Latest / The DEX Show: A Show for IT Change Makers / 🚀 From Zero Tickets to High-ROI: AI + DEX in 2026 (w/ Samuele Gantner and Vedant Sampath)

Transcript

🚀 From Zero Tickets to High-ROI: AI + DEX in 2026 (w/ Samuele Gantner and Vedant Sampath) · The DEX Show: A Show for IT Change Makers

• 0:00
  Speaker 1: Hello there, Welcome back to the Deck Show for twenty
• 0:03
  Speaker 1: twenty six, and I'd love to start the year with
• 0:07
  Speaker 1: happier news, but it really does behoove us to pause
• 0:11
  Speaker 1: for a minute before we introduce today's show to acknowledge
• 0:17
  Speaker 1: the passing last month of a friend of the show,
• 0:21
  Speaker 1: an absolute legend in the deck space in Alan Nance.
• 0:26
  Speaker 1: On a personal note, I've been working in the deck
• 0:30
  Speaker 1: space for around six years and nobody was more welcoming
• 0:37
  Speaker 1: or informative than Alan. Just a really lovely, kind, funny,
• 0:43
  Speaker 1: engaged and engaging person to work with, an absolute pleasure
• 0:48
  Speaker 1: to have on the show repeatedly down the years. I
• 0:51
  Speaker 1: think he was our second or third guest ever, so
• 0:54
  Speaker 1: it could be said, you know that he was one
• 0:56
  Speaker 1: of the guests we had in mind when we launched
• 0:57
  Speaker 1: the podcast. So sending as a show, Tim, myself, everybody
• 1:02
  Speaker 1: who's worked with Alan at next think sending our prayers
• 1:06
  Speaker 1: and our thoughts to his family, of course, but also
• 1:10
  Speaker 1: his friends and his colleagues. Rip Alan.
• 1:13
  Speaker 2: That's welcome it, change Makers to the Deck Show with
• 1:19
  Speaker 2: Tim Flower and Tom McGraw.
• 1:21
  Speaker 3: Let's get into it.
• 1:24
  Speaker 1: Happy new year, change Makers, Happy twenty twenty six. My goodness,
• 1:28
  Speaker 1: and regular listeners will know how we'd like to start
• 1:32
  Speaker 1: the year on the Deck show. It's with the return
• 1:34
  Speaker 1: of a very special guest, none other than Next Think
• 1:37
  Speaker 1: Chief Product Officer, Samuel Gantner. Sam's going to be joined
• 1:42
  Speaker 1: by somebody else also, but let's just Tim and I
• 1:44
  Speaker 1: welcome Sam. First of all, Sam, how are you doing?
• 1:47
  Speaker 1: Welcome back. Always good to have you, Sam, Hey and.
• 1:51
  Speaker 3: Awesome to be back. Thank you so much for the invited.
• 1:53
  Speaker 3: There is no better way for me to start the
• 1:55
  Speaker 3: year than being with you guys for an episode.
• 1:58
  Speaker 1: You're on one of our main love tradition. Sam.
• 2:01
  Speaker 3: It's good to have you.
• 2:02
  Speaker 1: And today it's not only welcoming back you know you
• 2:06
  Speaker 1: as a returning guest, but for the first time, unbelievably,
• 2:09
  Speaker 1: for the first time, Next Thing so in, chief Technology Officer,
• 2:13
  Speaker 1: is joining Sam for the show. Is the Dan sympath
• 2:16
  Speaker 1: the Dan. How has it happened? How have we gone
• 2:18
  Speaker 1: so long without managing to make this connection and get
• 2:22
  Speaker 1: you on the show. It's just some Next Thing is
• 2:24
  Speaker 1: just so full of full of guests you can overlook
• 2:26
  Speaker 1: one as good as yourself.
• 2:27
  Speaker 3: But welcome, Welcome to the show for the first time.
• 2:30
  Speaker 4: Thank you Tom it's great to be here. It's a
• 2:32
  Speaker 4: great way to start off the great to be with
• 2:35
  Speaker 4: Sam on the show as well. I'd like to keep
• 2:37
  Speaker 4: a little profiles, so I think I've managed to successfully
• 2:41
  Speaker 4: do that.
• 2:41
  Speaker 1: A little too successful maybe with that, but we've got
• 2:45
  Speaker 1: to We've got a real power couple on the show.
• 2:47
  Speaker 3: To him to kick off you and yes, sir.
• 2:50
  Speaker 1: So I know we got We've got kind of two
• 2:52
  Speaker 1: sections actually, which is pretty for the first time, and
• 2:55
  Speaker 1: an interesting structure to this discussion. And I'm going to
• 2:58
  Speaker 1: kick us off, guys. Obviously, it's special to have who
• 3:03
  Speaker 1: I think over the very least is next thing is
• 3:05
  Speaker 1: to technology captains with us together, and I thought it
• 3:09
  Speaker 1: would be interesting to start by talking about the last
• 3:12
  Speaker 1: two to three years of profound technological transformation and what
• 3:17
  Speaker 1: they have meant for your teams and the areas of
• 3:21
  Speaker 1: the business that you oversee. So Dan, as first time guest,
• 3:26
  Speaker 1: let's let's start with you. Dan, how has AI forced
• 3:31
  Speaker 1: next thing and forced yourself to rethink the product or
• 3:34
  Speaker 1: should I say technology in your case design process? So
• 3:38
  Speaker 1: what part of the traditional research cycle from prototyping to
• 3:43
  Speaker 1: iteration have had to be restructured? First of all, because
• 3:47
  Speaker 1: AI as a technology simply behaves so differently from conventional software.
• 3:54
  Speaker 4: So I look at it in two parts. Actually, the
• 3:58
  Speaker 4: first one is we make extensive use of AI coding tools,
• 4:03
  Speaker 4: and this is something that has changed in the last
• 4:05
  Speaker 4: couple of years. So these tools are fundamentally changing how
• 4:11
  Speaker 4: software is being developed. So my team, for example, makes
• 4:15
  Speaker 4: extensive use of tools that you may be familiar with,
• 4:18
  Speaker 4: things like cursor, windsurf, gethub, copilot, plod code, you name it.
• 4:22
  Speaker 4: There are many more as well, And like any company,
• 4:26
  Speaker 4: there are those who jump into using these tools and
• 4:29
  Speaker 4: others who need to be coaxed and they need to
• 4:31
  Speaker 4: get over their skepticism and kind of using new ways
• 4:35
  Speaker 4: of doing things. But I, for one, personally believe that
• 4:39
  Speaker 4: there's a massive impact in improving software development using these tools.
• 4:46
  Speaker 4: But it is a new muscle that you need to develop,
• 4:48
  Speaker 4: and it's like you're doing pair programming, for example, with
• 4:52
  Speaker 4: another developer or assigning tasks to a junior developer. These
• 4:56
  Speaker 4: things don't come naturally for most people. You often need
• 4:59
  Speaker 4: to be trained to kind of incorporate these tools into
• 5:03
  Speaker 4: your planning, development and review process. So we've done that.
• 5:06
  Speaker 4: We've offered a lot of training in each of these tools,
• 5:10
  Speaker 4: and we monitor adoption and assist people who may be
• 5:14
  Speaker 4: having difficulty, and we're seeing the impact of all of
• 5:18
  Speaker 4: that major productivity gains of helping with our innovation. And
• 5:24
  Speaker 4: this is something your I T listeners I'm sure going
• 5:27
  Speaker 4: through as well. I'm sure they're under a lot of
• 5:29
  Speaker 4: pressure from management to kind of monitor AI tool use
• 5:34
  Speaker 4: the value that AI is getting to their employees, identify
• 5:37
  Speaker 4: shadow AI and doing this remains a challenge for them
• 5:42
  Speaker 4: and the same thing for us on that one. And secondly,
• 5:45
  Speaker 4: and more specifically to the question you asked me, for
• 5:49
  Speaker 4: a company like next thing, we are building AI products,
• 5:52
  Speaker 4: right and so when you're doing that, it's a very
• 6:00
  Speaker 4: There is fundamental ways in which software development for building
• 6:04
  Speaker 4: agents is very different from traditional software developments. People talk
• 6:09
  Speaker 4: about growing AI products rather than building it, and that's
• 6:14
  Speaker 4: very true in enterprise software. For example, I've never previously
• 6:19
  Speaker 4: had to think of a tone of interaction, for example,
• 6:22
  Speaker 4: whether whether it's being empathetic or not. For example, are
• 6:27
  Speaker 4: all the guardrails in place? These are things as enterprise
• 6:30
  Speaker 4: software developers we never really thought of. But if you're
• 6:34
  Speaker 4: building a conversational agent, that is core to what you're doing,
• 6:37
  Speaker 4: you need to kind of make it front and center.
• 6:41
  Speaker 4: And the other thing is as fast as fast as AI,
• 6:44
  Speaker 4: and these frontier models are evolving. The AI tools to
• 6:48
  Speaker 4: support the development of AI. Application development is also evolving
• 6:53
  Speaker 4: equally fast. So the way we did development two or
• 6:56
  Speaker 4: three years ago completely different from what we're doing now.
• 7:00
  Speaker 4: Vector databases, platforms like lang chain to do evails, create benchmarks,
• 7:06
  Speaker 4: test and validate agents. Model context Protocol MCP, which I'm
• 7:11
  Speaker 4: sure your listeners have heard of, incorporate various types of
• 7:16
  Speaker 4: data sources into this AI making it easy for all
• 7:19
  Speaker 4: of that to happen. All of those have changed how
• 7:22
  Speaker 4: you develop AI products. But for us at next Think Innovation,
• 7:27
  Speaker 4: it changes core to our culture and our teams have
• 7:31
  Speaker 4: really adapted very well to this change. It's a great
• 7:34
  Speaker 4: time to be in this business.
• 7:37
  Speaker 1: Dam Before before I ask this sort of related question
• 7:40
  Speaker 1: of Sam just, I'm interested if you think of your
• 7:44
  Speaker 1: career or your role in the year preceding the lordch
• 7:49
  Speaker 1: of you, the unveiling of AI. You know, two three
• 7:52
  Speaker 1: years ago, right when we all started, it started changing everything.
• 7:56
  Speaker 1: How if you had to put a percentage to it,
• 7:58
  Speaker 1: how different is your role now to what it was then,
• 8:01
  Speaker 1: and it might just be five percent, it might be tenpcent,
• 8:03
  Speaker 1: it might be thirty percent. I have no idea, but
• 8:05
  Speaker 1: if you had to pull a number out, what would
• 8:07
  Speaker 1: it be.
• 8:09
  Speaker 4: I don't believe my role has changed substantially. Ultimate Ultimately,
• 8:16
  Speaker 4: you are building products. You are building products that solve
• 8:19
  Speaker 4: a specific problem. The tools you use and how you
• 8:23
  Speaker 4: build it has changed, and that has changed over a
• 8:26
  Speaker 4: period of time. When you went, for example, from building
• 8:30
  Speaker 4: on prem products to using the cloud, that changed substantially.
• 8:34
  Speaker 4: And then now it's in that same order that you're doing.
• 8:38
  Speaker 4: You have to use new tools. You have to understand
• 8:42
  Speaker 4: how these tools work. It's not like your typical software
• 8:46
  Speaker 4: development like I just discussed. So you are essentially building
• 8:53
  Speaker 4: a role rather than building tools and workbenches for certain
• 9:00
  Speaker 4: roles for certain users. So that is where it's fundamentally different.
• 9:06
  Speaker 4: It's not just work workflows, database or dashboards and.
• 9:10
  Speaker 3: All of that.
• 9:11
  Speaker 4: It's all of that bundled together into an intelligent agent
• 9:16
  Speaker 4: that does everything, including the workflow that a human might
• 9:20
  Speaker 4: have done, albeit at a certain lower level than a
• 9:24
  Speaker 4: human does. So human supervision may still be important, but
• 9:28
  Speaker 4: that's the fundamental way in which I think AI has
• 9:31
  Speaker 4: changed software development.
• 9:35
  Speaker 1: Excellent, sam as AI has been playing a sort of
• 9:41
  Speaker 1: larger and larger role in the product in product development,
• 9:44
  Speaker 1: What have been the major learnings for you in your
• 9:46
  Speaker 1: role around managing the issues of accuracy and quality which
• 9:51
  Speaker 1: are inevitable concerns for anybody working with AI. What did
• 9:54
  Speaker 1: you find with the hardest aspect of ensuring reliable, consistent
• 9:58
  Speaker 1: experience when you no longer you're dealing with what you
• 10:01
  Speaker 1: might call deterministic software.
• 10:06
  Speaker 3: Yeah, that's quite an adventure and quite a new challenge, right.
• 10:10
  Speaker 3: The key around that is really building the proper evils evaluations,
• 10:16
  Speaker 3: which are basically standardized methods that you put in place
• 10:19
  Speaker 3: to measure how well the agent is performing, which expect
• 10:24
  Speaker 3: to specific tasks. Now, to understand that how it works,
• 10:29
  Speaker 3: we need to look at for a second, how an
• 10:31
  Speaker 3: agent is constructed, right, and maybe different from what people
• 10:35
  Speaker 3: might think. An agent and an AI agent is not
• 10:38
  Speaker 3: a monolith. It's not a one thing, but it is
• 10:41
  Speaker 3: fundamentally composed by two things. One is really the core
• 10:44
  Speaker 3: agent kind of your brain that is doing the reasoning.
• 10:48
  Speaker 3: But then the brain uses a set of tools in
• 10:52
  Speaker 3: order to perform tasks to look up knowledge and so
• 10:55
  Speaker 3: on and so forth. So when you need to test
• 10:58
  Speaker 3: all of that, you need to test both. You need
• 11:00
  Speaker 3: to test that the reasoning of the agent is sound,
• 11:03
  Speaker 3: but then you also need to make sure that the
• 11:05
  Speaker 3: tools that it's using are given the correct answers, and
• 11:09
  Speaker 3: you need to test that. You know, you need to
• 11:10
  Speaker 3: test for accuracy, for reliability, for speed. There are so
• 11:14
  Speaker 3: many things. You can have a great agent, but it
• 11:16
  Speaker 3: takes two minutes to answer to answer a question and
• 11:18
  Speaker 3: then it doesn't feel great anymore. Right, So the way
• 11:22
  Speaker 3: that you do that is you start by testing the tools,
• 11:24
  Speaker 3: because that's the easiest thing, you know. I give an example.
• 11:28
  Speaker 3: One of the tools of Assist is a tool that
• 11:31
  Speaker 3: takes text and transform that into an antul query, and
• 11:35
  Speaker 3: testing that is pretty easy because basically you have a
• 11:37
  Speaker 3: huge catalog of you know, couple's text query and then
• 11:40
  Speaker 3: you just check that it matches. So it's not so
• 11:43
  Speaker 3: hard for the agent is much it's much more difficult,
• 11:47
  Speaker 3: right because these agents they are performing overall complex tasks.
• 11:50
  Speaker 3: So the way that we did it is that we
• 11:52
  Speaker 3: had to create a dex benchmark, so something that really
• 11:58
  Speaker 3: it's a whole evaluation of what it means to be
• 12:01
  Speaker 3: a very capable expert dex analyst. It's composed by a
• 12:06
  Speaker 3: hundred questions, and then we get the agent to take
• 12:11
  Speaker 3: the test right, and then we use another agent to
• 12:14
  Speaker 3: evaluate how well our agent took the test, which is
• 12:17
  Speaker 3: kind of interesting. And then at the end you get
• 12:18
  Speaker 3: a score and you and you understand what is how
• 12:21
  Speaker 3: it's happening. Now, this whole thing is actually a huge
• 12:25
  Speaker 3: amount of work. It takes a lot of development effort
• 12:29
  Speaker 3: to come up with all of these things right. And
• 12:31
  Speaker 3: what I noticed that often companies they don't understand importance
• 12:35
  Speaker 3: of that. You see very often, you know, companies they
• 12:39
  Speaker 3: start just coding the agents without doing all of these
• 12:43
  Speaker 3: background work or creating the evaluations and the testing and
• 12:46
  Speaker 3: all of that. And then what happens is that maybe
• 12:48
  Speaker 3: one day they have a great model and the next
• 12:51
  Speaker 3: day there is an update and the model becomes terrible
• 12:54
  Speaker 3: and they don't even realize it because it's very hard.
• 12:57
  Speaker 3: You know, maybe it still performs beat in a certain
• 13:01
  Speaker 3: type of tasks, but other tasks became much much worse.
• 13:04
  Speaker 3: So I mean, my point is you need to invest
• 13:06
  Speaker 3: into that, and you need to create these things before end,
• 13:08
  Speaker 3: and you know, absorb the cost because it's the only
• 13:10
  Speaker 3: way to do a great thing. Then the other thing
• 13:14
  Speaker 3: is one of the fundamental aspects that people interacting with
• 13:18
  Speaker 3: agents are looking at is the concept of consistency. And
• 13:23
  Speaker 3: this is a fundamental pact with the people using the agents. Right,
• 13:28
  Speaker 3: if you ask the same question to the agent twice,
• 13:31
  Speaker 3: you're not gonna get the same answer. It's not gonna happen.
• 13:36
  Speaker 3: I you know, you guys, I think you're using GPT
• 13:38
  Speaker 3: or German or whatever. You live in that every day, right,
• 13:42
  Speaker 3: But for us as human it's not so easy to
• 13:45
  Speaker 3: digest that thing. And the way that you that you
• 13:48
  Speaker 3: that I look at to that is that basically I
• 13:50
  Speaker 3: look at agents like very smart individuals that have a
• 13:54
  Speaker 3: similar knowledge. But every time you ask the question, you
• 13:57
  Speaker 3: ask the question to another individual, so obviously you're going
• 14:00
  Speaker 3: to get a different answer. And here the key is
• 14:03
  Speaker 3: that has as humans, we can influence that.
• 14:08
  Speaker 4: So if you.
• 14:09
  Speaker 3: Ask a very broad and open question, then obviously the
• 14:14
  Speaker 3: answer that you get is going to be very different
• 14:17
  Speaker 3: every single time. However, if you're very specific with your
• 14:22
  Speaker 3: prompt and you really nail the type of answer that
• 14:26
  Speaker 3: you want to get, then you're going to see much
• 14:27
  Speaker 3: more consistency. In the way that these agents are answering.
• 14:31
  Speaker 3: So but I think, you know, lots of effort for
• 14:34
  Speaker 3: the people that are building the agents. But then also
• 14:36
  Speaker 3: as as a population of humans that are starting to
• 14:38
  Speaker 3: interact more and more with agents, we need to understand
• 14:41
  Speaker 3: their way of functioning and adapt our habits, expectations and
• 14:45
  Speaker 3: our way of interacting, you know, to be able to
• 14:47
  Speaker 3: make the most out of that.
• 14:50
  Speaker 1: Guardrails effectively.
• 14:51
  Speaker 5: You know.
• 14:52
  Speaker 1: I think, however, whatever area you're working, that's an essential
• 14:56
  Speaker 1: thing when it comes to AI, and be down from
• 14:59
  Speaker 1: a seat your perspective, what have been the biggest architectural
• 15:04
  Speaker 1: or operational challenges that emerged when you shifted from building
• 15:07
  Speaker 1: those deterministic systems to AI powered ones, And what have
• 15:12
  Speaker 1: been the biggest internal transformation for your team structure and
• 15:17
  Speaker 1: governance and operational kind of mode.
• 15:23
  Speaker 4: Yeah, that's a great question, and it kind of relates
• 15:25
  Speaker 4: to some things that Sam kind of talked about earlier
• 15:29
  Speaker 4: as well and what I'd said earlier. So we are building,
• 15:32
  Speaker 4: like I said, a role an agent in our case,
• 15:36
  Speaker 4: Sam was talking about, we're building, for example, the junior
• 15:39
  Speaker 4: support agent. You're not building the workbench the tools like
• 15:42
  Speaker 4: I was talking about earlier.
• 15:44
  Speaker 3: I mean in other.
• 15:45
  Speaker 4: Domains, you may be building a junior underwriter, a junior paralegal,
• 15:49
  Speaker 4: and so on. That's the fundamental way of thinking that
• 15:53
  Speaker 4: has changed. But there are some things that remain the same.
• 15:56
  Speaker 4: You still have to build a lot of scaffolding that
• 16:00
  Speaker 4: is required to run the software agent. Take Spark our
• 16:04
  Speaker 4: employee support agent, for example. We need to integrate it
• 16:07
  Speaker 4: into teams and we have to have tools to load
• 16:11
  Speaker 4: knowledge bases into them. We need to build a cockpit
• 16:15
  Speaker 4: to configure and manage the agent so that the user.
• 16:18
  Speaker 4: Remember this is a junior agent, so we need that
• 16:21
  Speaker 4: it admin to kind of help with the conversations, nudget
• 16:25
  Speaker 4: at times, ensure that the conversation can be improved with
• 16:31
  Speaker 4: additional knowledge bases, and so on. All of these require
• 16:34
  Speaker 4: typical software design and development processes we are used to,
• 16:39
  Speaker 4: and I'd say that's about half the work that needs
• 16:42
  Speaker 4: to be done. Often it's very easy to build a
• 16:45
  Speaker 4: POC with AI, but when you actually take the trouble
• 16:49
  Speaker 4: to get something into production, you find that there's all
• 16:52
  Speaker 4: of this that you need to build. This is the
• 16:55
  Speaker 4: deterministic part of software development remains the same. Different and
• 17:01
  Speaker 4: Sam kind of alluded to a lot of these things
• 17:03
  Speaker 4: is that the way the core AI agent or the
• 17:06
  Speaker 4: intelligence is developed, it is probabilistic, as you mentioned, and
• 17:12
  Speaker 4: part of developing your agent using AI and LLLM models
• 17:18
  Speaker 4: is to be able to address the tonality, empathy and
• 17:22
  Speaker 4: all the other good stuff that I talked about earlier.
• 17:24
  Speaker 4: If you're building a conversational agent, so you need to
• 17:28
  Speaker 4: kind of have that capability to be able to iteratively
• 17:31
  Speaker 4: build that, improve upon it, and so on. One of the
• 17:35
  Speaker 4: benefits that next Thing we had is, as you know,
• 17:38
  Speaker 4: we manage petabytes of data in the next think infinity
• 17:42
  Speaker 4: platform and machine learning. Even before all the generitive AI
• 17:47
  Speaker 4: became popular, was core to how we analyzed data. We
• 17:52
  Speaker 4: had built the platform essentially to make accessing and analyzing
• 17:57
  Speaker 4: this type of data easy both for humans and for machines,
• 18:01
  Speaker 4: like a unified data model. We have something called next
• 18:04
  Speaker 4: thing query language NQL. So when we got to generat VII,
• 18:08
  Speaker 4: this isn't the after twenty two, twenty three, twenty four.
• 18:12
  Speaker 4: We started with the most difficult problem first, basically building
• 18:17
  Speaker 4: a natural language interface to that query language I talked
• 18:20
  Speaker 4: about to NQL, and this is something that LMS cannot
• 18:24
  Speaker 4: do quite as effectively on their own. So we had
• 18:27
  Speaker 4: to fine tune and an LLM to be able to
• 18:30
  Speaker 4: do that, and I'm happy to say that the one
• 18:33
  Speaker 4: that we've got passes the next thing, NQL certification, which
• 18:38
  Speaker 4: isn't an easy test, I may add, So once we
• 18:41
  Speaker 4: mastered that, the rest became a little easier. We built
• 18:45
  Speaker 4: a number of additional tools to assist the reasoning process.
• 18:49
  Speaker 4: A tool to understand the user context, who you are,
• 18:53
  Speaker 4: which is your device, One to get kind of a
• 18:56
  Speaker 4: time based view on things. What happened on the devices yesterday,
• 19:00
  Speaker 4: what happened in the last week, what happened in the
• 19:02
  Speaker 4: last two weeks. Helped these freezes and crashes been a
• 19:06
  Speaker 4: recurring thing if you need to kind of find what
• 19:08
  Speaker 4: the problem is. And we combined all of that, including
• 19:13
  Speaker 4: the knowledge bases that you may have to the foundational
• 19:17
  Speaker 4: knowledge LLLM in our case anthropic, and you had this
• 19:21
  Speaker 4: junior support agent who's getting smarter every day, so essentially
• 19:27
  Speaker 4: software development intelligent AI agent that is conversational. And we
• 19:32
  Speaker 4: have this junior software agent jor's support agent. But challenges remain.
• 19:40
  Speaker 4: As Sam mentioned, we need extensive benchmarks. Every time you
• 19:45
  Speaker 4: move from one version of the LM to another. There
• 19:50
  Speaker 4: are subtle things that change that impact conversations, how things happen.
• 19:54
  Speaker 4: So we need to have all of those benchmarks. We
• 19:57
  Speaker 4: need to make sure every time that the guardrails remain
• 20:00
  Speaker 4: effective and most importantly, that you can trust that you
• 20:04
  Speaker 4: have a secure and accurate agent and you're protected from
• 20:09
  Speaker 4: all of the hallucinations and other things that lms are
• 20:12
  Speaker 4: known for. Hopefully that gives you the picture of what's changed.
• 20:18
  Speaker 3: And by the way, by now, the support agent is
• 20:21
  Speaker 3: very senior, right, he's no longer he's not a.
• 20:23
  Speaker 5: J new junior agent. He's graduated from the helpout.
• 20:26
  Speaker 1: We should have him on LAM. You know we can
• 20:29
  Speaker 1: do We did that a while back.
• 20:30
  Speaker 5: We should Okay, you have that thing.
• 20:32
  Speaker 1: We did, we had a system.
• 20:35
  Speaker 5: So Tom, I want to highlight what a unique opportunity
• 20:38
  Speaker 5: this is for folks that are listening in to have
• 20:40
  Speaker 5: not just our chief product officer, which is always a highlight,
• 20:45
  Speaker 5: but also our CTO and Vida and Sam here together
• 20:49
  Speaker 5: to give that really unique perspective of the next thing, history,
• 20:53
  Speaker 5: the next thing, foundation, our approach, but then also the
• 20:55
  Speaker 5: outcomes for our customers. So, guys, before we get into
• 20:58
  Speaker 5: a little bit of Q and a I want to
• 21:00
  Speaker 5: set kind of the transition from great information on next
• 21:04
  Speaker 5: Think foundation and our approach into the outcomes. So I
• 21:08
  Speaker 5: think the main point that we collectively have been singing
• 21:10
  Speaker 5: here at next Think for a while, and the rise
• 21:12
  Speaker 5: of AI makes even more important and makes it even
• 21:17
  Speaker 5: more clear. There's three dimensions of corporate AI right, and
• 21:21
  Speaker 5: they define some really important directions that every IT team
• 21:25
  Speaker 5: needs to pursue to really drive successful digital transformation. The
• 21:29
  Speaker 5: first is AI for it, and this is automation taken
• 21:34
  Speaker 5: to the next level, the creation of autonomous experiences and
• 21:38
  Speaker 5: the enablement of preventative operations. It creates that IT org
• 21:42
  Speaker 5: that shifts from reactive to proactive ultimately to zero tickets.
• 21:46
  Speaker 5: So AI for it is the first dimension. The second
• 21:50
  Speaker 5: is AI for employees. This is about coaching and the
• 21:53
  Speaker 5: flow of work. If not you talked about it a
• 21:54
  Speaker 5: little bit. With that empathy and sympathy for employees and
• 21:57
  Speaker 5: the struggle that they have the self help and conversational
• 22:00
  Speaker 5: resolutions of both technical issues and knowledge gaps, it helps
• 22:05
  Speaker 5: get employees the instant contextual support they need and drive
• 22:09
  Speaker 5: that AI adoption. So AI for employees. The third is
• 22:14
  Speaker 5: AI for the business that it supports creates the visibility
• 22:19
  Speaker 5: needed for a proper governance and achievement of ROI. We're
• 22:22
  Speaker 5: hearing all over the place now. One of the biggest
• 22:24
  Speaker 5: issues with AI is the return on the investment. Is
• 22:28
  Speaker 5: it worth the money we're spending to get back the benefits?
• 22:32
  Speaker 5: But they I for the business creates that enterprise wide
• 22:36
  Speaker 5: intelligence layer for secure governed high ROI transformation. So I
• 22:40
  Speaker 5: wanted to set that foundation because I want to ask
• 22:42
  Speaker 5: you each a question about each of those three pillars.
• 22:45
  Speaker 5: So stamp and let's start. And I'm going to kind
• 22:48
  Speaker 5: of give you a question, but feel free to add
• 22:51
  Speaker 5: your thoughts wherever you want to take it. So AI
• 22:54
  Speaker 5: for IT, experience management, and proactive IT. And I'm looking
• 22:59
  Speaker 5: at my history, Salmon, you were out with me when
• 23:02
  Speaker 5: I was a customer and looked at how we were
• 23:03
  Speaker 5: implementing at the Hartford. Right, it was still a pretty
• 23:07
  Speaker 5: heavy lift back then AI doing discovery and analysis, but
• 23:13
  Speaker 5: the automation still needed a fair amount of it touch. Right,
• 23:16
  Speaker 5: you had to build scripts, you had to do the
• 23:18
  Speaker 5: next level of what you wanted to implement. We're now
• 23:21
  Speaker 5: moving to truly autonomous management where IT teams can move
• 23:26
  Speaker 5: from creating problems to preventing them and as a customer,
• 23:31
  Speaker 5: my team's able to do that without modern AI capabilities,
• 23:34
  Speaker 5: but it was an effort. What innovations make preventative operations
• 23:39
  Speaker 5: and the path to zero tickets not only achievable but
• 23:42
  Speaker 5: hugely efficient for it teams today. And then a part
• 23:45
  Speaker 5: b on that what's next? What still needs to be solved?
• 23:49
  Speaker 3: Yeah, it's a great topic. Look, when you look at
• 23:53
  Speaker 3: these there are really two aspects, right, there is proactive
• 23:57
  Speaker 3: work and then there is reactive work. And the way
• 24:00
  Speaker 3: that we define it is proactive work is everything that
• 24:03
  Speaker 3: happens without an employee having to say anything, and reactive
• 24:09
  Speaker 3: work is it can still be automated, but the trigger
• 24:13
  Speaker 3: for doing something is the employee saying something to your
• 24:15
  Speaker 3: own And then of course these two dimensions. When you
• 24:19
  Speaker 3: when you want to automate, you have two dimensions. Right,
• 24:21
  Speaker 3: there is a deterministic or augentic. In the past, it
• 24:25
  Speaker 3: was there doministic in the future is more augentic, but
• 24:27
  Speaker 3: there is still a space for deterministic. So when we
• 24:32
  Speaker 3: look at at the pure deterministic, what is the problem
• 24:36
  Speaker 3: with that? For it? The problem is that every problem,
• 24:41
  Speaker 3: every incidence that you want to solve, you need to
• 24:46
  Speaker 3: hardcod it. Give an example, you want to make sure
• 24:50
  Speaker 3: that people never have a problem with the VPN, and
• 24:53
  Speaker 3: so somebody needs to write a scripts that gets in
• 24:56
  Speaker 3: there and say, Okay, if the VPN is not installed,
• 24:58
  Speaker 3: then you need to install it. If it's not connected,
• 25:00
  Speaker 3: you need to connect it. If it's not authenticated, then
• 25:02
  Speaker 3: you need to authenticate it. If that doesn't work, you
• 25:05
  Speaker 3: need to look at the configurations and so forth. Now
• 25:09
  Speaker 3: that's okay because you can still avoid a huge number
• 25:12
  Speaker 3: of problems. Right, But all of that work that you
• 25:15
  Speaker 3: put in there only tikeses VPN problems. What about everything else?
• 25:20
  Speaker 3: And here it's interesting because when you look at all
• 25:22
  Speaker 3: of these things, the eighty twenty rule applies, But the
• 25:27
  Speaker 3: tea teams start still spending a huge amount of time
• 25:29
  Speaker 3: on the long tail, and those often are the most
• 25:32
  Speaker 3: difficult things right pastword re SATs. Okay, now everybody has
• 25:35
  Speaker 3: automation for that, but what about the five hundred other
• 25:38
  Speaker 3: issues that are happening from time to time. Then the
• 25:42
  Speaker 3: other problem is also even if you have the automation,
• 25:44
  Speaker 3: you still need to know when to trigger it. Right,
• 25:48
  Speaker 3: what is the signal that is telling you that you
• 25:49
  Speaker 3: need to do that with the eye. The difference is
• 25:52
  Speaker 3: the following. You just need to build the building blocks.
• 25:57
  Speaker 3: A building block is install an application. Another building block
• 26:02
  Speaker 3: is start an application, another one is set a configuration.
• 26:07
  Speaker 3: But everything else is parametric. And what is important is
• 26:11
  Speaker 3: that your chestration of these building blocks, you don't do
• 26:15
  Speaker 3: it by hand anymore, but the agent, the AI agent
• 26:19
  Speaker 3: on the fly does that for you. And so you
• 26:23
  Speaker 3: go from having to write five hundred automation to cover
• 26:28
  Speaker 3: five hundred problems to maybe writing fifty building blocks to
• 26:33
  Speaker 3: cover five thousand problems. So you can see all much
• 26:36
  Speaker 3: time it is going to save around all of that.
• 26:39
  Speaker 3: And by the way, the building blocks the other advantage
• 26:41
  Speaker 3: is that the building blocks are very similar company to company,
• 26:45
  Speaker 3: So even for these there is a much more reasonability. Right,
• 26:48
  Speaker 3: it's less it's less specific stuff. So what is the
• 26:52
  Speaker 3: outcome of that is that I take an automic more
• 26:56
  Speaker 3: and you know, ike personal. I can focus on approving
• 27:00
  Speaker 3: the building blocks and deciding what are the boundaries in
• 27:03
  Speaker 3: which AI can operate rather than having to do all
• 27:07
  Speaker 3: of these manual efforts. And then obviously the other outcoments
• 27:11
  Speaker 3: that the experience for emplass is so much better. And
• 27:14
  Speaker 3: this is similar between between practive and reactive. By the way,
• 27:17
  Speaker 3: the is kind of the same principles. Right now, we're
• 27:20
  Speaker 3: saying that there's still a space for deterministic actions. Is
• 27:25
  Speaker 3: when you want part of your process to be executed
• 27:29
  Speaker 3: every time exactly one hundred percent in the same way.
• 27:33
  Speaker 3: Give an example, let's say that you want to you know,
• 27:36
  Speaker 3: you want to automate software provisioning, but for programs that
• 27:40
  Speaker 3: are particularly expensive or things that are not usually used
• 27:43
  Speaker 3: within our department, you want manage our approval. Right, that's
• 27:47
  Speaker 3: not something that necessarily your your key to leave the
• 27:51
  Speaker 3: AI choosing. You want a deterministic thing that looks at
• 27:54
  Speaker 3: who is exactly the manager of this person? Right, you
• 27:57
  Speaker 3: want to execute the best step and there in that
• 28:00
  Speaker 3: In that case, it's where you're combined to a process.
• 28:03
  Speaker 3: So one of your building blocks is get approval to
• 28:07
  Speaker 3: install this software, and within that you still have a
• 28:10
  Speaker 3: more traditional workflow. That is, you know, if the nets
• 28:13
  Speaker 3: or or whatever visual editor kind of thing. Now, the
• 28:18
  Speaker 3: difference is that even in that case, you can simplify
• 28:22
  Speaker 3: by using AI to generate these workflows, so that rather
• 28:26
  Speaker 3: than having a developer right in the code or an
• 28:29
  Speaker 3: engineer put in the building blocks, THEE is going to
• 28:33
  Speaker 3: prompt it and vibe coded, and then that part is
• 28:35
  Speaker 3: gonna get It's gonna get easier and faster as well. Right, So, yeah,
• 28:40
  Speaker 3: that's a bit the situation. We did a huge amount
• 28:43
  Speaker 3: of work on the reactive piece agentic. We have an
• 28:49
  Speaker 3: incredible solution for the proactive piece, deterministic. One of the
• 28:54
  Speaker 3: things that we're going to focus on next year is
• 28:57
  Speaker 3: to do the proactive piece in a agentic way.
• 29:00
  Speaker 1: Sweat.
• 29:02
  Speaker 3: It's going to be very exciting.
• 29:04
  Speaker 5: Yeah, I think it's for our IT audience. I think
• 29:07
  Speaker 5: it's really intriguing to take some of those fundamental practices
• 29:13
  Speaker 5: of software coding into delivery of it. Right, the concept
• 29:18
  Speaker 5: of core code and subroutines, the least common denominator and
• 29:21
  Speaker 5: top down programming into the deterministic subroutines that can be
• 29:26
  Speaker 5: either reused or can be specific to a unique problem,
• 29:30
  Speaker 5: and then come back to the core part of it.
• 29:32
  Speaker 5: I think the building block analogy is perfect for it.
• 29:36
  Speaker 5: Automation and to have that agentic AI in the background
• 29:39
  Speaker 5: doing some of that decision making for you is really intriguing.
• 29:45
  Speaker 5: For the second pillar, employees are interacting with AI directly
• 29:51
  Speaker 5: through coaching and in the flow of work, and for
• 29:54
  Speaker 5: conversational remediation. I've never really liked, just as maybe a
• 29:59
  Speaker 5: bit of a side commentary, this construct of shift left
• 30:03
  Speaker 5: a big ITIL proponent. It puts structure and rigor to
• 30:07
  Speaker 5: things that we do in IT. But shift Left over
• 30:10
  Speaker 5: the years basically became outsourcing it to the employee. Right,
• 30:14
  Speaker 5: it is too expensive, Let I employees fix their own stuff.
• 30:17
  Speaker 5: They want to fix their own stuff anyway. But that
• 30:19
  Speaker 5: shift left for me was never ideal. ITOL was really
• 30:25
  Speaker 5: conceptually right in the fact that you needed to get
• 30:29
  Speaker 5: the solution information closest to the source of final remediation.
• 30:34
  Speaker 5: If only level one, if only level three knows about
• 30:37
  Speaker 5: the fix, then essentially level one and level two are useless.
• 30:41
  Speaker 5: Get the solution closer to where the problem happened. It
• 30:45
  Speaker 5: avoids escalations, cost, and delays. How do you think about
• 30:49
  Speaker 5: designing AI that employees actually trust and want to use
• 30:55
  Speaker 5: versus feeling like they're just a delegated member of the
• 30:58
  Speaker 5: IT team. And then a part be to that, How
• 31:01
  Speaker 5: does next thing's approach accelerate both zero tickets and broader
• 31:05
  Speaker 5: AI adoption?
• 31:07
  Speaker 4: Great question, Tim and one hundred percent wholeheartedly agree with
• 31:12
  Speaker 4: your issue with shift left. To me, at its worst,
• 31:17
  Speaker 4: shift left is like telling users you think this is easy,
• 31:20
  Speaker 4: here go fix this issue yourself.
• 31:22
  Speaker 5: And it's one we broke, right, it's it broke you
• 31:25
  Speaker 5: and now we're going to tell you how to fix this.
• 31:27
  Speaker 4: And by the way, here are a few kbs right,
• 31:30
  Speaker 4: you're welcome. So that to me is shift left.
• 31:33
  Speaker 5: And here's a servey so to ask you how we did.
• 31:36
  Speaker 4: Not exactly a recipe for happy employees. Well so, but
• 31:41
  Speaker 4: on the other hand, people really don't like to open tickets.
• 31:46
  Speaker 4: I mean, that's got to be the last resort for people.
• 31:49
  Speaker 4: It's like I just cannot solve this. I need to
• 31:52
  Speaker 4: escalate this. And the other thing is, I mean if
• 31:56
  Speaker 4: you look at the digital first generation, I mean they
• 31:59
  Speaker 4: prefer were to resolve the issues themselves. So you're kind
• 32:04
  Speaker 4: of stuck whether you wire it or not. You better
• 32:07
  Speaker 4: create a good shift lift strategy because that's what they're
• 32:10
  Speaker 4: calling for, but not the old way of doing things.
• 32:14
  Speaker 4: So a generic so what do most people do? Hey,
• 32:18
  Speaker 4: I've got a chat pot. Now it's going to be
• 32:20
  Speaker 4: much better. So I'm going to throw in a few
• 32:23
  Speaker 4: knowledge bases. This barely takes you halfway there. We've got
• 32:29
  Speaker 4: expercience with that. I mean we have device information, application information,
• 32:35
  Speaker 4: network performance data, all of the things that is core
• 32:39
  Speaker 4: to what we do to troubleshoot issues. And in addition,
• 32:42
  Speaker 4: you may need to access the Internet for specific information.
• 32:46
  Speaker 4: Hey is the system really down? I may be looking
• 32:49
  Speaker 4: for all kinds of network information, but guess what the
• 32:51
  Speaker 4: application is having an issue? And most importantly, you need
• 32:57
  Speaker 4: automated remediations that that execute securely and reliably these actions
• 33:03
  Speaker 4: that Sam was talking about earlier, and you need all
• 33:07
  Speaker 4: of this in what I was talking about early, having
• 33:10
  Speaker 4: an agent which is conversational, which does it with empathy
• 33:13
  Speaker 4: in the right tone, and all of that that stuff
• 33:15
  Speaker 4: was also important. So that's exactly what we did and
• 33:19
  Speaker 4: what we've built with Next think in next thing with Spark,
• 33:22
  Speaker 4: basically it understands your problem with real device level data.
• 33:29
  Speaker 4: Does multi turn reasoning with that information? Meaning I reasoned
• 33:33
  Speaker 4: with the data. I asked another question. I can create
• 33:36
  Speaker 4: a plan. I incorporate your specific knowledge basis that you
• 33:41
  Speaker 4: have that is owned by you into your reasoning, and
• 33:45
  Speaker 4: then with your permission, with the user's permission, runs the
• 33:50
  Speaker 4: scripts to remediate the issue. It eliminates that friction and
• 33:54
  Speaker 4: gets you back to work and hopefully puts a smile
• 33:58
  Speaker 4: on your face. It's basically today's it agent, not your father's.
• 34:03
  Speaker 4: I till if I may say, sir.
• 34:06
  Speaker 5: So.
• 34:07
  Speaker 4: Also another thing, I think this was the other point
• 34:09
  Speaker 4: that you're raised, which is today work gets done in
• 34:14
  Speaker 4: SaaS applications and your typical employee needs to use multiple applications,
• 34:20
  Speaker 4: and when they have difficulty using features within these applications,
• 34:24
  Speaker 4: they waste a bunch of time trying to figure out
• 34:26
  Speaker 4: how to fix things. And if they cannot do that,
• 34:30
  Speaker 4: there isn't a ticket that coming your way. Maybe there is,
• 34:33
  Speaker 4: that's probably a good thing, but most of the time
• 34:36
  Speaker 4: they just stop using the application if they can get
• 34:38
  Speaker 4: away with it, and you're not getting the productivity you
• 34:42
  Speaker 4: expected when you invested in all these tools and applications.
• 34:46
  Speaker 4: So we have again like spark, an AI based application
• 34:51
  Speaker 4: and adoption guides that assist users in their journey to
• 34:54
  Speaker 4: effectively use these applications and tools. You provide so empower
• 35:00
  Speaker 4: or in employees to solve issues themselves and identifying patterns
• 35:05
  Speaker 4: because remember you talked about Sam talked about one is
• 35:08
  Speaker 4: having employees resolve it themselves. If the employee identified one
• 35:14
  Speaker 4: problem and that exists, the same thing exists with ten
• 35:17
  Speaker 4: thousand other devices, do you want to wait for those
• 35:21
  Speaker 4: tickets to be created. That's an opportunity on the proactive side,
• 35:26
  Speaker 4: in an agentic way to identify that there are ten
• 35:29
  Speaker 4: thousand devices for which I can run that same remediation
• 35:33
  Speaker 4: that I ran for that user and solve it proactively.
• 35:37
  Speaker 4: And that's next thing's approach to zero tickets both give
• 35:40
  Speaker 4: the user the capability that they need identify patterns eliminated
• 35:46
  Speaker 4: across your estate.
• 35:47
  Speaker 5: I think that's a really important distinction to make zero
• 35:51
  Speaker 5: tickets is zero reactive incident management tickets. Can you imagine
• 35:57
  Speaker 5: finding a thousand blue screens and then opening up one
• 36:00
  Speaker 5: thousand tickets so the automation can go solve a thousand issues.
• 36:04
  Speaker 5: We're now flipping to proactive automated problem management, one ticket,
• 36:09
  Speaker 5: automated automatically resolving all of those issues and then closing
• 36:12
  Speaker 5: it out with the record. So it's just moving to
• 36:15
  Speaker 5: a more efficient logical way to operate. And then preventative
• 36:19
  Speaker 5: is don't cause the blue screens to begin with SAM.
• 36:24
  Speaker 5: The third pillar is why we're in this business right
• 36:28
  Speaker 5: It's sometimes you have to do it for it's sake,
• 36:31
  Speaker 5: but it really should be doing it for the business sake,
• 36:35
  Speaker 5: make the business more profitable, drive revenue secure, and manage risk.
• 36:41
  Speaker 5: This is all about This third pillar is all about
• 36:44
  Speaker 5: total visibility, governance, achievable ROI and as companies scale these
• 36:50
  Speaker 5: AI initiatives to get the most out of this technology,
• 36:54
  Speaker 5: lack of ROI is a huge problem and with growing
• 36:58
  Speaker 5: skepticism about how to really rely on this ad scale,
• 37:02
  Speaker 5: Like when I was talking, people are just going to
• 37:04
  Speaker 5: stop using it. If there's another way to do it,
• 37:07
  Speaker 5: it might be perceived as easier if AI is tough
• 37:10
  Speaker 5: to learn or giving them trouble. What blind spots do
• 37:12
  Speaker 5: you see emerging now? And how is next thing helping
• 37:16
  Speaker 5: orgs quantify impact on adoption, govern AI safely and ultimately
• 37:22
  Speaker 5: drive that enterprise wide value.
• 37:25
  Speaker 3: There is a lot to say about this topic. You know,
• 37:28
  Speaker 3: before getting into that, maybe we need to make the
• 37:31
  Speaker 3: distinction between two things. There is what we call everyday AI,
• 37:36
  Speaker 3: which is your two sixty five copilot chg APT and
• 37:40
  Speaker 3: some support. And you know when you look at industries
• 37:44
  Speaker 3: and those are internal analysis that can lead to five
• 37:47
  Speaker 3: to twenty percent efficiency improvement. Right, So it means that
• 37:53
  Speaker 3: maybe out of my working day, I get one hour
• 37:55
  Speaker 3: back of productivity because I can do many tasks more quickly.
• 38:00
  Speaker 3: And then there is transformational AI, which is as important
• 38:03
  Speaker 3: for the company. And this is really the things that
• 38:06
  Speaker 3: are making a major shift in an entire process. For instance,
• 38:11
  Speaker 3: it could be you know, automating your customer support. You
• 38:15
  Speaker 3: getting rid of an entire minor process. For an engineer,
• 38:19
  Speaker 3: it could be code generation. For a product manager, it
• 38:21
  Speaker 3: could be vibe coding. You know, in my team, product
• 38:24
  Speaker 3: managers now can do prototypes without having to borrow engineering time,
• 38:29
  Speaker 3: which is a great acceleration. So that's that's transformationally I
• 38:33
  Speaker 3: and I think it's important to look at them a
• 38:35
  Speaker 3: bit separate because they're kind of different beasts. Right when
• 38:38
  Speaker 3: it comes to EVERYDAYI one of the main problems for
• 38:43
  Speaker 3: companies is being able to understand adoption. That's the starting point, right,
• 38:50
  Speaker 3: If you want to get to that five to twenty
• 38:51
  Speaker 3: percent efficiency improvements, you need people to adopt it. But
• 38:56
  Speaker 3: how do you manage it, how do you measure it,
• 38:58
  Speaker 3: how do you know what's going on? It's very hard.
• 39:02
  Speaker 3: Just by the way, why we launched our aidrive product
• 39:05
  Speaker 3: and why in two months we got to fifty percent
• 39:08
  Speaker 3: customer penetration because it's such a fundamental thing. Right. But
• 39:12
  Speaker 3: then at the point in which you can start measure adoption,
• 39:16
  Speaker 3: which means you can identify where you need to do enablement,
• 39:18
  Speaker 3: where you need to deploy licensees and all of that,
• 39:20
  Speaker 3: then the second question you mentioned that is absolutely important.
• 39:24
  Speaker 3: What's the ROI? Am I in the five percent? Am
• 39:29
  Speaker 3: I in the twenty percent? How many tools do I need?
• 39:32
  Speaker 3: How many licenses do I need? This is additional and money, right.
• 39:36
  Speaker 3: I mean, nobody had budget for AI two years ago.
• 39:39
  Speaker 3: They had to create these budgets. They needed to come
• 39:41
  Speaker 3: from somewhere, right, And so you need to be able
• 39:44
  Speaker 3: to quantify that. That's of fundamental importance, and there are
• 39:47
  Speaker 3: many techniques. We picked an approach that is a mix
• 39:50
  Speaker 3: between analytics and also asking feedback to people because at
• 39:54
  Speaker 3: the end of the day, you know, you can come
• 39:56
  Speaker 3: up with very complex math formulas. Nobody's gonna believe that.
• 40:00
  Speaker 3: But if I'm telling you, hey, today saved one hour
• 40:04
  Speaker 3: because they have to write that wrong document, I mean
• 40:06
  Speaker 3: that's a factor, right, I said one hour. And then
• 40:09
  Speaker 3: the other aspect on EVERYDAI is giving AI adoption teams,
• 40:16
  Speaker 3: which is something that we're starting to see in every
• 40:18
  Speaker 3: large enterprise company, the ability to know what they need
• 40:23
  Speaker 3: to do to drive more adoption, so to come up
• 40:26
  Speaker 3: with a plan that they can track over time that
• 40:29
  Speaker 3: will drive more adoption. And this once again something that
• 40:31
  Speaker 3: you do based on the feedback. But just this thing
• 40:35
  Speaker 3: about knowing where you are is so important, you know,
• 40:37
  Speaker 3: I was thinking about an analogis if you don't have
• 40:40
  Speaker 3: a way to measure it. It's like if I ask
• 40:42
  Speaker 3: your team to take your motorcycle drive one thousand miles
• 40:47
  Speaker 3: without Google Maps and without traffic signs, and every time
• 40:51
  Speaker 3: that you want to know where you are, you need
• 40:53
  Speaker 3: to stop and ask somebody on the road. That's how
• 40:55
  Speaker 3: people are doing it. You need to ask people every
• 40:58
  Speaker 3: time it doesn't work, it's think right. So yeah, that
• 41:03
  Speaker 3: was a big part of what of what we were
• 41:05
  Speaker 3: doing right now. By the way, on this whole topic,
• 41:09
  Speaker 3: there is another opportunity of discovering which tools people are
• 41:13
  Speaker 3: adopting on their own. And you remember ten years ago
• 41:18
  Speaker 3: everybody was talking about shadow shadow I tea now now
• 41:23
  Speaker 3: as shadowy I and you know, be forward looking people
• 41:27
  Speaker 3: at the time were saying, look, shadow I tea is
• 41:29
  Speaker 3: not a problem. Shadow It is a huge opportunity because
• 41:33
  Speaker 3: it's telling me, you know, the smart folks in my organization,
• 41:37
  Speaker 3: what are they doing to get more more productive? Right,
• 41:40
  Speaker 3: shadowy is the same thing, is absolutely the same thing.
• 41:44
  Speaker 3: So for me that obviously there are security risks and
• 41:47
  Speaker 3: data data risks, and but it can be a big
• 41:50
  Speaker 3: opportunity because people tend to get creative and when they
• 41:54
  Speaker 3: want to get something done. So that's every day. The
• 41:58
  Speaker 3: other aspect is transforming the eye. And here the number
• 42:03
  Speaker 3: one question for every company is where is my opportunity?
• 42:09
  Speaker 3: Where do I need to go and invest to make
• 42:12
  Speaker 3: a difference. What are those processes, what are those things
• 42:15
  Speaker 3: that are leading to the biggest return. And like in
• 42:18
  Speaker 3: my previous analogy, the only way that they have today
• 42:22
  Speaker 3: is to go and ask people they do service to
• 42:26
  Speaker 3: try to figure out where the bottomnecks are. And you know,
• 42:30
  Speaker 3: then they pick, they pick some things that they feel
• 42:34
  Speaker 3: is gonna is going to generate a return. They automate,
• 42:36
  Speaker 3: they bring an eye, but then you don't have an
• 42:38
  Speaker 3: objective way of measure before after whether it's really gener
• 42:42
  Speaker 3: everyting that value that you're hoping You know, are you
• 42:45
  Speaker 3: are you putting your efforts in the right place, because
• 42:48
  Speaker 3: let's face it, embedding the eye is it doesn't come
• 42:51
  Speaker 3: from free. You might need additional tools, you need the effort,
• 42:54
  Speaker 3: you need, developers, you need, you need to put everything together.
• 42:58
  Speaker 3: And so I believe that huge opportunity in twenty twenty
• 43:02
  Speaker 3: six we'd be to automate that part. So to provide
• 43:06
  Speaker 3: the reference of information that says, look, those are the
• 43:10
  Speaker 3: ten things in where we're seeing that there is huge friction.
• 43:14
  Speaker 3: And if you put AI or if you put automation in,
• 43:16
  Speaker 3: then in those ten things, you're gonna have a huge gain.
• 43:21
  Speaker 3: And we're also gonna give you a way to track
• 43:24
  Speaker 3: how the gain proceeds. It's something that we're gonna look at.
• 43:28
  Speaker 3: You know, maybe maybe you can invite me in a
• 43:31
  Speaker 3: few months in another dexter and we're gonna talk about
• 43:33
  Speaker 3: that topic.
• 43:34
  Speaker 5: Yeah, absolutely, I love it.
• 43:38
  Speaker 1: I really like the next thing started talking about these
• 43:40
  Speaker 1: three pillars. I think they flesh out in a very
• 43:43
  Speaker 1: lucid way, just just just the kind of the pressure
• 43:49
  Speaker 1: on it leaders in particular, right, because if you don't
• 43:52
  Speaker 1: tend to all three of them, you're not doing AI
• 43:55
  Speaker 1: transformation properly. I think that's quite clear, and I love
• 43:58
  Speaker 1: where you guys have unpacked them so far. But let's
• 44:01
  Speaker 1: end with view Verdan. I mean thinking about that complexity,
• 44:05
  Speaker 1: even intimidating complexity, how is it? The next thing specifically
• 44:09
  Speaker 1: makes them all realistic and realizable and achievable together.
• 44:16
  Speaker 4: So one of the things that next thing that we've
• 44:19
  Speaker 4: always put front and center is that we need we're
• 44:23
  Speaker 4: driven to ensure good IT outcomes for our clients. So
• 44:29
  Speaker 4: we started with I mean, you need the ability to
• 44:32
  Speaker 4: collect data, be able to form the foundation for all
• 44:37
  Speaker 4: of the AI you need to put in place. Next thing,
• 44:40
  Speaker 4: Infinity is the platform that can be easily deployed and
• 44:44
  Speaker 4: you can start getting value very quickly. So with AI,
• 44:48
  Speaker 4: we just kind of continued that heritage that we had,
• 44:53
  Speaker 4: and essentially we've built AI agents that are easy to
• 44:57
  Speaker 4: deploy and natural to use for all three pillars that
• 45:01
  Speaker 4: we talked about, for example, be's workspace and assist for
• 45:05
  Speaker 4: IT admins. Sam went to do a lot of detail
• 45:09
  Speaker 4: out there, Spark for employees and it self remediation. I
• 45:14
  Speaker 4: talked about that quite a bit, and what Sam was
• 45:17
  Speaker 4: just talking about AI drive to measure and drive AI
• 45:22
  Speaker 4: adoption and adopt for optimal application used in your enterprise,
• 45:28
  Speaker 4: so it's you already have the basis. If you already
• 45:32
  Speaker 4: are a client, you've deployed it, you can just enable
• 45:35
  Speaker 4: it and start using it. And on top of that,
• 45:37
  Speaker 4: for anything in the enterprise, we offer extensive online training
• 45:41
  Speaker 4: and support to ensure success. So we've made sure that
• 45:45
  Speaker 4: with all of the agent agentic stuff that we're depulying,
• 45:48
  Speaker 4: that we support that with whatever documentation to get it
• 45:52
  Speaker 4: deployed quickly. Our services teams and CSMs are there from
• 45:57
  Speaker 4: the start and have been trained to help. So the
• 46:00
  Speaker 4: bottom liners, we're there to ensure you realize the value
• 46:04
  Speaker 4: and get the products up and running quickly. So everything
• 46:09
  Speaker 4: that we've always done, we're now doing with AI for you.
• 46:15
  Speaker 1: Basically very well said, and I have to stress that
• 46:18
  Speaker 1: anybody who was listening and maybe I hadn't heard about
• 46:22
  Speaker 1: Spark before, it's possible, right you missed experience, or you
• 46:25
  Speaker 1: just haven't come across it and you think I want
• 46:28
  Speaker 1: to know more, you are listening to the right show,
• 46:30
  Speaker 1: because it's going to be Spark Month in about a
• 46:33
  Speaker 1: week or two. I think we've got a series of
• 46:34
  Speaker 1: incredible shows lined up where Tim and I and the
• 46:37
  Speaker 1: reality Biased team are going to be just exploring and
• 46:40
  Speaker 1: unpacking in detail of this really remarkable game changing innovation
• 46:45
  Speaker 1: that we've teased a little bit today. But I think
• 46:48
  Speaker 1: for now, this is a good way to kick off
• 46:51
  Speaker 1: for you, gentlemen, the Dan thanks so much for your
• 46:54
  Speaker 1: first appearance with us today. It's been great to have
• 46:56
  Speaker 1: you on SAM. Thanks for coming back kicking off the
• 46:58
  Speaker 1: new year again. Happy new Year everybody, and I hope
• 47:02
  Speaker 1: we will have a fantastic twenty twenty six.
• 47:05
  Speaker 3: Can you believe it?
• 47:06
  Speaker 5: Thanks guys, I thank.
• 47:07
  Speaker 3: You for hever so much. Thank you to make sure
• 47:09
  Speaker 3: that you never miss an episode.
• 47:11
  Speaker 2: Subscribe to the show in Apple Podcasts, Spotify, or your
• 47:14
  Speaker 2: favorite podcast player, and if you're listening on Apple podcasts,
• 47:18
  Speaker 2: make sure to leave a rating of the show. Just
• 47:20
  Speaker 2: tap the number of stars you think the podcast deserves.
• 47:23
  Speaker 2: If you'd like to learn more about how next Think
• 47:25
  Speaker 2: can help me improve your digital employee experience, head over
• 47:28
  Speaker 2: to next think dot com.
• 47:30
  Speaker 3: Thank you so much for listening.
• 47:31
  Speaker 2: Until next time,