Kloud9 CIO Webinar Series Replay
Dion Rooney: The Future of Enterprise AI
Dion: Hi, Sami, how are you? I'm
Sami: really good. We'll, we'll start in a couple of minutes. Great.
Sami: It's kind of magical that things actually work straight away instead of the required 15 minutes of.
Sami: Can you hear me
Dion: now? Yeah. You know, all too common expression over the last couple of years.
Sami: So true. So true. I always thought that pandemic lasted almost three years and, and we still had a hard time getting everything working.
Dion: Yeah. Well, my favorite, one of my favorite expressions is you can, how do you train a two year old better than he can teach people?
Sami: And I think the, one of the big problems was always, especially with the enterprise, users was the firewalls and the, low level network security because all of these zoom and everything else requires different end points and ports.
Dion: Absolutely. It was a different, different world for sure.
Dion: But I do marvel at how quickly people got it.
Dion: There were still those that, that were challenged but, you know, people adapted so well throughout so many curveballs that came through the pandemic.
Sami: One of the, fabulous stories I heard during the pandemic was somebody threw their 70th birthday on on, on one of the virtual events platforms and practically everybody made it and it was like in, out of L A and, and all around the US.
Sami: So I thought it was fantastic.
Dion: That is great. Like I said, I think people were very resourceful and adaptive.
Dion: It taught a lot of people, a lot of lessons and lots of opportunities came from it and some of them, you know, persisted.
Dion: So that's really good.
Sami: Yeah. Absolutely. OK. It's let's see, we have 20 people on.
Sami: So let's maybe give it a minute or two. And then, then we'll get going.
Sami: We had a lot of registrations.
Sami: But what I've, I learned in the, during the pandemic as well is that a lot of webinars, people prefer to Netflix them, which is to register, get the recording and then grab all the best parts.
Sami: So we'll probably deliver this as well to to the people who registered but couldn't make it today.
Sami: Sounds great. Awesome. Awesome. Awesome. All right.
Sami: But hey, let's let's get going. So, I'm a Sami hero.
Sami: So I work for cloud nine and I have the absolute pleasure to introduce Dion Rooney, who's in fact, this is our kind of inaugural cio webinar series.
Sami: So we're going to have more to come.
Sami: But we're, we're starting with the best and then the rest will have to figure out how to outdo you because everything's always a competition, right?
Sami: And and, and it's, there's a lot to say about the ana let you talk about it.
Sami: But I think one of the, the really super cool things for me personally is to talk to him because I, I used to love to go to Toys R US with my Children.
Sami: when it was still a huge thing in the US.
Sami: But you've been a cio for a really long time.
Sami: So maybe you can not talk for a couple of minutes and, and tell us a little bit about your career.
Dion: Sure. Thank you and, and thank you for having me. I'm thrilled to be a part of this.
Dion: It's a certainly a passionate topic of mine. So thank you for that.
Dion: I have been a cio for a long time. You could tell by all the gray hair.
Dion: So, but many of those years were at toys.
Dion: and you know, toys was an interesting experience because it was a global company.
Dion: They had as many stores, if not more outside the US as they did inside the US.
Dion: And, you know, I think we all had those fond memories of either our kids or ourselves kind of growing up, you know, having that amazing toy store that you could go into and just kind of eyes open wide and, and see all that.
Dion: And then, you know, some of the nuances around both retail and I think to an extreme toys, which impacted kind of the, the technology role was the seasonality of the business.
Dion: And although every retail company and most companies say they have some degree of seasonality, I don't think many of them rival kind of what we saw when, when toys was during its heyday, you know, going from 20 week to 200 million in a day is quite an overnight change and, and volume and it does have implications on all functions including technology for sure.
Dion: But I, you know, I went to Hudson Bay and you know, Hudson Bay for those that don't know, you know, Saks Fifth Avenue off fifth at the time, Lord and Taylor Guilt was an acquisition as a startup.
Dion: and also Hudson Bay in Canada, which is a large department store chain.
Dion: And so lots of retail experience and then, you know, my last gig was in manufacturing at a company called G A F who makes roofing materials.
Dion: So lots of lots of relevant I think, experience and interesting experience.
Dion: And what I loved most about my career is, you know, it gave me the opportunity to learn continuously through those experiences and, and hopefully we can, we can share some ideas today and continue some of that learning.
Sami: That's That's fantastic. I actually, I didn't realize I used to live in Canada at one point.
Sami: I, I obviously know Hudson's Bay, but I didn't realize they own all those other brands.
Sami: So that's, that's fantastic. So obviously, you know, in retail data is absolutely critical and, and I guess a, I was a little bit in the early days if you can call it that maybe more of a rules based promotions and things.
Sami: So how, how do you see as, as you've kind of moved in these different companies, the role of data and, and headed into, into this world of A I now.
Dion: Yeah, you know, it, it is, it is incredibly interesting to see different organizations.
Dion: I love the expression big data because, you know, most people don't quite understand what big really looks like because everything is relative.
Dion: but there are so many use cases in retail and in, in manufacturing where data can, can provide such incredible insights, you know, things like the purchase path and, and, and you know, what, what's driving a consumer's behavior and how you can help them purchase more effectively and easier.
Dion: So, you know, the the the basis of all that data and how you can take advantage of it.
Dion: And, and with the technologies advancing and the compute power being cheaper and all of those things are converging to give you a much better, you know, view of what data can do and, and really how your business is operating.
Dion: you know, that I use that example of purchase path, but that path of kind of touch points and filling them in for either attribution purposes or the like is incredibly interesting and important.
Dion: You know, there were some experiences I had when, when I was at Hudson Bay, you know, where it wasn't quite a i by any stretch, but a lot of image rack and machine learning and an anticipation of what could help a consumer either find what they were looking for or make relevant recommendations.
Dion: So, but it, it is, it is an interesting field, particularly around both retail and manufacturing, I think
Sami: and because it's the data, like you said, it's it's underlying everything, right?
Sami: So, but I, I just, I was talking to somebody the other day and, and they were saying that they only this year they got their data lake done.
Sami: So essentially uns siloed their data and the company's been around for a long time and these technologies have been around since forever, so to speak.
Sami: I'm also gray and I was in one of those startups, data warehouse startups way back then.
Sami: It's but why do you think it's still so hard to get to the relevant data?
Dion: I think, you know, for me, many companies are probably not as well positioned as they think they are to take advantage of the data and you know, it sounds so attractive.
Dion: We've got data, how do we leverage our data?
Dion: But that data is likely been accumulated for different use cases and purposes, right?
Dion: It's to do a back office function, it's to accomplish some process.
Dion: And when you look at it from a different vantage point and try and do analytics that have not been anticipated in the past, you typically find some gaps and, and you know, so most are not well positioned to, to do that.
Dion: And you know, once companies start trying to leverage their data, I think it's analogous to a journey and, and a journey that people take without a map, right?
Dion: So there, there's no, there's no blueprint that tells them how to get where they're going.
Dion: But what I love about it is, you know, for each question you typically answer you, you, you find three or four more insightful or deeper questions to ask which then you know, cause you to either make a right or a left or go straight on your journey or turn around as the case may be.
Dion: And I think you see that so frequently with organizations that things they think they're data rich.
Dion: But because they haven't used it for the high end analytics or A I it it, they don't know what they don't know which is there's likely gaps or cleansing necessary around dirty data.
Dion: And, and I think even that topic itself, companies tend to take shortcuts to, you know, cleaning their data, you know, it's not preferential integrity, it's not requirements, it's, it's quality.
Dion: And, you know, I, I remember an example not too long ago where, you know, there was a critical element of a data set that was missing.
Dion: So the answer was, well, we'll just plug in a, a standard, you know, answer to all those questions which satisfies the requirement but doesn't achieve the desired result, which is having proper related meaningful data that you can do analytics on.
Sami: Yeah, that's probably one of the reasons why people like to say that the A I that they start to hallucinate.
Sami: And I'm, I'm kind of and I, I like that term because it's if you feed them something which is incorrect, they, they will start inferring incorrect answers as well, which sound perfectly logical and, and quite can be quite biased actually.
Dion: Yeah, it, it's a great, it's a great point.
Dion: You know, in the example, I was referencing, it was, it was a date that was missing for, you know, 50% of the, of the elements.
Dion: and plugging in a date is, is dangerous because as you said, it's not, it's not meeting the requirement.
Dion: It's going to actually give you likely the wrong answer to some of your questions that you're going to want answers to because it will think that's the correct data and, and you know, there's no, there's no asterisk next to it that says ignore it and, and in, in Candor, sometimes you need to ignore that data.
Dion: If you can't recreate it the way it's supposed to be.
Dion: Maybe you exclude it from your analytics or you think about it differently because the most dangerous answer is to populate it with some generic answer that will corrupt your, your future answers to the analytics you're doing out of your data lake or wherever.
Sami: Yeah, I can imagine.
Sami: It's like our production date typically is is one day or a shipment day is one or two days after we started production and then you will have a bunch of variances whi which you, you actually lose in the process
Dion: and, and, and the value in that example, you're referencing is in those exceptions, right?
Dion: How do you know you're likely gonna wanna double, double click on those exceptions to understand what's causing them, whether they're the positives or the negatives.
Dion: And if you homogenize it to a generic answer because you are missing the data elements. That's problematic.
Sami: Yeah. Now imagine when when everybody is in love now with chat GP T and all the kind of the A I variances.
Sami: So so what do you think of that in the context of kind of enterprise world,
Dion: you know AAA couple of things. One in general, I I'm excited by it.
Dion: I know there's the, the kind of doomsday fear and it certainly does make interesting, you know, news and interesting commentary.
Dion: and it, but it, it is an exciting time, you know, as technology advances, it, it is, it is always a good thing.
Dion: But having said that there are, there are challenges, you know, as you said, you know, leveraging that at the enterprise level, I think people are starting to take advantage of that.
Dion: Companies are certainly starting to take advantage of creating offerings to that because it is clearly the the buzzword right now is, you know, my product is, you know, chat T P T enabled or whatever.
Dion: But you know, to me, it is completely an exciting time.
Dion: I think that press references both the upside and downside of chat GP T and, and there's probably more myth than reality around it.
Dion: But things like, you know, rewriting marketing material, you know, doing students homeworks, right?
Dion: So, you know, everyone's fearful that our kids will be stupid because they're, they're using chat GP T to do their homework.
Dion: I go back again. I'm old.
Dion: Yeah, we had an encyclopedia set of encyclopedias that we used as a source.
Dion: It's really always a combination of, you know, the analysis people need to do and the data sources that they have, you know, one of my favorites is the example of blog creation, right?
Dion: So you can actually, you know, leverage things like chat, chat GP T to create blogs, but you still need to edit them.
Dion: You still need to do a bunch of work.
Dion: You need to, you need to, your job may change and the roles may change.
Dion: But if they don't go away and they still need some care and feeding one of the examples I'll give you with a, with AAA friend.
Dion: Is he literally connected a Google sheet with chat GP T?
Dion: So he could help filter out potential choices for his son to select AAA college to go to.
Dion: So it was an interesting use case where he had parameters and, you know, kind of filled in and again, it's a, it's a little bit of both and I think, you know, the fear mongering of, we've got to stop this, we've got too much going on where they're gonna take over.
Dion: You know, I don't think you can govern or, you know, prevent or, or ignore technology advances, I think every time you do, you're just creating more and more challenges and, and, you know, to think and, and I'm not gonna get into a political conversation but to think that the, the federal government can help us.
Dion: And, and the politicians in Washington can help govern chat GP. T scares me more than chat GP T.
Sami: Yeah. We used to joke way back then when somebody lost emails is that let's just go and check from NSA.
Sami: They have copies of everything. So Yeah, there is, the control and go governance is different from control, oftentimes.
Sami: And, and it's actually, it's really interesting because it's, yes, GP, t is one of these large language models and, and right now investments are going to a bunch of these other ones as well.
Sami: And there's an interesting one called Anthropic.
Sami: I don't know if you heard about that but it's, it's really geared into, supporting with customer service.
Sami: But they just, why it's interesting is that they, they are really big on pushing for, them not being biased in any way and being kind of honest and transparent and everything.
Sami: And they just wrote, raised 450 million in funding yet another round.
Sami: So it's, who, whatever people are saying about the, the V C community right now, it's a I is still collecting a lot of money compared to a lot of the other companies.
Dion: Yeah, I think, I think you're, you're right.
Dion: And I think that use case is a perfect use case, right?
Dion: We've all, we've all been on that, that death March of a phone call waiting for customer service.
Dion: And e e even the ones that have the chat bots, they're not very effective.
Dion: It's a huge opportunity to get answers to, to the questions many of us have in a more effective and efficient way there.
Dion: Again, the roles will be different right.
Dion: Someone's gonna have to teach you the answers, someone's gonna have to deal with the answers that fall through the, the kind of use cases or deal with the, the questions.
Dion: But those, those investments are great because I think that there's so much promise with this technology.
Dion: And I think just finding the right use cases and getting people started on that journey is what's going to help companies and organizations take advantage of it.
Dion: And you know, those that, that pretend it doesn't exist are gonna find themselves behind a a relatively long steep learning curve.
Sami: Yeah. No, absolutely. Right. So when we talk about data science, like it, it tends to be our A I it tends to be like a kind of a big term.
Sami: And now, you know, we have the large language models which help us with that kind of processing a large volume of content and a answering.
Sami: Yeah, let's say good questions, good prompts.
Sami: But then we have all the other stuff as well for benchmarking and other things. Do you see that?
Sami: What kind of balance do you see in the enterprise in the future?
Sami: Like between these two or should companies invest in both?
Dion: You know, I think I, I, I, I can't imagine it being an either or I do think, you know, the investment level may vary by organization and their maturity and what their strategic goals are.
Dion: But I think I think you need to do both.
Dion: You need to continue to run your business, you need to continue to invest in some of these newer capabilities to see what you can learn from them.
Dion: But I think it is it is a both answer you, you need to do strong analytics and, and even if you develop through L L M or something else, this this hypothesis, you know, you need to test it, you need to assess it.
Dion: I'm a big fan as I think probably most of the people on the call are, you know, measurement is important.
Dion: Data is important and the analytics that go along with it.
Dion: So I think if they go hand in hand,
Sami: yeah, that reminds me, I, I used to work with a CFO who used to say in board meetings is that in God we trust and others bring data which is also a little bit tricky because, you know, sometimes we have this confirmation bias, you know, so we go and try to rig essentially our questions in such a way that we get the answer we like or that fits our past experience, see that happen as well?
Sami: Or, or is it just me? Yeah, I
Dion: think it's in part the, the I'll call it the personality but the, the, the organization, you know, some organizations are very relationship focused, some are very data and analytically focused.
Dion: But I, I think it's one of the reasons why when you do a test or you have a hypothesis that you kind of define and it sounds so basic but define what does success look like before I execute the test?
Dion: Because you do find that, that confirmation bias where people attempt to rationalize or explain a way what they, what they've seen.
Dion: We did, we did some work with process modeling and you know, it showed this crazy behavior around a AAA circumventing process that was being executed and, and everyone just rationalized it.
Dion: They were like, yeah, that's because we do this.
Dion: Well, should we do that is a question that's harder for people to get their heads around and it's that infamous.
Dion: Well, we've always done it that way, right? The answer to some of these crazy questions.
Dion: But yeah, I think, you know, people don't like to be wrong and sometimes data prove you wrong.
Dion: I'll give you an example from my retail time, online ordering.
Dion: You know, could we give you a range of when your order was going to be delivered or should we give you a specific date?
Dion: And intuitively, the answer we all came up with was a consumer would love to have the specific thing that, that something is gonna be delivered.
Dion: So of course, we, we did A B testing and determined they actually hated it and they didn't like it because it was later than they wanted.
Dion: So the range even though it wasn't as accurate, gave them some hope that it may come earlier.
Dion: But who would have thought of that when, when you're kind of coming up with these tests?
Dion: Now, the test proved that the range was better. We stuck with the range for some period of time.
Dion: And again, this was a while ago, you know, shipping has gotten a little more sophisticated but it was counterintuitive the results and, and, you know, we talked about whether companies are, you know, emotionally or relationship based or data and analytics based, even if companies are data and analytics based, they need to be comfortable to acknowledge the fact that their hypotheses were wrong in order to iterate appropriately to get to that right answer.
Dion: And I think those are all kind of parts of the maturity of an organization that companies have to go through.
Sami: Yeah, that, that's, that's actually so true because I, I had a personal experience with that a couple of years ago, I worked in a, a company as well.
Sami: And and we build a predictive model for renewals because, you know, it was a business.
Sami: We worked that was pre pandemic though.
Sami: So we worked a lot with retailers around the world and figuring out like, you know, which ones are more likely to turn and to stay and so forth.
Sami: And I had a very clear hypothesis on what are the variables that will drive.
Sami: the retention or not. And once we run the algorithms and all of our data, it, it quite surprised us because there were some completely different engagement metrics that were more of the driving factors than that we that what we anticipated.
Sami: And typically, you know, it's like, you think it's like things like usage and and things like that, but not really because it's it's not always the thing that people appreciate because again, we sell to humans.
Sami: You know, even though now, I guess with some of the, widgets, on, on GP, t you can actually tell it to do stuff for you.
Sami: And, and of course, you know, I've had a bunch of, Amazon Echoes at the house for, for years and every so often we talk about something, we don't even use the trigger word and it just announces that it added something in the shopping list and it is kind of creepy.
Sami: But, but yeah, it's, but it helps us, you know, obviously it, it empowers us.
Sami: It, it saves us time and convenience wins until, until it takes over. Right?
Dion: You know, it, it is one of the, one of the interesting dynamics retail in, in particular, right?
Dion: That whole, how, how do I, how do I give people what they want?
Dion: How do I anticipate, how do I personalize the experience and how do I allow discovery at the same time, right?
Dion: Because I think that is part of the fun part of shopping is discovery.
Dion: You know, so you want personalization yet you want some level of discovery, but you gotta just stop short of that creepy factor.
Dion: Because that's where people I think are a little put off. I hear the story all the time.
Dion: You know, my wife and I were talking about something and all of a sudden I started seeing the display ads.
Dion: I'm like, well, someone's listening, right. So anyway, I, I do think those experiences are, are very effective.
Dion: But again, those are the use cases that I think finding your data, understanding the original source.
Dion: Where should you source that data from? Is it clean? What are your hypotheses?
Dion: What data can you use to help validate or refute those hypotheses are all part of this journey that I think many organizations have taken and, and many still have yet to take or begun.
Dion: And again, it's not a, I'm gonna start and be done in six months or a year.
Dion: It is literally a journey. And I love the fact that, you know, people try and answer questions that they have today, but with some of this new capability, you can actually answer questions you haven't really thought of yet.
Dion: And that's the iterative part of this.
Dion: That's so exciting because it, it really does unlock things that you didn't necessarily think were possible.
Sami: Yeah. And you could also say that with the, with the new technologies and the new approaches in the data science, particularly like machine learning, you can address things which were kind of too small to spend energy before.
Sami: But now because it comes easier and we've been working on this ghost inventory problem with with a huge retailer essentially.
Sami: And it's it's meaningful, you know, it's several percentage point points that they are losing sales because of it.
Sami: But it's too small to address traditionally because, you know, when you have 10,000 plus stores and 100,000 plus queues, the combinations are, are kind of insane.
Sami: So the traditional approaches with with B I tools like or Power B I or others and even some statistical models are, are a little bit painful.
Sami: But once you hit it through an M L model, you can identify the individual products that are probably going to be out of stock.
Sami: Even everybody thinks they should be in stock.
Sami: They kind of conversely, it's not, we're not talking about Home Depot, but you know, the Americans, know that, you know, you go to their website, they say they have five units of something, you go there and try to buy it and most likely they have one or none.
Sami: and then they start calling around, he's like, who's got it? You
Dion: know, and, and in fact, the, the employee in that store says, if it says five, we probably don't open it, right.
Dion: They, they, so they acknowledge it.
Dion: And I think that that particular example is so important now because many of the retail organizations are using that store's inventory to ship to home as well.
Dion: So a consumer who's made the purchase is expecting it to arrive.
Dion: And you know, it's kind of like the, the analogy I'll use is someone shopping through that Home Depot store and, and ac a an employee actually pulls the product out of their cart before they can get the check out.
Dion: We so the dependency on having accurate inventory data in a, in a retail location is, is so much more important now than it was when it was hidden from you.
Dion: You still had consumers coming in trying to buy something, but you didn't know when they left without it or when they left displeased.
Dion: Now, you know, they placed an order, they expected to ship it from the store, it didn't exist.
Dion: And, and you know, we ran some analytics in my previous life where we tried to ascertain if a store said they didn't have what we were asking them to ship to a customer.
Dion: Did they sell one subsequently at the point of sale without having received additional ones?
Dion: So we could understand was that they couldn't find it or it really didn't exist.
Dion: And again, you're peeling back the layers of this data and analytics to understand what is really going on and you know, to do it for every item, for every location is mind numbing.
Dion: But to put, you know, machine learning behind it and actually understand trends, when does it occur, will help you understand what could be the root cause for when it does occur.
Dion: So I think there's so much there and I think it's more and more important for retailers and I say important but probably no more visible would be a better way to say it because I think that problem has existed probably was worse previously, but it was hidden to them because they didn't necessarily understand a consumer coming in and not finding what they were looking for and just walking out with either not making that purchase or having to substitute.
Sami: Yeah. No, you're absolutely right.
Sami: And some companies end up investing a lot in technologies like R F ID to track items and have more real time data.
Sami: It's I think the the European sports goods chain called the Decathlon is probably one of the leaders because they actually, because they also do manufacturing, you know, you could, they could do crazy things like they put the R F ID tag inside a basketball already at the factory and they can track it.
Sami: And I was recently talking to somebody from Levi's and that was a fascinating story because their in-store inventory accuracy was 70% prior to pandemic And then, during pandemic, when they were closed, they used the stores as the last mile delivery route and the pick up in store, buy online pickup in store.
Sami: And they, they built a lot of machine learning models during the pandemic and they actually also deployed them and then also stuck our ID tags and everything.
Sami: They went to 99 plus, accuracy and, they still can't tell you where on the rack they five oh one of certain sizes are, they know that they are there.
Sami: They just don't know. So they still need to go and, and search the piles and stuff.
Sami: But it's, it is really, really interesting how you, you just increase more volume of data and you increase the accuracy and then you can model behind it.
Sami: But actually, here's a, here's an interesting question for you because, we've heard from a lot of companies who's been doing a lot of kind of machine learning modeling, but they're struggling a little bit getting these things into production.
Sami: Like, have, have you seen that as well?
Sami: And do you know if, if it's like guys playing with new toys?
Sami: I'm not sure if it's gonna work or what's behind it?
Dion: You know, I, I think, you know, if I go back to, you know, image rec and machine learning.
Dion: So, one of the things in the apparel world was, you know, consumers are looking at a particular and I'll use an example, a, a dress, right?
Dion: And, and it's a floral dress. Those, those consumers might want to look for alternatives.
Dion: Now they can go back to the full product array and look at all the other addresses.
Dion: But they could also say show me similar items, right?
Dion: Because I want something like that and it's, it's traditional, you know, image rack, machine learning attribution that can be fully automated.
Dion: I and I think those things are, are pretty straightforward, you know, you can test them, you can validate them, you can do a lot of things to make sure you're on the right track.
Dion: But then you get into some of the interesting nuances that we also found around environment appropriateness for a particular garment.
Dion: So, is this a appropriate dress for a cocktail party or is it an appropriate dress for a wedding?
Dion: There you get much more into the subjective and it took us a long time to get the model to learn because the the the help it was getting in its learning process was somewhat subjective, right?
Dion: So it took longer for that model to learn. And then the model was much more consistent, much more appropriate.
Dion: So I've seen those challenges but I I think the other tranche of challenges people face is that that confirmation bias you mentioned before was that it can't be right?
Dion: It can't be giving me this answer.
Dion: So I, I wanna test some more or I wanna change the expected results till I get an answer.
Dion: I'm more comfortable with. But I think the technology today around machine learning and, and teaching models is pretty straightforward.
Dion: It's really just the quantity and quality of the data that you can leverage.
Dion: I have seen instances where the frequency of occurrence creates challenges with the the lumines of the data if you will and, and those things just surely take time.
Dion: But that brings up the, you know, are you willing to share data?
Dion: And, and that's a whole another conversation that I think organizations have struggled with in the past is there's been some, you know, close to the vest nature around data.
Dion: Whereas if you want to take advantage of others data, you know, even if it's anonymized or whatever, it is truly more powerful, if we can leverage a larger data set, you know, and, and it comes back to the size of companies and, and how much data they accumulate on what frequency, you know, I've seen some technologies related to machine performance and making sure you don't have unexpected downtime.
Dion: Those scenarios are, you know, if you're, if you're gonna use a, a partner with a S model, they're gonna take data from other manufacturers or other companies who have similar machines, which means your data is already rich and your learning time before you start to see value can be significantly reduced versus trying to have enough bearing faults on a particular machine for you to predict it effectively.
Dion: So I think those things are all reasons why people have struggled, not willing to share their data, having confirmation bias.
Dion: or, you know, believing in, in the answer that the algorithms are coming up with,
Sami: you know, that's really interesting because that's the co mingling or, or whatever you call it really, it's the because then that comes with the GP T for instance, comes as a challenge because we're a little bit nervous if we want to enrich it or refine it with our own content.
Sami: Can we really trust that when it goes into that one ginormous data, data set that there is a flag somewhere that says, keep this separate from everything else.
Dion: Yeah, I, I think, you know, it, I think that's a huge challenge.
Dion: You know, you have contracts with, with some of these providers that, you know, promise to anonymize your data, but also they commit, they're gonna use your data, right?
Dion: Because that gives a better answer to the whole.
Dion: And I think it's very easy and noncompetitive examples, but when you get into either vertical integration or competitive environments, people, even if they intuitively believe it's good for the all still struggle with being willing to take on that risk that you're alluding to, which is, you know, what if somebody saw my data, what would that do?
Dion: And you know, the, the analogy that comes to mind when we talk about this is, you know, people's willingness to give out their own personal information.
Dion: If it makes their life more convenient.
Dion: If I give them a 10% discount, they're giving me anything that I want. Right.
Dion: It's, it, there, there has to be value in it and I think if we can do a better job as an industry and defining and, and laying out the value to the groups, I think you'll begin to break down some of those barriers.
Sami: Yeah. No, it's, it's, it's very true. So I have an, a little bit of an example.
Sami: There is Adobe they have within their data platform.
Sami: They also have this marketplace where you can essentially trade segments or sell segments, they're anonymized.
Sami: But those people who are using the Adobe stack essentially have this unique id of a person.
Sami: So so I could essentially dub my physician behavioral data in an an anonymized fashion for to pitch them some exotic vacations where if they are like neurologists and stuff like that.
Sami: And it, but it is a, it is getting kind of to the point of being a little bit creepy at, at times as well.
Sami: But you know, it's, it's really interesting with this security and safety is you've probably seen that Aws they recently launched their foundation models that called Titan and deliver through the bedrock deployment system.
Sami: And what is really interesting for us and a lot of our customers is that is the ability to take essentially a copy of their large language model and run it in your own virtual private cloud essentially in your own servers and then you can enrich it.
Sami: And I think it's a little, it feels at least a little more trustworthy than just dumping it into potentially a common pool.
Dion: Yeah. And, and, and you know, I think that, I think that my, my belief is that that model will continue to evolve and they will support this hybrid where I've got a kind of a leg in two different data lakes or however, you want to look at it where the results can be commingled real time without having to maintain duplicative data sets.
Dion: And I, I think those are, I think the technology will evolve to that because I think there will always be some element that people will not be willing to share.
Dion: I do think they should share more than they probably do today, but I do think the model will evolve to a more hybrid like situation where you don't have to duplicate the data, you don't have to duplicate the models.
Dion: You can, you know, kind of look across the various data lakes and, and you know, combine the best of both worlds where I have some proprietary and some nonproprietary data.
Dion: And I think that's another interesting area for many organizations.
Dion: I think there's lots of, you know, nonproprietary data that most companies don't even know exists that could augment their analytics and, and their A I experience.
Sami: Yeah. No, that's that's a really good point.
Sami: And it's interesting what you mentioned earlier about the machinery because about two years ago or so, I was working with with some A I stuff around plastic extrusion machines.
Sami: And and there were some newer machine vendors who would essentially say, hey, we've got this built in A I essentially.
Sami: but what they always forgot was that you don't have a, a plant that only has one new line or one new machine.
Sami: You always had like the, the old stuff and the older stuff and then the brand new stuff as well and figuring out how all of those can work together.
Sami: And and you would end up having multiple analytics tools or models to monitor them or ideal in, in that particular case, there's there's a company who essentially brought all the sensor data together.
Sami: That kind of help people produce a little more effectively.
Sami: Well, plastic is quite easy because, you know, if you screw up, you can always take it back and melt it rather than, than go back to the Granules.
Sami: But still, it's it stops the line and it's it's annoying and a lot of time human factor actually caused a lot of those problems.
Sami: And then once you bring all of that as well into your model, it's interesting to, I saw some of the data, which was quite fascinating is that sometimes on weekends they had less problems because people run them slower because they didn't want to mess up their weekend work.
Sami: My, I'm here on my shift. I don't want any problems.
Sami: So I'll run it, you know, higher temperature and slower. So no problems comes up.
Sami: But then they, you know, the model finds all that and then of course, you know, you can go and fix the, look for the root causes and fix them.
Dion: Yeah. And I think, you know, with it reminds me of, you know, some of the scenarios where, you know, the model is being taught to deal with certain levels of tolerance.
Dion: And, and I think those you're comparing that to what was maybe done, you know, by a human again, with that variability that's innate and, and, and subjectivity, which is innate in some of these people.
Dion: They, so when you start to tune the model, they're like, well, that's not really a defect or that's not really a problem.
Dion: Well, but you told me it was.
Dion: And then so, you know, you get these different levels of, of inconsistency through some of these tools that I think is just again, part of that journey around learning and understanding and tuning, you know, you may have tolerance that are set because you know, there's inherent variability in human assessment that you can now fine tune more effectively because you've got some level of automation doing that assessment.
Dion: And I think those are the things that are in incredibly interesting.
Dion: And, and again, you, you mentioned the machines in their varied age.
Dion: And again, if I can share data, and I know how old this machine is versus this machine.
Dion: And there's, you know, probably more data on the older machines.
Dion: And you know that that analogy of, you know, predicting a failure of a particular machine was done previously by somebody who would walk by the machine and kind of listen to see if the, the pitch was wrong or, or the, you, you've now quantified that.
Dion: And if you are willing to share data in that an anonymized fashion, you've got adequate data samples and a model tuned for that specific device or machine.
Dion: and you can save lots and lots of money with unexpected downtime.
Dion: Your point about raw materials is certainly relevant.
Dion: But you, you're now producing, you know, twice for 11 sellable unit because you had to start over.
Dion: And again, I think ideally, you get to the point where you can predict those things occurring and you can eliminate them or cluster those work activities in a particular scheduled downtime for instance to get them done effectively and create greater consistency.
Sami: And I think rolls Royce, the not the cars but the engine manufacturing manufactured for for the airplanes, they've done it pretty well in, in terms of real time tracking of all the mechanics.
Sami: So in in the engine and doing this kind of predictive maintenance and things, which is like if you compare it to the older times is like why change a perfectly good good piece.
Sami: But you don't think that in four hours or 40 hours or 100 hours it's going to to break and cause something else.
Sami: And, and, and a lot of that stuff would be quite impossible to do with the traditional approaches.
Dion: Yeah. And I think, you know, they, they, the collection of the data is so much easier today.
Dion: I mean, it, it's, it sounds low tech, but literally they can take a sensor and glue it onto the side of the machine and it can, you know, speak to the, the network and, and populate that data in a more automated way than we ever thought was possible.
Dion: And, and I think those, those endeavors and, and those types of things are truly going to allow companies to better tune their capacity needs and their performance.
Sami: No, that's very good. So I, I, I wanna save some time for questions as well.
Sami: But since you've been cio for a long time and if, if you would give like a couple of let's say three things for another cio to think about like how to, how to address the, the future of a, you know, future of A I, the A I is here.
Sami: It's not future so much anymore. But what would you do tomorrow?
Dion: You know, you know, my experience is you, you need people focused on it, you know, within, within the technology world and, and then, you know, partnerships or select functions within the organization where you think there's, there's an appetite to learn and to iterate to me establishing that partnership and, and nailing some early wins.
Dion: those wins. I find fascinating because if you socialize those wins, you then create a, an appetite for more.
Dion: And you can tell people that, hey, we didn't think this was possible, we did it, which helps others think more broadly and more openly about things they may not have considered.
Dion: So to me, having a team focused and well trained on doing this and then having them establish, you know, one or two or three key partner functions where you can really create some interesting dynamics and, and some wins.
Dion: And I think ultimately that will create more demand, more, more needs, more opportunities.
Dion: But the third piece of this is also measuring those types of results to help quantify the benefits.
Dion: But, and not because we want to, you know, watch every dollar and, and you know, cross every T and dot Every I but because if you're successful it will help justify the additional funding needed to do that more broadly or more deep in functional areas.
Sami: Do you ever find the need to, to train the, what business people, you know, air air quotes in technology and what, what is the art of possible?
Dion: A a absolutely. I think, I think it is, as I said, it, I called it earlier.
Dion: It's a journey and it's a journey for the entire organization and there'll be times where the technology team leads and then there'll be times where your partners across the organization will lead.
Dion: And that's the beauty of that collaboration in my mind.
Dion: You know, neither, neither function will have the best answer but together between that healthy conflict about what are we trying to accomplish?
Dion: How effective is it to get done?
Dion: You will get in compelling answers and, and to me, it's, it's, it's technology, but it's more a people challenge than it is a technology challenge.
Dion: It's, it's setting them up and again, I find people who you and you know, approach about doing more data and analytics or A I and they're like, well, if you could just give me this column on this report and you know, you have some work to do, right?
Dion: You need to, you need to say, OK, no, that's not what this is about.
Dion: This is about how do we answer those questions? You didn't think you could ever answer?
Dion: And and that is the learning. And again, there's no, there's no exclusive on that.
Dion: So that's why I love the fact that there'll be times where the, the function is pushing the technology team and vice versa.
Dion: And I think that's the beauty of that collaboration.
Sami: Oh, that's, that is so true.
Sami: I remember vividly when I had a couple of jobs ago, I, I owned a corporate data platform and I had a little over 15 terabytes on, on Amazon red shift.
Sami: And then one executive came to me and said, hey, do you mind giving this all to me on Excel, I'd like to understand what's in there.
Sami: And I was like, wow, let's talk about it. What are you really trying to achieve?
Sami: But that's a a and I'm with you.
Sami: It's kind of like what I'm, I believe is that, you know, people who are on the kind of the line of business side of things in marketing or sales, particularly in marketing, I think it would be quite helpful by the production in other places.
Sami: It, if they have a little bit better understanding of technology and then ideally also, and the people they engage on the technology side and the in the cio office is that if they have some tech technology, people who also are interested and understand the impact to the business, then the conversations happen a little bit better and, and they're more realistic and we have kind of less finger pointing.
Sami: Like usually what happens between sales and marketing.
Dion: True. You know, I used to, I still believe that, you know, the best technology professionals know at least as much about the business as their partners across the organization.
Dion: Otherwise you're, you're, you're, you're either trying to find a problem for a solution you have or you're, you're missing the mark and there's so many turns that we all take during these initiatives that understanding the objective and having that strong collaboration are, are critically important on both sides.
Dion: And, and that's why I'm, I'm, you know, it's not an either or it's a, it's a function of, you know, collaboration rather than one lead and one problem.
Sami: Yeah. No, absolutely. Partnership is always good. Awesome. So let's see if we have any questions.
Sami: So there is handy, handy questions and answers on Zoom if you want to try it or if you want to try raise your hand and pop on the video, that's all all good as well.
Sami: Otherwise, if nobody has any questions, I have one more question
Sami: Nobody, anybody, you can use the the chat too if you're afraid of Q and A no questions.
Sami: Good questions, right? It's so, in fact, actually, I have, I have one more serious question and then I have one kind of I promised to my team that I will not ask any Star Trek related things.
Sami: because I just finished, finished the Picard, which was awesome.
Sami: But so how close do you think we are to Skynet?
Dion: I, I don't think we're really close.
Dion: And, and I, and I think, you know, I, if, if we were and we chose to take some actions, you know, if you follow the kind of terminator plot, you know, they'll just do time travel and change our decisions anyway.
Dion: So it's not gonna do us any good to worry about it.
Dion: But I don't think we're, I don't think we're, we're very close and I do think, you know, it, it, the, the, the interesting part about it is, it reminds me of the story several years ago where the Facebook bots were, had created AAA shortcut or a language or a code that the humans couldn't understand and everyone's freaking it out and the rumors are they had to turn them off and they just had to tell him we'd prefer you to use English because we don't understand what the heck you're talking about.
Dion: But it, it is that, you know, that fear and there's been so many movies and so much press around, you know, the, the robots are gonna take over the world and they're gonna do this and they're gonna do that.
Dion: I think we're pretty far away from that.
Dion: and, and, but I do think, you know, I think we do need to govern and understand what we're doing.
Dion: I just don't think it's, I don't think we're in a position where we have to worry about the, the robots taking over.
Dion: Not for a while.
Sami: Yeah. Well, the first people they would probably take out would be the Boston Robotics testers, you know, if you to slap the robots with, hockey sticks and things like that.
Sami: And it's actually, it's fascinating when you think about how long this has been going on because, Asimov was writing about this in the forties and fifties with the three laws of robotics and things like that.
Sami: But we did get a question.
Sami: So, and this was actually going to be my last question practically as well, which was, so a role of A I in e-commerce to what extent will it be a game changer or in developing, go to market strategy.
Dion: I, I think for many companies, it will be, it will drive dramatic change.
Dion: I think it's, we're on the on ramp to that.
Dion: I don't think we're, we're at the, the top of that bell curve by any stretch.
Dion: but I do think it's going to change the way organizations are structured, how they run their processes, the rules they have, you know, I, I, again, my, my old time story, you know, when ATM S were orig originally invented, that was the advent of no more tellers, right?
Dion: But what you saw for years after there were actually more tellers statistically than there were prior, their jobs evolved.
Dion: They did higher end work, they probably did more work around different offerings that change in, in how companies are organized and the roles they play.
Dion: The other one is we talked about chat GP T generating blog content. That's great.
Dion: Someone needs to review it. So the writer now becomes an editor and it's just a change in an evolution of roles.
Dion: I think those efficiencies that will come from it, which is one thing. And I think that's the minor benefit.
Dion: What I really think is the, the insight fullness and that the, the value of the answers to drive different decisions and outcomes across businesses will clearly be a game changer because I don't think they've had that luxury and I think the gut intuition is going to diminish and it'll be greater data enrichment to make those decisions.
Sami: Excellent. And we got another question, which is, what is your view of Jeffrey Hinton's recommendation on temporarily posting on A I
Dion: I intuitively, I don't agree because I don't understand what the pause would do for us.
Dion: And you know, it comes, I, I, I don't think we can go, we, we can't, we can't slow down technology, it's not going to happen even if we try and do it, it's gonna pop up somewhere else.
Dion: It's going to continue to evolve. There's, there's really no ethical reasons to do so today.
Dion: And I think I would rather us focus on understanding it more and creating and understanding where and if we need guard rails versus stopping or slowing it down or pausing it, I, I don't really understand what that will do because I think it's still gonna come if the answer was we need to better understand whether there are risks or whether there are.
Dion: That's ok. but the pause makes no sense to me.
Dion: It, it just seems again like something that the press has kind of taken to heart saying it's that big bad evil machines are taking over the world.
Dion: We have to stop them.
Sami: Yeah. It, it feels a little bit and this is a little politically, topic on the tiktok banning in the US.
Sami: we know that Facebook and all the other vendors, Google is reading our emails and everything else probably still, it's, it's, if we stopped it, if we, in my opinion, if we paused it in the United States, for instance, the Chinese and the French and the Brits and everybody else, many of the other nations would continue doing it whether openly or, or in hiding.
Sami: But, yeah, this, this thing would, definitely keep on going. Yeah, I agree.
Sami: Ok, we have one more question and then that's probably going to be a wrap up then.
Sami: So, does the organization need to consider empowering end customers using a I for better decision making, especially in the retail world or, or otherwise?
Sami: And how will it impact?
Dion: my opinion is, of course, right, we, we want to put the power in the decision maker.
Dion: You know, the, how many of us have been there trying to buy a car?
Dion: And the guy says, or the lady says, I have to go check with my manager, right?
Dion: When you know, they're just going to get a glass of water and come back and give us the answer, they were gonna give us anyway.
Dion: But yes, we want to put that power in the decision makers hands and empower them to make the right decisions.
Dion: So all of that is absolutely true.
Dion: You know, a customer service agent or a sales person that's empowered to understand where their guard rails are and the net result of making a decision, they know that that these, these technologies are going to enable companies to be more thoughtful about what the impact of those decisions are and making them and then figuring it out isn't going to help.
Dion: But if I can influence the person who's making that decision to give them an influence, maybe the wrong term, but inform them about the ramifications, both positive and negative of those decisions is incredibly powerful and it, and to me that's what should be fueling our empowerment drive is to allow people to make the right decisions.
Dion: Understand when maybe they're making a less consistent decision and what are the ramifications of that?
Sami: I know that's, it's kind of democratizing the access to analytics as well or the data overall.
Sami: And maybe, maybe the GP T and other tools will actually help us or a chat kind of a chat.
Sami: But a conversational A I can help those people make better decisions about
Dion: democra. The democratization is, is so important.
Dion: You know, we've, we've all, many of us have lived in the environment where you had those kind of data heroes where, you know, they kind of had all the data and you had to go to them and get the answer and we either they lived in I T or lived in a function.
Dion: It didn't really matter, but they, they had their data, right?
Dion: And it was probably not consistent with other data sets across the organization and you couldn't share across and there were so many problems with that.
Dion: So democratizing and allowing the decision maker to do their own analytics is, is so important.
Dion: And, and the other reason why I think it's important is as I've said before, it's, it's very iterative when I get an answer, it, it hopefully will light a light bulb that says, well, what about this?
Dion: And that's another question and, and to give me, the ability to iterate through to a conclusion is, is incredibly powerful for organizations.
Sami: Absolutely. Absolutely. And it, it is a tricky area because then you have some executives who are not so much growth mindset, maybe who are like, ok, what's the return investment on all of this?
Sami: You know, how are you going to measure it?
Sami: And it's, it's a fun exercise, same as the marketing attribution.
Sami: You know, it's, it, it, it becomes impossible because of all the touch points. We have nowadays.
Sami: Yeah, for sure, for sure. But you have to have an answer, right?
Dion: You you do and, and you know, to, to, to say, wait or pause or you know, to me the, the the best advice I can give people is start the journey, waiting is not going to help.
Dion: There's so much opportunity and so much to do and it's, it's likely harder than most people think they're gonna find things in their data they didn't expect.
Dion: But starting that journey sooner rather than later is going to be a competitive advantage for companies.
Sami: Absolutely true. Do you have time for one more question which we got?
Sami: So there is so he is saying I'm from web U I background nowadays, natural language processing to voice U I is evolving in user interface domain.
Sami: Agree what would be the impact of A I in future for web user interface think?
Dion: Yeah, I think I think you're seeing some of it with, you know, some of the, the speakers that you see in your homes and those types of things.
Dion: But I, I do think, yep, we need to meet people where they wanna be if they want to speak to us or if they wanna, you know, manipulate a mouse or whatever.
Dion: I think those things are gonna continue and I don't think you're gonna have one answer fits all.
Dion: you know, as, as a, a spouse of a voice texter, I think there's some work to be on there.
Dion: I, I joke that I don't wanna have to edit your text because you spoke it instead of typing it, but I do think that that user experience for voice is going to continue to evolve.
Dion: Well, I, you know, we've seen it in the last few years with, with all of the, how do I add my item to my Amazon shopping list?
Dion: And those types of things are absolutely gonna continue.
Dion: And I think people are going to, I think you're gonna see it in both environments that the, the, the beauty of that evolution though is they're gonna get better and better and better.
Dion: You know, I'm not gonna have to slow my words down.
Dion: I'm not gonna have to say them in the right sequence to get what I want.
Dion: I'm literally gonna be able to have a conversation which I think will make it much more palatable for people to use voice.
Dion: And I think you'll see the connective house and connected store much more in the future.
Sami: Yeah, absolutely. And the beauty of all of this is that as the technology gets better, the platforms get faster, you know, like Amazon investing in their own silicon to doing things, things get cheaper.
Sami: So it, it, they actually are now more affordable for small and mid size companies as well.
Dion: Yeah, I think it is literally the perfect storm.
Dion: If you can accumulate the data, the technology is becoming, you know, the compute power is becoming cheaper, the the software is becoming better and, and more sophisticated.
Dion: It really is the time to jump on.
Sami: Fabulous. All right. So look, we, we took a few minutes over your time, but thank you so much for being so gracious with your time and really, really interesting insights today, it's it's been fabulous talking to you again about this topic because yeah, everybody is talking about it, but not enough people are doing it, I guess.
Dion: Yeah. Well, it was my pleasure. Thank you again for having me.
Dion: I certainly am passionate about the topic and love to talk about it. So thank you for
Sami: that. Absolutely. Thank you so much.
Dion: Thank you.
Sami: Thanks, bye bye and thanks everybody for for participating today and joining us.
Sami: Thank you.