Ep 1 - Why Isn't Our AI Pilot Turning Into Anything Real?

(upbeat music) - Hello everyone, welcome to the first episode of the new series, "Operations is the Alpha" from forever forward. The first question that we take from, obviously inspired by the time I spent at the Proptic Connect Europe is the question that most people who've invested in AI over the last 80 months or 24 months have on their mind that why isn't the AI pilot turning into anything real? And in bed in mind, I'm also not currently talking about the return on investments. It's just that there's this valley of bed and where all these pilots kind of go and nothing really happens from the operations point of view or the business's usual implementation point of view. Now, what's the difference between the pilot that worked and one that's about to scale? A pilot that scales was really, or should have actually been only sculpted about a real decision or a real business flow into it and not a feature demo. A pilot that usually dies has really been done to prove a technology and not the business change around it. So could be as a reason of the push that happened from top to bottom that something needed to be done AI and you ended up doing it. But fact of the matter is that there was no thought around what one real decision is it gonna change and to end? Or even alter, transform, whatever it could be. And the ones that made it to, made it like as a scale, had the people who would actually use the output involved from day one and not as an afterthought. I have been a personal like backer of the idea that unlike all the other transformations, digital transformation where your IT teams and CIO has taken the lead. I think in this case, especially when we look at operations, I think the people who actually operate need to have a very formidable seat at the table and probably be driving it as well. IT, you know, and the information technology team or the information security team, all of them are extremely important because they help us make the right technology decisions. But I think the outcome needs to be on the operations and not by the ones who are custodians of innovation. And scaling needs the underlying data and provenance to be already reliable but the data layer has not been around consistently just has just been patched for the pilot and you obviously won't survive the production volume. However, I must also say that we are also for, I'm also for belief that, you know, you can wait the eternity to get your data perfect in a perfect condition and then say, then we are going to implement because the fact the matter also is that you are operating your acid facility with whatever data you have. Now, I would have, you know, it's hard to believe that you're going to not trust that data because you're operating at those conditions, right? You could say that my confidence is not extremely high which is where we are going to work to improve the data quality. But again, data quality does happen to be one of the points where it kind of goes, goes a little, I mean, it doesn't scale. And then when you look at, you know, that you have IX that interest everyone and everyone was extremely happy is because the first time, I mean, sometimes it does happen that the credibility is fragile which means that the first time an output looks wrong to an operator which means, you know, it just takes far more effort for the team to bring them back onto the table than, you know, than it takes to impress for the first time. So I think that is where it is very important to set the right, you know, expectations from day one. And also, I think when you look at pilots, you know, on budgets for the overall goal, the time, effort, governance, all of that will have to come into play. One scopes that out, very few people scope that out for a pilot and leave it and say, okay, we'll see what to do after the pilot. And by the time when you are coming into a stage after the pilot, you're already saying, oh, this is actually going to be so much of an effort, right? And the other third point is, is, you know, the output of the pilot needs to live inside systems, people already trust, they use daily, and more often than not, unless you're changing your entire technology stack. If you're just thinking one more tool, and saying, hey, you would also need to use this, why you use five, six, seven other systems is probably not gonna work much. That's one thing, you know, for example, we've tried to stay away, we say, no, no, no replace. And there is obviously there are some use cases where you would have to have your technology stack, but the moment you are wanting to make too much of a change to the people who operate, and the tool ends up being one more tool in the armory, then I think that's gonna be a problem. And also, at times when people look at the output, the reasoning behind the AI generated output is fundamentally important and not just the answer, because people are skeptic in nature, and that they're rightfully so. So there are some sort of personas who really need to know why were certain decision made, what was the reason why that decision was made or recommended, and it's almost like us talking or working with our colleagues, we don't just take the answer in most of the cases, we want to understand the reason behind it. If it is an important decision to be made. So in my humble opinion, these are like top cohorts of reasons why the pilots fail, and as we conclude the episode one, there's one thing that I don't actually use with all our customers is that I never use the word proof of concept. I'm saying it is not a concept that we're proving, we are gonna prove the value to your business. So it's always good to say it as proof of value versus saying proof of concept. Obviously, you guys wish any people listening to the podcast, you would have your own views, suggestions, please feel free to chime in, and we are gonna be back again with the episode two, where we pick up is the pilot not being successful, and that implementation is a technology problem, or a people problem wearing a technology costume. Till then enjoy this episode, and let me know what you thought of it, and we'll be back again soon for episode two. Thank you. (upbeat music)