 20, 20 vision and you've worn contexts, you probably have heard of an outcome product. You got a parent who's cataract age, or you, you know, you yourself are Cataract age probably got an outcome product. So. A very large eyecare company. And honestly, that's. Yeah. I think back to that, that Simon Sinek start with why like working for a company like Alcon is really easy to get behind because you think about your census, right? I think if you asked most people, they're going to put vision number one, your vision is so important to you. So to be able to help people with their vision, it doesn't hurt when you think about getting out of bed at the beginning of the day, right?
Ari Kaplan:Yeah, totally. I heard that not just for humans, that some of the people use it for their dogs or cats.
Robby Dally:Yeah. So, you know, it's, it's funny you say that. several years ago my dad found a stray dog and the little dog was in pretty rough shape and my dad took care of him. He's great. Now he's really happy, but he actually, he had cataracts. So. The little guy got cataract surgery. And I don't know what model of lenses in his eyes, if it's Alcott or not, but it seems to be a happy little dog who can see. So that was great. And then the other funny one is my little dog Pocco, who's running around here somewhere. He actually has really bad dry eye. So, we have to give him cyclosporin, which is a, dry eye drop that stimulates his eye tear prediction is ridiculous. It's twice a day, I have to give him drops. And then he also gets sustained, which is. One of our products, so it lubricates his eyes. but he's, uh, he's a good little dog, so I don't hold that against him too much.
Ari Kaplan:we appreciate it. You spoke at the date of robot AI, experience event. And, uh, you said, Alcon as a company and you, as your department kind of like a sports team or are you a big fan?
Robby Dally:I am. Yes. Big, big time fan.
Ari Kaplan:Yeah. In what way you see yourself as a team?
Robby Dally:Interesting. So, yeah, I think the analogy I was making there was that, like a sports team or like an athlete, we were actively performing repetitions of the same thing. Right. And I think you know, just for context, for folks that didn't see that top, We have a, like an attire, a mega project that we run once a month. And it consists of on our last run, actually 600 individual data robot projects, which are all sort of handled differently. And then we run this pretty Epic stitching process where we put it all together. it's, it's pretty complicated, like in how all the pieces work you've got Altryx with Python in the middle with data robot and then Python again, and then Altryx again and then AWS data Lake, So. My analogy there. It was like, like a good athlete. What we're getting our reps in. Right. It's just a, I'm a big basketball guy, basketball and football. So like, this is like, yeah, I know you don't want to shoot a hundred free throws a day, but if you shoot a hundred free throws, you know, you step up and it's crunch time, you got a better chance of making them, right. Because it's just muscle memory. So that was my analogy. And I've got people, my team who, who likes sports growing up. So it's an easy one for them to get behind. Although they're, probably get tired of me saying we got to get our reps in.
Ari Kaplan:good way to think about it. I'm a huge sports fan. people hear me talk all day and night about it. So. Being a team, you know, you're there to help win, as the sport and then what you're doing, you know, winning with 600 projects you're talking about, but you know, making each one of those wind by better predicting something or classifying something, setting a price or a marketing. So what, what are like some of the key uses of, either data, robot or AI in general?
Robby Dally:we're using it for automating a lot of our forecasting processes. So, you know, the one I talked about data robot AI virtual conference was our sales process, but there's certainly other areas of finance where we can automate forecasting or, you know, manual back office type challenges. So we're actually, getting through a series of forecasting. Solutions, and then we'll be moving on to figuring out sort of what do we tackle next? And, you know, there again on that 600 projects, those are all time series. So there was a very unique and challenging project in itself. And like to think we're pretty darn good at time series now. And, going back to the team analogy. Um, we broke, broken into pieces, like, someone has this, somebody handles R Python, you know, somebody handles our. Altryx flows. And then somebody handles like integration with our destination system because we get our processes end to end automation and we load into an existing legacy system. uh, like any good team, we kind of divide and conquer there and we trust each other that things will work well all. And, and they have right. I mean, again, you have to have things pretty ironed out to, to run 600 projects. each running two autopilots with, forecast distance, average blenders, you know, among other things. Right. And then have that all processed in 12 and a half hours. So, we're kind of a well oiled machine there and, Yeah. The other good thing too. Right. And this is something that I think is kind of gets lost is that we went through all this hard work to get to where we are now in this project. But we've got all these learnings. Now we can apply those to other projects and we've dramatically reduced time to implement even just like on, the part where we take the data, parse it, send it into data robot, pull it back. Right. That's all like a really impressive Python package. And, time series is so nuanced. Right? And so we've learned all these rules. So if we were building a new time series product from scratch, we would note little things like can't have a negative. If you want a zero inflated model. That's just one of those things you just don't learn until you go do it. Right. can't read books for everything and know every rule without just going in and getting your hands dirty.
Ari Kaplan:Yeah, totally. And you learn on the fly, you learn by actually doing something, you know, so sometimes we see people trying to do complex data science projects kind of from the beginning You know, just in general, most data science projects don't succeed or are extremely challenged. so it was kind of the balance of complexity. do you have the data or not? How easy is it to implement? So, what's the thought process of coming up with new ideas? Is it there's like so much pent up demand or are you. Uh, sitting in a room, being creative on a walk with your dog and it comes to you.
Robby Dally:Uh, some of the, sometimes the inspiration will strike on the dog walk. Um, actually, it's kind of interesting lately, so we. we probably have a list of like what I would say, 10 to 20 really good use cases. That's already more than we can do right now. Right. That's a, that's a lot of use cases because some of them are pretty big. this is where, you know, we've worked with our, team from data robot. you know, Matt, he helps me out a lot in, in kind of thinking through use cases and. You know, you mentioned it, which is like, all right, do we have the data? Is it value? Add, is it super complex? Right? Like something could, could be like a really cool idea and it provide a lot of value, but it's so hard. It's going to take us 10 years and it's probably not worth it when we can do something that gives us like 25% of the ROI, but we do in six months. So I think learning how to frame ideas has been really important for curating a backlog of, AI ideas. And look, the other reality is that. How come it's been around a long time, been very successful outcomes, very early in the machine, learning data science, AI revolution. And so a big part of that is there, there are more use cases I don't even know about, that are happening at other groups and within al-khan and they're bubbling up. As people start to learn what is possible. And, you know, we presented at a town hall on our recent project and I had somebody come to me afterwards and. They said, Hey, I've got a really great idea. What do you think? And I'm like, that's a phenomenal idea. And I, I never would have known, they just saw the presentation. They were inspired by the, you know, applying machine learning to transform processes and now we're going to talk to the next week and start to flush out a use case. So, you know, inspiration, we can come from a lot of different places and, and in most cases it's not coming from our, data scientists at Atkins, coming from our folks who are. Data experts and understand our problems and yeah, maybe we'll have 600 use cases before I know it, which is a good, bad
Ari Kaplan:projects. Yeah.
Robby Dally:Yeah.
Ari Kaplan:And a dozen or so use cases and yeah. Cause some of them, you, you were mentioning time series. So for listeners, no, no. There's artificial intelligence, which is the general bucket machine learning data science and time series is like one, special aspect. but, Robbie, since you've been successful at it, can you explain, you know, what is time series and then, what makes it uniquely valuable? And then, you know, kind of some of the lessons learned.
Robby Dally:temperatures is a difficult beast. I like to say, and it's super valuable. So here's the challenge time series, right? Everybody's thrown data and Excel and fit a trend line and be like, boom, time series, model. This isn't so hard, right? I mean, at that definitely done it. When you start introducing all these variables, you start introducing all these different, like unique series. In our case, products, we'll be able to have different patterns. They all have different seasonality. They all have different, orders of magnitude. And what you realize with time series is that like, Nuance matters. That's the thing we've realized, right? Like, I don't want series that have something with a hundred thousand per month. And it was something that had 10,000 per month because of the way it's going to score on them. It's going to always overweight the big ones, but maybe I wouldn't actually get a good score on the little one. Right. So there's just a ton of nuance and time series. But if, you're trying to predict something that's cyclical and you want it to be time aware, you could use OTBs I guess, but we do all time series modeling time where modeling and, the easy stuff is easy. The hard stuff is where, you gotta think hard and I think long and get creative and, um, we sell capital equipment. For example, capital equipment in a smaller company is pretty infrequent. In terms of how often we sell it and that's a challenging one to solve. So like, how do you handle that? Or how do you handle a new product launch? Right. So time series, machine learning is inherently using past observations to make future predictions. Right? I have a new product that's launching, I have zero past observations to make the future prediction. So how do you handle that? and look, we haven't solved all these, by the way. We're still working through them. And, you know, with the scale where we've gotten, we've been really lucky. We've got a lot of input from. The data robot team, um, even some, uh, like the, super users on time series, like Jay Sherman, right? we joking in the calm to God, a time series and he's helped us solve a few problems and given us some really good ideas that we've implemented along the way, and, regularly with what Dave, our CFDs. And we just keep iterating through it. Like every time we write it, we discover new stuff. So that's why I see time series is challenging, but also super rewarding because. You learn more, you get better at it and you see the value of it at the end of the process.
Ari Kaplan:with time series, we talk about, regime changes in time. So how has like COVID and that total shift of purchasing patterns and health patterns how's that affected your
Robby Dally:Yeah, look. Yeah. Uh COVID right. I mean, it's probably can't have a single one of these without mentioning COVID I've got all this stuff that is super consistent and stable and seasonal. And then I introduce, the shutdown of elective surgeries in the United States. Right. And for two months, all of these patterns are now destroyed. So what do you do about that? we fund some work arounds throughout that, where we blended. So what we call hybrid models. And we, we look at our pre COVID runs with our post COVID runs and, we blend these things together based off of what we think rate of recoveries look like. And, and that's where you're kind of infusing definitely a little more art than science, but we've just, we've got to get past COVID. because honestly, like, does anybody really know what's happening right now with COVID? And like, we think we know, but. Things could change on a dime, right? get, I go back to the elective surgery. So if you think of cataract surgery, that's an elective surgery. You can certainly put that off for awhile. So if the state, let's say Texas and California and Florida shut down elective surgery for two months, like the procedures are gone. But the flip side of that is, it's a procedural driven business. People don't stop having cataracts. So at some point they'll come back. So. That's where we are right now. Right? Like how do we blend these things together and understand the nuance? And when do we go back to a world where it's sort of back to normal, and this is what we've talked a lot again with, Dave is like, how do we handle that kind of cutoff? And at some point, do we just sort of take the COVID months and slice them out, set them aside and say that was a bad trip. let's not look at that data anymore because it's just so. Disjointed with everything else. So like I said, we don't have all the answers yet. We've found pretty good solutions, but every month is like refresh the models and let's see what happened.
Ari Kaplan:Yeah. Yeah, Maybe nobody has all the answers. Maybe nobody has the key answers, but it's a bounce one direction. And then as people start coming back to elective surgeries, you know, maybe it's different slices and segmentations of the public. for a lot of our customers, it's like the challenge of. Knowing what data drifted, and by how much and what data did not drift. So I know you don't do everything, but do you put some thought into understanding what's drifting and what's not, and is it a monthly refresh or a three month
Robby Dally:a good, it's a great point. Um, it is a monthly refresh, so w we kind of think of it along like area of the business and, therapeutic area to patients. So like, look, somebody that's dry eye, dry eye doesn't stop because COVID has happened. So there's still no pride by the dry products now. Do their behaviors change for a while? Do they go pantry stock? Right? Kinda like, the run on toilet paper and paper towels and Lysol wipes. Does everybody think, Oh crap. I got to get my dry eye drops because what if there's a disruption and I don't want to deal with that. So yeah, you have something like that. And then you have like vitreoretinal surgery I mentioned earlier is, usually trauma or like a serious eye condition like that stuff. Can't wait, that's not elective surgery. So. That's more or less going to be, well, it's going to be less impacted, right? There's always, of course still going to be impact, but certainly would be less impactful. cataract surgery, you know, it's purely elective. So what happens there? And then, so that's one country. Now let me order of magnitude this around the globe, because we do a global forecast. So we've got South Korea on the one hand, pretty much unaffected, right. Or, or relatively unaffected. Italy shut down for a while. It's just like the U S so did the UK, but then Germany was last in fact to see, you know what I mean? Like it gets complicated and that's why every month is like, okay, let's, let's see what happened. Let's adjust as we need to. And then hope fingers crossed. Hope this thing goes away, uh, at some point where we can get back to normal and, go from there. Cause it'd be much easier. I want it to go away for many reasons. One of which a very low importance, one on the global scale of things, is it what just make time series modeling so much easier?
Ari Kaplan:Yeah, I love talking about time series, so great. We're bringing that up. yeah. Any, any other, like, how do you quantify the value of, know time series insights?
Robby Dally:so the way we're looking right now with this project, we just rolled out and several more that we have coming is we're actually looking at it kind of from a continuum of efficiency and effectiveness. And we've been harping on this a lot. So are we efficient and making a forecast? So if I tell you, we go from a process, fully manual Excel based, country involved around the world. And we go to one where we, still have the countries involved, but instead of doing like a bottom up build and Excel, they have like a base level forecast that they tweak. What I would say is one of those is much more efficient than the others. Now that efficiency doesn't matter. I'm sorry. Actually, I w. Kind of, it has to go along with effectiveness. So we have to have like a pretty good forecast as a starting point because the forecast is trash and they spend twice as much time fixing it, it wrong forecast than they would of just building a brand new one. We failed the efficiency criteria. So it really is a marrying efficiency and effectiveness. And at the end of this, right, and this is interesting. I think people can probably relate to this who are early in this journey. Like. I framed this as, like, this is not an and full disclosure. People who don't know data robot that well, and how can they hear data robot, but then inadvertently, they ended up calling it the robot. So I framed it. I'm like, look, this is not human versus the robot. Right. What we want to do is we want to put these together. We want to take out the manual, work through automation and machine learning. Let you fine tune, add, add knowledge. And we get to something that's even better than we get standalone either way with much less work. So that's where those things come together. That's efficiency and effectiveness. And I don't have like perfect KPIs to give you right now because we're pretty early into it. But, that's where we're going to be. And I, people, honestly, I've been very happy. Like people are excited about that and I don't know why I thought they wouldn't be, I don't, I don't know whether I had a thought one way or the other. Like, I think people always would get concerned if you're sort of changing their job. Automation and machine learning, but people have been very excited They're eager to help improve the forecast, which is a again that will only make things better as we go forward.
Ari Kaplan:Yeah, totally. And then when people learn, it's automating the mundane repair. Additive, you know, we brought up the word fund before, but non-fun parts. Okay. The job then, you know, there's buy-in and in the end, you, can greatly accelerate how you produce insights.
Robby Dally:Yeah. And you know, actually you asked me a question earlier about, are we talking about like, it person wants to be a data scientist? I think actually, because we're in finance, right. I don't know that most machine learning groups are in finance or the finance has machine learning groups. Although I suspect it's probably a bigger number than I think, but, you know, what's, what's been interesting to me is I've seen a lot of finance folks talk to me and say, wow, this is super cool. I would love to have access to data robot or, or help you improve the models and give you more data. And let's think about upskilling our finance community. Right. There's this whole automation component where we have a data Lake and we have X, right. But then there's the concept of taking manual processes and put them at data robot. And I've been really encouraged to see people respond to that so positively and want to, you know, I mean, there's nothing wrong with being a traditional finance person, but be a finance person who maybe can dabble and citizen data science components. So that's been really, really exciting for me to see that, start to emerge and. I was talking to somebody the other day and they said, yeah, I'm actually taking Python right now during the pandemic. It's like, this is just like a normal finance person. They're doing a Python course. It's awesome. So if we inspired that at all, I think that's amazing. And it's only good for the company. Good for that person and people who are doing that. Right.
Ari Kaplan:You personally are inspiring. A lot of people, they want to be like you, and you know, let's take those Python courses. Let's take data science let's let me become a citizen data scientist. the title of this podcast is a more intelligent tomorrow. So curious, five or 10 years from now, we were having a conversation. What are we going to be talking about?
Robby Dally:Wow. Yeah, that's interesting. I've been so in the here and now that I haven't, I haven't considered where some of that is going. Right. So like, Kind of like on a personal side of things. I'm a big Tesla guy. Okay. and I look at that and, and I look at how hard it's getting to get to full self-driving. Right. but at some point they will, And then the world just changes. Right. What's a car can drive itself, you know, and again, I'm, I'm done my Tesla rabbit hole here. Right. So self-driving taxis, for example, The world has changed fundamentally. So I don't know for sure what it looks like, but I do know that there's a lot of smart people who talk about what AI and machine learning could mean and sort of my higher level, like 10, 15 years type things. Like how, when AI starts to becoming incredibly smart versus sort of narrow AI that we have right now, in our world though, like if I kind of zoom out of the conceptual and. The world of Ilan. If I kind of come back to alkaline world, you know, like I really hope five to 10 years from now. We've got just like tons of machine learning, data science implementations happening, and we've reduced a lot of mundane. Like you mentioned, early mundane work and replaced it with work. That's just more impactful and, and value driving. And you know, it's going to change the company. I really, I believe it'll change the company, many companies around the world and, Um, personally, I'm just really excited that my company will be on the forefront of that. I believe versus, you even like where we were two years ago, we weren't in the world of AI and machine learning at all. Really? And yet here we are now diving head first, so good. I suspect most companies are, are in that same path. And, um, yeah, and if you're not, that's probably not going to put you in a very good competitive situation.
Ari Kaplan:Yeah, so, I love Tesla and the whole concept of self-driving and automation, both in. Enterprise artificial intelligence and then kind of non-business uh, have you ever been in a Tesla?
Robby Dally:Um, yeah, I actually own a Tesla. I have a, four year old model S and I, I love it. I look, I love the concept of electric vehicles in general. super fun to drive. And I am fascinated by the world of, computer vision and artificial intelligence driving the car. And yeah, I think that the societal impact right of the cars could drive themselves how much safer the roads could be so there's, there's job transformations, of course, that go along with that. But the world is a safer place and That's an amazing idea, right? That you could change the world and make it better. And that's sort of like the moonshot projects, right? So we were talking enterprise AI and where we're at, and it's a lot more, tackling, right. To use the sports analogy. It's sometimes it's not glamorous, but it's the stuff that enables you to do the really cool stuff and the glamorous stuff.
Ari Kaplan:Yeah. As kids who would have thought, you know, self-driving cars are here. And all the other technology. Would you, do you feel comfortable now know, sleeping a long haul and some self-driving car?
Robby Dally:I'm probably not there yet, unless the road is perfectly paved with perfect lane striping. but look, I've done some road trips and I actually, my, my car being a little older has the older version of autopilot. But like, I can, you know, if I go drive the doubts, for example. Yeah. I did that yesterday and I think I drove one mile out of the 25 miles that I was on the highway and that's because I was going through a construction zone. So, you know, a lot of times they talk about sort of the, curve of innovation. So like early phase one or stage one technologies they have their bugs and flaws, but then there's this big, huge jump that happens. Yeah. Anyway, that's you kept me, you're on a subject I love, right. Which is Tesla and self-driving cars. And I could go on for awhile. I don't think we have enough time for that one.
Ari Kaplan:Yeah, no, it's super fascinating. And you know, one of the themes we hear around artificial intelligence these days are like trusted AI. So like the Tesla, you know, trust at this point, you know, to sleep for three hours in it. Um, but you know, on a more, you know, day by day basis, There's no trust or lack of trust and insights. So, what challenges have you faced with AI trust? Like what either for you personally, or people you're trying to convince makes it.
Robby Dally:I think where you run into that is where you have like very knowledgeable, Business users, right. Who really understand their data. And, common trap you run into at first is you make a prediction and they're like, well, yeah, I could have told you that. Right? Like this, this and this. It's a very obvious thing. Uh, the flip side of that I would say is, look, if the machine learning doesn't get the obvious one. Right. Well, okay. Maybe there's a reason. It didn't because it wasn't so obvious, but actually that's kind of proof of validation You know where we are again on this journey is I think about some of our upcoming projects. our goal is not to be perfect machine learning and answer every single observation. Like what we really want to do is, pick where we've got like a thousand observations that have come in and that would have to all be manually reviewed. And let's use AI to cut that down to 150, that truly required investigation. Right. And. to your point, you got to get to the trust side of that. I think a lot of times you, you run into like a parallel process type situation to build that trust. And then again, w we're early, right? So we'll see how this pans out, but I think as you start to stack up, examples of machine learning, doing a really nice job and, proving value, I think You know, you build that credibility up and that inherently increases the trust, right? As you get, it's just not shocking. I mean, somebody's not going to trust AI if they feel like they understand the business and they do a good job at it already, Like I don't need that. But then again, that's part of the whole value proposition. That's actually our project we launched. We started that project by talking to value proposition and we showed a stack chart that was like, time-span up to a hundred percent on average. And like 60% of it was spent on forecasting and 40% on, business, uh, partnering with the business. And we said, what if we could take that instead of 40 to 60 to 70% of your time partnering business and 30% forecasting or 80 and 20, right. We don't know what the exact number will be. We'll figure that out, but we're going to, again, kind of flip how these things go and that value proposition makes them want to trust it. So then you go, prove it, show it works well. And now you have advocates. I think.
Ari Kaplan:Yeah. Makes sense. Yeah. One of the things, as you're talking, I was thinking is, Al con was it a spin-off of, you know, much larger company? Um, I dunno, the sizes involved, the big versus the other, but do you see like how people implement data science differ, you know, much larger companies or maybe multi-business line companies? Versus a smaller focus company.
Robby Dally:I'm a little out of my element only because I get my experiences mainly just alcohol, right. But I have seen, just being around these conferences and hearing other folks talk like, you know, a lot of times what ends up happening. We talked about this a lot. Like where are we evolving to an icon in terms of a structure around governance for AI and machine learning, data science. And I think it kind of depends on where you evolve too. But I think a lot of like really big companies, they probably end up with sort of like functional area machine learning groups, right. Just because of the sheer size. And a company like outcome, which is a pretty big company, but, you know, in our CEO calls it like a, multi-billion dollar company, that's a startup. Right. because of that, we're, we're pretty flexible. We're pretty nimble. And I think we have the opportunity to sort of bring all those ideas together. Again, it's kind of bubbling up organically throughout our organization, including, you know, the, the group that I lead the finance team. So. I'm not sure this is quite the answer that you were looking for, because I just don't know that I've seen it a lot, but I, I think it just depends on how quickly you can stack up wins. And does the leadership team have a big vision in it? So we're really lucky in that our leadership team at Alcon believes in this, they believe in analytics, they believe in, applying data science to problems, and that also helps provide credibility when you have it from the top down.
Ari Kaplan:Probably thank you so much for being a guest on a more intelligent tomorrow or indeed a robot, podcast. really appreciate going over huge, a variety of topics. It's been a lot of fun, so much appreciated.
Robby Dally:Yeah, thanks so much for having me. I really enjoyed it.
Ari Kaplan:Thank you.
Ben:okay. So I think don't close