 Hello, everyone, and welcome back to theCUBE's live coverage of UiPath Forward 6. I am your host, Rebecca Knight, along with my co-host, Lisa Martin. We are joined by two guests for this segment. We have Alan Pellsharp. He is the founder of Deep Analysis. Welcome, Alan. Thank you. And Luke Palamara, VP AI product development at UiPath. Hello. Thank you so much for coming on. Thanks for having us. So we're going to be talking about applying AI to the enterprise, and Alan, you literally wrote the book on this topic, Practical Art Diffusion of Intelligence and Enterprise Playbook. Why don't you start by giving our viewers the lay of the land and where you see the current state of the market in relation to AI, applied AI? Well, first of all, I wrote a book. There are others, but yeah, it's an interesting thing. I mean, AI in a sense isn't new. It's been around a long time, but the last few years it's become just so much more accessible, and frankly, a lot smarter than it was before. And so people are still sort of thrashing around a little bit trying to figure out what to do. I think in very, very practical terms, there's still a world out there that thinks this is for people with advanced math, data science, and actually it's not. It's incredibly accessible now. So we're in the early phases of figuring out what to do with it, and some of those things are very basic, undertaking administrative tasks that were previously manual. It's not very exciting, but it saves a heck of a lot of money, and it saves a lot of headaches, right? So hopefully in things like insurance, in healthcare, in customer service, just making life easier for everybody. There's many more complex and fantastic and amazing things, but those are the things in business that we're first encountering, and that's actually pretty cool. Yeah, yeah, and you said that a lot of people are still reticent or skeptical. They just haven't tinkered with it. As you said, we have this idea that it's for experts, for people who are good at this stuff. Yeah, and in truth, it is for experts. It's very complicated stuff, right? But just a few years ago, essentially you were building your own AI, it's not the case anymore, right? You come to a company like UiPath or whoever. They've already built it for you. So there's that element to it, and I don't know if it's about skepticism. I think it's just, we don't know where to start. This sounds clever, this sounds great, but where would we start with this in our organization? There is a secondary challenge that it's a cliche, but it's a truth, garbage in, garbage out, right? So if your data's pile of junk, it ain't going to work too well. So there are some very, very, very practical to use that term yet again. Challenges for organizations, but hey, it's there, you can use it, and if you're prepared to roll your sleeves up, you can start using it immediately. Luke, you know, one of the things that we heard during, I loved the customer, the keynote this morning, customer examples, Rebecca and I and Dave were talking about that this morning, and it was a really dramatic, bold story told consistently through the voices of customers this morning on stage. I don't, we don't see that to that, I haven't, to that level of like dramatization, but I thought it was really impactful. But one of the things that I wrote down that I loved that Rob said is anything that AI can do, AI and automation can do better. Talk about, Mary talked about, you know, some people are still, you know, Ellen to your point, is it good, is it bad, is it neutral, what is it? Obviously it varies, it depends. The answer is D, y'all will be above. But give us, Luke, some perspective on AI and automation better together from UiPath's perspective and how that's going to impact your customers. Yeah, Lisa, so, you know, I think, actually Mary said it well this morning around, you know, if you think of AI as sort of like the, the brain and the automation as the muscle, these two things together really make AI, put AI to work in the enterprise. You know, I think a lot of the, you know, there's a lot, obviously a huge amount of excitement around gen AI right now. And of course you heard Graham talk about earlier today specialized AI and generative AI. But with generative AI, it's sort of the use cases that you see today really have generative AI locked in this personal productivity type of use cases where it's helping you, it can do, you know, it can do things like write texts, it can summarize texts for you really well, it can analyze data, but it's sort of relegated to these personal productivity improvements. And what really with UiPath, what we're focused on is going beyond just personal productivity and actually making it productive for the enterprises. And you know, I think that's, when you look at the capabilities that we bring to the table to help do that, first, you know, what is it, so what does AI need to actually, you know, make it work in the enterprise? First is context. So you have to bring context to the AI in terms of information so they can make the right decisions. You know, how valuable, we heard from Walter Isaacson earlier on the keynote stage as well, and he's an amazing author, I love Walter Isaacson, but how many people would actually hire him and put him day one answering customer inquiries and writing back like responses to customers? Well, you probably wouldn't because he doesn't have the context, he knows a ton of information but he doesn't have the context to know, you know, how do I respond to this customer to this type of request? Where do I get the information? Where do I enter information to be able to do what they're asking me to do? So in the UiPath, this context comes from our integration service and bringing that data in. And then the other elements are really around action. So being able to give AI the capability to take action is really what UiPath is all about. So it has to go beyond, to put AI at work and make it productive in the enterprise, we have to do more than just allow it to make decisions with that context, you actually have to allow it to orchestrate the movement of information from system to system and enter it into, you know, do an invoice processing use case, enter that data into a system. So that's really what UiPath is focused on and, you know, context and providing context, enabling it to take action, that's where we amplify AI with automation. I want Walter Isaacson as my customer service rep. No! I think he might be a little pricing. Luke, sticking with you, you mentioned the word integration and it reminded me of last year, August 2022, re-infer was acquired to enhance everyday customer conversations through automation. What's progressed on the integration fronts? What are customers benefiting from one year later? Yeah, so we've been super thrilled to have re-infer as part of UiPath and in April we announced the true integration of that into our platform as communications mining. So communications mining has unlocked a ton of new use cases for the platform. So around unstructured communications from customers, then being able to look at what are the communications that our customers are emailing us about, for example, being able to understand that voice of the customer, but then actually going beyond understanding the voice of the customer, but actually being able to automate what based on those inquiries of, you know, can I cancel a policy? Can I upgrade my policy? Can I modify my insurance policy? Things like that. We can now automate those things inside UiPath platform. Whereas before, you know, re-infer could route emails to different people based on who was emailing in about what topic, but now we can actually take action and actually orchestrate the movement of the data and the automations downstream. And so this combination with our orchestration and automation, but also our document understanding product has allowed us to expand our intelligent document processing suite to beyond just documents, typical documents you would think about to also customer communications. So when that customer communication comes in and has an attachment, we can now act on that attachment beyond just like what's the content in the email. And so, you know, it was reflected, I think, well in the Everest report that came out earlier this year around intelligent document processing, which ranked UiPath as the clear leader in that space. But one other, you know, one other aspect to the integration of the communications mining into the UiPath platform, this goes back to something Alan was just talking about, which is how do we make applied AI for more than just like the data scientists in an organization? And really we've taken the active learning capability in communications mining, which allows a non data scientist to create machine learning models without knowing any of the science behind it and rapidly train these models to understand the specifics of their domain. And we've taken that, and you heard an announcement earlier today from Graham that we've introduced active learning and document understanding inside of UiPath platform as well. So not only have we taken the technology and integrated it very deeply, but we've also take the learnings of how we can take these non data scientists to train AI models and infuse that in other products inside of our suite. So we've benefited in multiple ways and we're seeing great adoption of the communications mining capabilities in our platform. Awesome. Alan, I'd like to ask you what your take is, what your perspective on UiPath's progress with IDP and communications mining and how they should be thinking about making sure that these IDP systems are fair and transparent and sort of what are the other considerations to take into account? There's lots of considerations to take into account. I think the one thing you could take a step back, so intelligent document processing, it's relatively new. Document processing is really old, it's 30, 40 years old. Just wasn't very good before. Again, the last few years advances with AI have meant that the accuracy level, say, for reading the text on a document in very simple terms, way more accurate than a human, right? So that's a big threshold to get through. So that opens up a whole world and the initial targets are basic, basic, basic business documents, invoices, bills of lading, statements, whatever, the basic currency. But what things like re-infer do is they open up the possibilities more, right? So you have IDP for what we would call long-form documents, like very long contracts or whatever, that's cool. But most communications are not big, chunking formal contracts and leases, they're email messages, they're text messages, they're, hey, Bob, did you see that? Shall we do something about it today, tomorrow, whatever? And so that starts to open up those possibilities and I don't think we've even really started to explore what those possibilities are yet. That's how early we are in the market, that's exciting. So how are you tracking the progress of these initiatives and what metrics are you using and how are you determining? Really hard, that one. Because it's such a, it's a very English phrase, but fighting with fog. Every time you try and do something, it's changed and it's moved, right? So there's those kind of things. I think there are metrics, I mean, there are metrics in the sort of formal world around bias and whatever and there's a whole world, sort of sub-world, if you like, evolving around AI ethics and benchmarking and maturity and accuracy and everything. But outside of that, when, you know, I use the term enterprise, but it could be a hospital, it could be a government department, but large organizations, right? When they're deciding whether they're going to use AI, well that means they're going to have to invest, right? Because it's change and change costs money. I think that's where there's a lot of shift going on at the moment because unfortunately we see things like AI and RPA in the past and the current probably as, oh, they do the job of a human. So we need less people. That's not a good business metric. No. It's not very nice for one thing, right? But it's actually not very accurate. The reality is, you start these projects, oh, we're going to reduce headcount. No, no, no. They're probably still going to be there at the end of the project. You just spent some money, you've still got the same people. It's not a good metric. Improving accuracy, reducing errors, improving customer experience. You know, I won't say the name of the hotel, but I was trying to check something at reception this morning on the phone, right? Press one for this, press error for the, oh my heavens, right? Those kind of experiences. You've all been there. We're there every day, right? Yeah. Those are the kind of experiences when something isn't paid on time, when your appointment's been canceled, when your insurance claim is in limbo somewhere, right? These are the kind of things where you take correspondence, so re-inferred, you take correspondence, you take messages, you take those documents. And I'm not saying we're quite there today, but you can get to that point where you can get an answer to a question. You can get a response. And as a human being, alleluia, right? Whether you're a citizen, whether it's your insurance claim, whatever it is, life stuff, that's great for you. But that's awesome for the organization too. Because errors, angry clients, angry citizens, delays, inaccuracies, that's not fun for anyone. That's expensive, yeah, but it's also just pain. Right. But you're trying to get rid of it. Yeah, yeah. Luke, what, I'm curious. This is a question I always grapple with because there's so much negativity in the news about AI. And I think, you know, I feel like those of us that are in technology have some sort of sense of obligation to help on the awareness front to really start lauding all of the positives that are happening from it. How do you, in customer conversations, what are some of the things that you talk to them about in terms of, you know, not to Alan's point, not reducing headcount, not eliminating jobs, but really enhancing, empowering. Where are customers with that good, bad, neutral kind of perspective? Yeah, I think this also ties into sort of Rebecca's question around the, how do you track the success of these sort of endeavors? And you're right. So we look at this as freeing up people's time to do more creative work, right? If I'm constantly replying to the same email that's coming in to my help desk, that's a very boring task, right? We want to take away that drudgery and free up. Unless it's Walter Isaacson. Unless it's Walter Isaacson. Yeah, yes, of course. I would love to get an email from him. But, you know, I think on one side of the coin there's, how do we free up people's time to do more creative work at work, right? And there's different ways to measure that. And that is, I really like to add, like, you know, it sounds a little boring, but average handling time of an employee, right? So if you're reducing the, so when we have these AI processes, we bring humans in the loop to actually validate what the AI is doing oftentimes, especially the more important or critical process, the more you want the humans to do validation. So you're not always reducing the work to zero. You're actually just allowing them to do more, but actually, you know, with less of the drudgery, they're just checking what the AI output, right? And so in that way, they're still handling the email, but their average handling time of those tasks goes way down. So that's one side of the coin. We can measure that in a bunch of different ways, but I've seen average handling time of a process be very, you know, of the human interaction in that be one way to measure how much time you are able to free up that creativity. The other side is customer satisfaction. So if you're able to actually, you know, process a customer request and get something done that they asked for it faster than you were before, where they might have been waiting days for somebody to do something that is working through a backlog of emails, you're going to get better net promoter score up from your customers, right? You're going to get the customer saying, yeah, I love working with them. I love the type of customer support I get from that company. So that's really the other side of the coin beyond just, you know, freeing up people's time to be able to do more creative type of work. So yeah, that's really like how we see that the ROI on those two different angles of it. Yeah. Last question for both of you. What are your thoughts on the next evolution of IDP and Com's mining? The next evolution. So I think there's lots of different angles you can take on this. Like more specific to IDP, the AI is getting better and better and better. I think we're going to reach a point, probably not in the two-distant future, where every document, intelligent document processing platform, it's sort of like the AI itself somewhat gets commoditized because it's reached thresholds of how accurate it can be, right? It becomes a point where there's diminishing returns on actually what the AI itself can do. And it becomes more about all the tooling and capability that surrounds the AI. Like being able to bring a human in the loop to validate what the AI's output is, being able to monitor those success metrics of how well the process is performing in your automation. Like what types of, you know, how much average handling time are you able to reduce in that process? What's the customer satisfaction? So and we think UiPath is really well positioned there because we're not just an IDP vendor. We provide this capability around it that allows you to measure the success of the process you're automating and be able to do things. We're working on things like to allow you to simulate how your process is going to do before you even deploy it. So that tooling around it, human in the loop, simulation, being able to monitor and audit what's happening in your automation, those are the things that are really going to make the difference in the end because the AI itself will reach a point of diminishing returns and what it's able to do. Alan, final word? I'll just take it from a different angle. I think, you know, at the moment, the traditional world of document processing has been about things like accounts receivable, accounts payable, really, really boring stuff, but very important stuff. I think now it's becoming democratized. I mean, I was a different thing last week and there was somebody talking about a real case study in South Carolina, I believe it was, where it was to do with foster children. And they moved from an 18-year-old access database and handwritten paper and forms to an electronic system. When you go out there, and as an analyst, I try very much to do that, keep one foot in the sort of, if you like, the Silicon Valley world, another foot in the real world, you go into supply chain, you see the most amazing technology on cargo ships, then you go into the office and explore to ceiling with paperwork, right? So that's what I'm excited about is there's a whole world of paper out there. We live, when we're in the tech world, we live in a world where we seem to think everything happens digitally these days. That's not the case. It's just not like that out there. And the opportunities are almost endless. And again, that brings benefits to everybody if we can get that right. So that's great. Great note to end on. Alan and Luke, thank you so much for coming on theCUBE, a good conversation. I'm Rebecca Knight for Lisa Martin. Stay tuned for more of theCUBE's live coverage of Forward Six. We'll be right back.