Latest / The Adults In The AI Room / Massood Zarrabian on Innovation in Enterprise Search with AI - Episode #3

Transcript

Massood Zarrabian on Innovation in Enterprise Search with AI - Episode #3 · The Adults In The AI Room

• 0:36
  SPEAKER_01: Well, welcome to our podcast today.
• 0:38
  SPEAKER_01: Today, our guest is someone who has spent his career helping to improve how organizations and individuals retrieve information while finding joy in positively impacting both the bottom lines of organizations and human lives.
• 0:52
  SPEAKER_01: He's currently the CEO of BA Insight, an innovator in the AI-driven enterprise search market.
• 0:58
  SPEAKER_01: Please welcome search and information retrieval pioneer Massoud Zarabian.
• 1:02
  SPEAKER_04: Hi, everybody.
• 1:03
  SPEAKER_04: Thank you for very much for your time and thank you for the audience who's going to be listening to us.
• 1:07
  SPEAKER_01: Chris, why don't you uh uh things uh kick things off here today?
• 1:11
  SPEAKER_03: I would love to get a you know get a take because you have a very interesting uh background, Masood, where you know, going to uh MIT, nice work.
• 1:21
  SPEAKER_03: And but you have a degree in civil engineering, correct?
• 1:24
  SPEAKER_03: Yep.
• 1:26
  SPEAKER_03: So now that you're um, you know, I I would love to get your thought process too in terms of what was your original motivation for you know graduating coming out of MIT?
• 1:38
  SPEAKER_03: What did you what did you believe you were going to change in the world?
• 1:41
  SPEAKER_04: Well, my original motivation to go to MIT, as you know, it was I came from Iran, was to become an expert in stuff, go back to Iran and change the world and from in the country that was at the time, a third world country probably still is.
• 1:58
  SPEAKER_04: So that was the that was the motivation of coming here, that was the motivation of studying here, that was the motivation of becoming a civil engineer.
• 2:04
  SPEAKER_04: We had a lot of things to do to build a country, right?
• 2:08
  SPEAKER_04: So and uh and when when I came here, I stayed here, I went back a couple of times, but somehow this country uh bit me.
• 2:19
  SPEAKER_04: I got to like it, I like the freedom, I like the I like all the things that a human human being has here that didn't have there.
• 2:29
  SPEAKER_04: So once I graduated, I changed my mind and decided to stay.
• 2:33
  SPEAKER_04: Unfortunately, at the time, the civil engineering market was really bad.
• 2:38
  SPEAKER_04: So it was really hard to find a job, right?
• 2:42
  SPEAKER_04: And and my journey changed, right?
• 2:44
  SPEAKER_04: My journey changed to okay, I'm gonna go back to business school.
• 2:47
  SPEAKER_04: Through that, I found a co-op job, I had a company in computer vision, I got new software, and the rest is history.
• 2:54
  SPEAKER_03: You can, you know, take the flavor of the engineering and change them, but you can't take the engineer out of the person, right?
• 3:00
  SPEAKER_03: It's basically how it goes.
• 3:01
  SPEAKER_04: So that's you know that that you got.
• 3:03
  SPEAKER_04: Yes.
• 3:04
  SPEAKER_03: Yeah, that's awesome.
• 3:06
  SPEAKER_03: So um, yeah, I I I actually I really love the the background because um most of what what we find, especially with those who will actually make a big difference in the changes from something that was uh that they grew up with that was I don't know whether you say it was life changing, traumatic, whatever that thing was, right?
• 3:25
  SPEAKER_03: To actually get you to a point where where the theme of of what you're doing is changed, but not in terms of uh still wanting to take on and really provide enablements and positive change of the world, right?
• 3:40
  SPEAKER_03: And so when I think of it in terms of where you are now, you know, and and just trying to get either more and better insights out of speech and natural language processing and some of those aspects that you know BA is into, um, you know, what what what would you say is is like the the key motivators now, right?
• 4:00
  SPEAKER_03: As you've gone through, you know, business and you've gone through some of these areas, you know, how how has that outlook changed and and really what's what's the uh the goal now for Massood?
• 4:11
  SPEAKER_04: I mean, I I don't think it's changed through my career.
• 4:13
  SPEAKER_04: Uh, you know, I like I like uh a bunch of things about work.
• 4:18
  SPEAKER_04: I like uh helping people grow.
• 4:20
  SPEAKER_04: I'm lucky enough to have gone company to company and have people who have decided to join me in the second company.
• 4:27
  SPEAKER_04: I'm proud of having that kind of loyalty.
• 4:30
  SPEAKER_04: I'm proud of people that I worked with that grew and someplace else became CEOs and CTOs, even though we didn't never work together again.
• 4:38
  SPEAKER_04: I'm proud of that.
• 4:39
  SPEAKER_04: I'm proud of making companies grow.
• 4:42
  SPEAKER_04: I always uh I always uh advocate that it actually is not people first, it's a company first.
• 4:49
  SPEAKER_04: Because if a company doesn't do well, people cannot.
• 4:53
  SPEAKER_04: We will not survive if a company is not profitable.
• 4:55
  SPEAKER_04: So so we we have this vision in multiple companies I work with, which is you know, if if we don't take care of this thing, which is our community, right?
• 5:05
  SPEAKER_04: Make it profitable, make it grow, make it have impact on our customers, keep our customers, we none of us will have jobs.
• 5:13
  SPEAKER_04: And I think that motivation that is balanced with a love technology.
• 5:17
  SPEAKER_04: I think a lot of technology can do a lot of things to improve lives, to improve things, to simplify things.
• 5:25
  SPEAKER_04: So even though I'm not a great software engineer, I've learned over a period of years to hang around great software engineers and try to help people to think about what the end result is and how you get to it is is not is critical, but that's not the vision.
• 5:44
  SPEAKER_04: The vision is the end result.
• 5:45
  SPEAKER_04: Like and I and yeah, you guys have heard me, right?
• 5:48
  SPEAKER_04: That that Google bar changed everything, right?
• 5:50
  SPEAKER_04: That's the end result.
• 5:52
  SPEAKER_04: The technology behind it is great, but the result is even better.
• 5:56
  SPEAKER_01: Well, we're gonna jump into that.
• 5:57
  SPEAKER_01: Uh, but I I have one other question about your background.
• 5:59
  SPEAKER_01: You have an interest in theater and film and humanities, and how does that manifest?
• 6:04
  SPEAKER_01: How you know we're number one, you you know, is it uh were you an actor, director?
• 6:08
  SPEAKER_01: What were you what was your role?
• 6:10
  SPEAKER_01: And then how does that also influence what you're doing in terms of the space?
• 6:13
  SPEAKER_04: So so let me let me tell you how I got into humanities.
• 6:16
  SPEAKER_04: When you go to MIT, you have to take a bunch of humanities courses, and I actually got to like it.
• 6:20
  SPEAKER_04: So I took more, I took more of it than normally people would.
• 6:24
  SPEAKER_04: So that's how how my interest actually started.
• 6:27
  SPEAKER_04: Okay.
• 6:28
  SPEAKER_04: And then years, I I've always liked movies, I especially like old movies.
• 6:34
  SPEAKER_04: Uh I, you know, that was a a relaxing thing to do.
• 6:39
  SPEAKER_04: It was, you know, everybody has something that they're interested in.
• 6:42
  SPEAKER_04: Uh years later, I realized you have to get your top picks, by the way.
• 6:46
  SPEAKER_01: You'll have to get your top picks at some point.
• 6:48
  SPEAKER_04: All right.
• 6:48
  SPEAKER_04: My topics are pretty simple.
• 6:50
  SPEAKER_04: It's James Bond movies, I love him.
• 6:52
  SPEAKER_04: And then my my real topic is Casablanca, which is an old old Humphrey Bogart movie, and and and and part of the topic in that is because my wife's name is Elsa, which is Ingrid Breckman's name in that movie.
• 7:05
  SPEAKER_04: But uh, but I that my my interest in movies in time grew because I started seeing a correlation between how movies are built and the good ones, and how companies are built.
• 7:17
  SPEAKER_04: Movies are directors, producers, right?
• 7:19
  SPEAKER_04: Movies have different actors, different roles.
• 7:22
  SPEAKER_04: Some spend less time visibly, some spend more time visibly, right?
• 7:27
  SPEAKER_04: Companies are like that.
• 7:28
  SPEAKER_04: Companies have directors and producers, they have CMOs and CTOs.
• 7:33
  SPEAKER_04: And a a great movie or a great play is a combination of how those people collaborate, produce, deliver.
• 7:40
  SPEAKER_04: And I think a company is the same thing, is it's how that group together does it.
• 7:46
  SPEAKER_01: Yeah, and what's interesting about a movie is you get all these people together who've never many times not work together, and then they have to produce a zero-defect product at the end of the shoot, right?
• 7:56
  SPEAKER_01: The end of the editing.
• 7:57
  SPEAKER_01: So yeah, it's pretty impressive.
• 7:59
  SPEAKER_01: Exactly.
• 8:00
  SPEAKER_04: And then the other thing about a movie is then there are critics that have never made a movie that tell you this tough about your movie, right?
• 8:06
  SPEAKER_04: That's right.
• 8:06
  SPEAKER_03: So let's just hope you don't design in the uh the antagonist, you know, you're okay.
• 8:15
  SPEAKER_01: You mentioned Google, and and that's really what we want to talk about today is search and Google search.
• 8:19
  SPEAKER_01: And you know, when people ask why can't it be like Google, what what is your refrain from that?
• 8:24
  SPEAKER_01: I mean, you know, my perspective is if people spend as much money optimizing content as you do optimizing content for Google for the enterprise, then you would have searched like Google.
• 8:33
  SPEAKER_01: But what do you tell us your thinking around uh around search and and that Google experience?
• 8:40
  SPEAKER_04: Well, I think there are bunch of issues.
• 8:43
  SPEAKER_04: I think one part of one issue is that we, at least in my life of growing up in the enterprise around enterprise search, we started with Fast and SharePoint 2010 and SharePoint 2013.
• 8:57
  SPEAKER_04: And and trying to make those things to be Google is really hard.
• 9:00
  SPEAKER_04: It's really, really hard, right?
• 9:02
  SPEAKER_04: And and I think that then caused people to think that enterprise search to become like Google is really hard.
• 9:09
  SPEAKER_04: And and I I I agree that it was really hard at the beginning of the last decade, but but as things changed, I and things became easier and easier, uh, people still think it's hard.
• 9:22
  SPEAKER_04: And and you know, I I guess anything you do in business has some level of degree of difficulty, but now things are not as difficult as they used to be, right?
• 9:30
  SPEAKER_04: Doing things around content services have become much easier, doing things with natural language have become much less expensive and much easier.
• 9:38
  SPEAKER_04: Doing things with with you know, taking advantage of all the cognitive services that everybody has that are getting more and more commodized, right?
• 9:46
  SPEAKER_04: Those are the things that Google used to have and was not available to the enterprise that is not available to the enterprise.
• 9:52
  SPEAKER_02: Right.
• 9:53
  SPEAKER_04: So all of a sudden, technology can do things that used to take really knowledgeable people.
• 10:00
  SPEAKER_04: And I'm not saying that it doesn't take those now, now the expertise has to be even higher, but the number of people evolved is much smaller, right?
• 10:08
  SPEAKER_04: Right, right.
• 10:08
  SPEAKER_04: The strategy becomes easier because you don't have all these impediments.
• 10:14
  SPEAKER_04: And and I think that I think that the that the the Google search for experience is getting to the place where it is very, very possible to do it at much lower cost and a much lower time than it used to be.
• 10:27
  SPEAKER_01: I I remember a conversation with Jeff Freed, who was one of your prior uh CTOs, um great guy, known him for many years.
• 10:33
  SPEAKER_01: And one of the things he said to me is, look, you could do a lot of this stuff five years ago with enterprise search, but it would cost you 20 million dollars.
• 10:41
  SPEAKER_01: Right.
• 10:42
  SPEAKER_01: And what's changed over those years is the things that used to have to be handcrafted, hand-coded, hand integrated, that you needed to build all of these, you know, UI functions and all of these, you know, uh classification structures and all of the you know uh semantic structures that you had to build by hand are now in many cases either as uh on a platform service or built into the infrastructure, built into the tools that are available today.
• 11:08
  SPEAKER_01: Do you want to speak more to that?
• 11:10
  SPEAKER_04: Yeah, exactly.
• 11:11
  SPEAKER_04: So so you know, like let's just go back.
• 11:14
  SPEAKER_04: Like in the old days, if you had a bunch of systems and you had to go on and do content services, specialist knowledge about people define content services for folks who may not be familiar with that content.
• 11:25
  SPEAKER_04: To me, content services is is set up amount of content that you have where you don't have knowledge about what that content is because it's all this legacy.
• 11:35
  SPEAKER_04: You let the users tag it and they tagged it the wrong way, right?
• 11:39
  SPEAKER_04: You just don't know all this the wealth of information available to you.
• 11:44
  SPEAKER_04: So people call that dark, dark data sometimes.
• 11:47
  SPEAKER_04: Um that's that's the that's a great way of thinking about it, right?
• 11:50
  SPEAKER_01: Yeah, you don't know what's there.
• 11:52
  SPEAKER_04: So so what do you want to do?
• 11:53
  SPEAKER_04: Hire 100 people, have them go and open each file one by one, one one, and tell you what's in it.
• 11:58
  SPEAKER_04: I mean, that's that's the world we used to live in, right?
• 12:01
  SPEAKER_04: Which is not the world of web even 10 years ago, because 10 years ago the web was pages, pages were tagged, because if the thing wasn't tagged, you wouldn't find it, everybody was scared of not getting fine.
• 12:10
  SPEAKER_04: Found them the whole SEO thing got created, right?
• 12:14
  SPEAKER_04: Because of that.
• 12:15
  SPEAKER_04: So that was difficult.
• 12:18
  SPEAKER_04: And if you cannot find the right content, enterprise is gonna fail, right?
• 12:23
  SPEAKER_04: But if you cannot tag the thing, and if you don't understand what it is, and then you're gonna fail.
• 12:29
  SPEAKER_04: What has changed is a lot of those things can be automated.
• 12:32
  SPEAKER_04: This doesn't mean that you don't need some expertise to review, approve.
• 12:37
  SPEAKER_04: It just means a lot of the work that was mechanical, tedious, go system by system, repetitive, yeah.
• 12:43
  SPEAKER_04: Yeah, and all other stuff can now be done automatically, and you end up, you know, you get quickly to 80% of the way there, and then you improve it and improve it.
• 12:52
  SPEAKER_04: So it's much cheaper, it takes a lot less time, and then once you have done it, it's continuous, right?
• 12:58
  SPEAKER_04: You don't have to every year come back and say, Okay, let me start over again, right?
• 13:01
  SPEAKER_04: You probably would do some rework every year to because the world is changing, but what you did stays in place and moves for built upon, right?
• 13:12
  SPEAKER_04: And there are a bunch of learning models that are getting created that lets the system learn from you know and user behavior and user clicks.
• 13:20
  SPEAKER_04: So it is becoming much more viable to do a Google kind of experience in an enterprise at much, much lower cost.
• 13:28
  SPEAKER_01: Yeah, and the question is.
• 13:31
  SPEAKER_03: I think the one thing to highlight, Seth, is that what we're really talking about is unstructured data, right?
• 13:36
  SPEAKER_03: And the difference between structured data and unstructured data is what you're getting at is it's uh it's you know files and formats that are locked down.
• 13:46
  SPEAKER_03: And you know, how do you expose that information that that somebody may you know have on a desktop and is pushing to, let's say, a central you know repository or something, but you know exactly what we're saying, we don't we didn't have for a long time, and I think this is the essence, it's one thing to have a you know, a uh data structure that's there and well understood and is is normalized and whatever else, but then you get into now the proliferation of graph systems and the proliferation of you know of other miners, and then take your data repository of non of unstructured data, throw that in there and say, okay, make it searchable.
• 14:22
  SPEAKER_03: Yeah, right.
• 14:23
  SPEAKER_03: But where's all the metadata about it?
• 14:25
  SPEAKER_03: Right.
• 14:25
  SPEAKER_03: Well, it's there, but how do we expose it in a wrong way?
• 14:28
  SPEAKER_03: So we want to make sure that we're talking about this, you know, the apple, it's apples and oranges in terms of structure or subject.
• 14:34
  SPEAKER_04: Absolutely.
• 14:34
  SPEAKER_04: Yeah, go ahead, Masoo.
• 14:36
  SPEAKER_04: Yeah, Chris, that's a very that's a that's an excellent observation.
• 14:38
  SPEAKER_04: That's exactly right, because structure is much easier to address, right?
• 14:41
  SPEAKER_04: And and a lot of unstructured becomes more difficult.
• 14:44
  SPEAKER_04: And as time has gone by, we have gone from just having Word and PDFs, and we have you know, Jira this, and we have GitHub that, and we have Confluence the other thing.
• 14:54
  SPEAKER_01: So so and and and and somebody told me GitHub that confluence the other thing, Slack this, Slack whatever, right?
• 15:01
  SPEAKER_04: And and somebody told me this morning that they have tried to force to to discipline people to go and put the content in the right systems, so they have rules about where to put the content, and they actually have people who are monitoring it, and they say we are nuts, that's not making it.
• 15:15
  SPEAKER_04: We just we are not able to get the users right to put the content in the right place.
• 15:20
  SPEAKER_04: So instead of doing that, we're gonna go the other way around.
• 15:22
  SPEAKER_04: Let them put wherever they want to.
• 15:24
  SPEAKER_04: Is there a way of finding all of it?
• 15:26
  SPEAKER_03: Yeah, the most expensive because it's knowledge workers that are generating this, it's probably the most expensive data that any organization has, yet the least available to the organization.
• 15:39
  SPEAKER_03: Yeah, right.
• 15:40
  SPEAKER_03: So that's you know, I mean, and and that to me defines the crux of what we're talking about here is is really trying to get at the most most valuable and expensive information that we have, and yet we can't do it very easily.
• 15:54
  SPEAKER_01: I I have a quick story about that, and then I'm gonna turn it over to Massood for his comments.
• 15:58
  SPEAKER_01: But I remember working at a large uh global manufacturer in the aerospace industry, they spend$250 million on their structured data, on their ERP implementation.
• 16:08
  SPEAKER_01: $250 million.
• 16:09
  SPEAKER_01: Then when it came to looking at the knowledge and the content, they they the there's a project of about$1.5 million, and they're like, whoa, that's way too much money for us, right?
• 16:21
  SPEAKER_01: You spend$250 million here in this, and you want to take less than 1%, like a half a percent, and spend it on the knowledge and the and the content and the unstructured.
• 16:31
  SPEAKER_01: And organizations have had this history and legacy of while data is important and transactions and accounting, and you would not go to the accounting organization and say and have people say, oh my goodness, all those numbers, you have to put them in the right place, they have to add up.
• 16:45
  SPEAKER_01: Oh, it's just too much work, right?
• 16:47
  SPEAKER_01: You wouldn't find that because people know how valuable is.
• 16:50
  SPEAKER_01: The problem with unstructured information is there's this lack of true understanding of that that core value that is within knowledge and content.
• 17:01
  SPEAKER_01: And as you say, Chris, it's the knowledge workers.
• 17:03
  SPEAKER_01: Masoo, do you want to speak to that in terms of are organizations starting to come around to this?
• 17:08
  SPEAKER_01: You know, do you still see this kind of big dichotomy and and and divergence of thinking?
• 17:14
  SPEAKER_04: Yes, so so let me let me just come back to this because I have a view on this about time.
• 17:19
  SPEAKER_04: If you go, if if if we were sitting here in 1990 and talked about the same exact thing, people would not spend 250 million bucks on an URP system.
• 17:28
  SPEAKER_04: You will be trying to convince them that a data warehouse is a good idea.
• 17:32
  SPEAKER_04: Because but that started in the 70s and it took 20, 30 years to get there, right?
• 17:36
  SPEAKER_04: Yeah.
• 17:36
  SPEAKER_04: The problem with enterprise search, the problem with this whole idea of self-service, because enterprise search is a part of something much bigger, which is self-service.
• 17:45
  SPEAKER_04: Whether you do it internally, customers, websites doesn't matter.
• 17:49
  SPEAKER_04: The problem is it started, at least I think it started from my experience.
• 17:54
  SPEAKER_04: I joined a company named ServiceSoft in 1999, and we had one of the first uh self-service knowledge managers, knowledge management system for customer support.
• 18:02
  SPEAKER_04: We use a search engine at a time named Verity.
• 18:05
  SPEAKER_04: Okay.
• 18:06
  SPEAKER_04: And we were trying to sell people this idea we don't need 100 people on the call, right?
• 18:11
  SPEAKER_04: You can actually do self-service.
• 18:13
  SPEAKER_04: Uh 10, 15 years later, that has kind of become acceptable that customer portals self-service.
• 18:20
  SPEAKER_04: I think part of the problem we have here is the beginning of our market that started in in 2002-2010 with portals.
• 18:28
  SPEAKER_04: And portals were good in first generation solutions that kind of failed.
• 18:32
  SPEAKER_04: But from there, we went to SharePoint and FAST.
• 18:36
  SPEAKER_04: Right?
• 18:36
  SPEAKER_04: In other places, you went from there, you went to a system that was next level.
• 18:42
  SPEAKER_04: So part of our issue is patience to get people to understand what we do.
• 18:47
  SPEAKER_04: Now, having said that, the reason I tell that story is because for the first five or six or seven years I was here, it was a struggle to get people to understand how valuable this could be.
• 18:57
  SPEAKER_04: I'm not gonna say it's not a struggle, it is much less of a struggle.
• 19:00
  SPEAKER_04: We now have people who come in that already have figured a bunch of things out and now are trying to find somebody who can play both the role of an advisor and a technology vendor with their partners or by themselves so that they can take the next level.
• 19:16
  SPEAKER_04: And I'm very optimistic.
• 19:18
  SPEAKER_04: And I think and I think uh the pandemic and COVID actually exasperated the problem.
• 19:25
  SPEAKER_04: I think a lot of people realize that you know, enterprise search or or this Google like search bar can do some of the stuff that coolers did, right?
• 19:34
  SPEAKER_04: Right.
• 19:34
  SPEAKER_04: You know, I can now ask without worrying about are you on the system?
• 19:38
  SPEAKER_04: Are you available?
• 19:38
  SPEAKER_04: Right, and I and I think things are starting to mature.
• 19:41
  SPEAKER_04: Um I'm very optimistic that in the next five years we're gonna go away from trying to spend so much time convincing people that this is the right solution to one that is much more like what you just talked about, which is acceptable.
• 19:55
  SPEAKER_04: And hopefully we will not make it so that it costs 250 million dollars to install it.
• 20:00
  SPEAKER_01: Let me let me one little one little uh before you go, Chris.
• 20:04
  SPEAKER_01: I just want to throw one other little anecdote out.
• 20:06
  SPEAKER_01: And I remember in the early 90s meeting with a company, and the problem they said was they needed to get more information online back then.
• 20:14
  SPEAKER_01: They made more information online, that was their big problem.
• 20:18
  SPEAKER_01: You're not gonna get anybody saying that now, right?
• 20:20
  SPEAKER_01: Exactly.
• 20:20
  SPEAKER_01: So the nature of the beast has evolved.
• 20:22
  SPEAKER_01: Sorry, Chris, go ahead.
• 20:23
  SPEAKER_03: I I think I think one I'd love to get your take on this on the evolution of the persona of who's actually, you know, getting this information.
• 20:33
  SPEAKER_03: Because before, like you said, I mean, the the complexity of uh putting up, let's say, an elastic search environment or something like that is a developer that has to do that.
• 20:43
  SPEAKER_03: And a lot of times, uh, you know, the the core of good technology is just trying to make it easy and accessible for the business to get to it.
• 20:53
  SPEAKER_03: And uh, you know, I can't have an army of developers actually, you know, go through and create some custom thing in order to get this information.
• 21:01
  SPEAKER_03: I've got to meet you know, meaning that the business doesn't care how the sausage is made.
• 21:05
  SPEAKER_03: They just need to be able to get access.
• 21:07
  SPEAKER_03: Yep.
• 21:08
  SPEAKER_03: So, you know, I think that's another another point that I'd love to get your take on, Masood, is uh is the evolution of not only the technology, but the evolution of the user, the personas, these, you know, the the folks that are actually getting this done now have evolved like on the Simeon scale, right?
• 21:25
  SPEAKER_03: From you know, from from you know, core little APIs and objects into now, you know, full-on business solutions.
• 21:33
  SPEAKER_04: So yeah, you know, when I when I joined V Insight in 2013, what amazed me is how many people with coding experience, even we needed to install configure and make our software work.
• 21:48
  SPEAKER_04: I I I was looking at this thing now.
• 21:50
  SPEAKER_04: Uh somebody will install our software via somebody who could do JavaScript and this and that and the other thing, and they will spend weeks with that customer, and then we end up with uh what people Said was a software solution, but it was 20, 30, 40% customized.
• 22:05
  SPEAKER_04: Right.
• 22:06
  SPEAKER_04: Which was, you know, and so wait a minute, you know, if if if if you decide to build a building and every wire, every every pipe had to be customized, you never finish one.
• 22:17
  SPEAKER_02: Right.
• 22:17
  SPEAKER_04: So even after the idea of user interfaces that let people cust configure things, we have the idea of look, let's not try to solve all the world's problem, let's solve the 80-90 percent that really matters, and it's okay if you cannot do the rest of it, we'll figure it out.
• 22:34
  SPEAKER_04: Yeah, and and we started building components that we we call modules that can play together.
• 22:40
  SPEAKER_04: And we also said, let's not be some big platform that everybody has to have, right?
• 22:44
  SPEAKER_04: So instead of being SAP, ARP, let's build modules that combine into SAP or ARP, and someone needs it to do that.
• 22:50
  SPEAKER_04: So our solution now, oh, it took us a while to get there, but our solution now supports different search engines, different uh cognitive services from different vendors.
• 23:01
  SPEAKER_04: We have 90 connectors, and it's kind of like a Lego game, it's like a plug and play.
• 23:06
  SPEAKER_04: And what we have done is we have created UIs in the back end that a business analyst can do things with, they as opposed to some IT person that knows Python.
• 23:18
  SPEAKER_04: Right, because an IT person who knows Python is rare and they have a lot more important things to do.
• 23:24
  SPEAKER_04: Right.
• 23:25
  SPEAKER_04: So as you do that and you learn from it, you end up with a product, with a more of a product view, as opposed to oh, the only way to build enterprise search is get a bunch of engineers to build something around Elastic, and then they leave, and the next person doesn't know what to do, and something falls apart, and and the thing gets a bad reputation.
• 23:45
  SPEAKER_04: So the strategy should be for anybody who wants to do any kind of enterprise-driven search, in my opinion.
• 23:52
  SPEAKER_04: Think about it as software.
• 23:54
  SPEAKER_04: Like, don't think it think about as software with advisors that are smart and help you.
• 23:59
  SPEAKER_04: Don't think about it as engineers that code and customize to build a custom solution for you because that's going to eventually backfront.
• 24:06
  SPEAKER_01: Right, right.
• 24:07
  SPEAKER_01: That becomes too costly and unsustainable.
• 24:10
  SPEAKER_01: And what you're saying is all that plumbing in the background no longer has to be coded.
• 24:16
  SPEAKER_01: It is simply connected with web services and APIs and really these built re pre-built modules that take care of a lot of that stuff in the background that used to be the heavy lift, right?
• 24:28
  SPEAKER_01: That's no longer the heavy lift.
• 24:30
  SPEAKER_01: And now we can do things like do classification, entity extraction, text processing, uh, intent identification, you know, uh uh uh utterance resolution.
• 24:41
  SPEAKER_01: Now, what's interesting is when you start thinking of all those pieces, uh, the last mile, and you talked about cognitive services.
• 24:48
  SPEAKER_01: And what I always believe is that whenever you're doing anything that is going to impact usability or the ability to interact with information, what you're doing is you're reducing the cognitive load on the human, right?
• 24:58
  SPEAKER_01: You're making they have to think less, they don't have to, you know, go spelunking through all this content.
• 25:03
  SPEAKER_01: You're saying, here's the answer, here's what you need, right?
• 25:06
  SPEAKER_01: So you can focus on higher value stuff.
• 25:08
  SPEAKER_01: But when you start thinking about, and that's the the misnomer of cognitive computing, it's not that the system thinks, it's that it's helping reduce the cognitive load.
• 25:16
  SPEAKER_01: Now, this is the future, right?
• 25:18
  SPEAKER_01: This is the evolution.
• 25:19
  SPEAKER_01: We're starting to see more cognitive assistance, you know, conversational interactions, but those are only as good as those mechanisms to connect those to those data sources and then the ability to kind of process that content so it can be served up in those services, in those bots, in those virtual assistants.
• 25:37
  SPEAKER_01: I always look at those as information retrieval problems.
• 25:40
  SPEAKER_01: And really, search is it's an information retrieval problem, just like search is an information retrieval problem, right?
• 25:46
  SPEAKER_01: Chatbots are a channel.
• 25:48
  SPEAKER_01: So when you think about it, these next several years, when these things get better and better, it really is the last mile is really going to be about the knowledge and the content and getting it in shape for these tools to expose them to these different types of virtual assistants, right?
• 26:04
  SPEAKER_01: So, what's that last mile in your mind, in your opinion?
• 26:07
  SPEAKER_01: So you have this infrastructure, you have these search technologies and tools that can process the content, that can make sense of the user's intent.
• 26:16
  SPEAKER_01: You know, what is the piece that really needs to be handled uh on an organization by organization basis?
• 26:22
  SPEAKER_01: So, in other words, you know, the plumbing is going to be done, but it's going to be different for different organizations.
• 26:28
  SPEAKER_01: Where's that value add?
• 26:30
  SPEAKER_01: Where's that work that needs still needs to happen uh for these enterprises?
• 26:34
  SPEAKER_04: So that there are a bunch of things.
• 26:36
  SPEAKER_04: I mean, clearly the use cases it forever are going to be a critical time.
• 26:40
  SPEAKER_04: Because no matter what you do, the world is changing, and the use case of today is going to be completely different five years from now.
• 26:46
  SPEAKER_04: Yep.
• 26:46
  SPEAKER_04: So one part of the last mile is to me figuring out that the solution you have can evolve as the world evolves.
• 26:55
  SPEAKER_04: Because if it doesn't, and every time something happens, you have to lift and shift, you're gonna you're gonna be in big, big pain.
• 27:02
  SPEAKER_04: I I I look back at the number of calls we got when Google decided to get rid of Google Search Appliance.
• 27:08
  SPEAKER_04: I I cannot imagine, I'm I'm sure it was hard for Google to make that decision, right?
• 27:15
  SPEAKER_04: And painful, but I cannot imagine the pain that customers went through getting moving away from that to something else.
• 27:21
  SPEAKER_04: And I cannot imagine how many people afterwards complain to them, but it doesn't do everything Google did because that's what happens when you go from system one to system B, right?
• 27:29
  SPEAKER_04: So so I think the use case view is a critical one.
• 27:33
  SPEAKER_03: I think oh keep going, keep going.
• 27:36
  SPEAKER_03: I've got I've got a follow-on, but keep going.
• 27:38
  SPEAKER_04: Yeah, I think that figuring out what are you gonna do with your content, no matter what happens, in terms of understanding your content, tagging your content, making it findable, it's absolutely a huge access, right?
• 27:50
  SPEAKER_04: And and my advice is figure out on one side, have experts that can advise you, and on the other side, understand the technology is gonna move faster than yours and and keep keep an eye on it and take advantage of it.
• 28:02
  SPEAKER_03: I think that's I think I wanted to I wanted to tug on that thread a little bit and just get your opinion for the organizations who are looking about looking at it from the perspective of I've got thousands and thousands of you know of pieces of information.
• 28:16
  SPEAKER_03: They're in thousands of different systems, and you know, getting an advisor is is super important.
• 28:22
  SPEAKER_03: But what do you think the barrier to entry is for these folks who are just like, hey, where do I start?
• 28:28
  SPEAKER_03: Right?
• 28:28
  SPEAKER_03: What can what do I need to do first?
• 28:30
  SPEAKER_04: I think what they do is they think about you have to go to each system and figure it out system by system.
• 28:38
  SPEAKER_04: Okay, think think about it.
• 28:39
  SPEAKER_04: And I it's a metaphor I use the other way.
• 28:41
  SPEAKER_04: The customer think about uh forget about internet availability.
• 28:45
  SPEAKER_04: Let's assume there were a bunch of stores and you want to buy something, but you don't know what's in each store.
• 28:49
  SPEAKER_04: What are you gonna do?
• 28:49
  SPEAKER_04: Go to a shopping mall and go to every store to figure out which one sells a shirt and which one sells, you know, uh diet coke and which one sells this.
• 28:57
  SPEAKER_04: That's the problem.
• 28:58
  SPEAKER_04: So what they do is they the problem becomes bigger.
• 29:01
  SPEAKER_04: But coming back to the information retrieval, if you then attempt to change that into let's just connect those system to a central index and use the index to figure it out, then that expert has one place to go with information from all the system and just and they can figure it out.
• 29:15
  SPEAKER_04: Yeah, you don't you don't now you don't need any more a documentum expert and a this expert and a that expert, you just centralize it.
• 29:22
  SPEAKER_04: It's kind of like what we did we did years ago with data warehouses, or now we are doing with the next generation and stuff like that.
• 29:28
  SPEAKER_04: We don't go to the source, we put it one place and we go figure out what to do with it in that one place.
• 29:34
  SPEAKER_04: So, so search as a technology of giving you what you need from multiple systems in a homogeneous way that then lets the expert look at and help you is enormously powerful.
• 29:47
  SPEAKER_04: And as a part of that, right, you could even do there are there are people who do things like, hey, now as we are doing that, there are 50, 60 things we found that repeat, let's use this to take advantage of NLP to train systems so I don't have to do this again.
• 30:04
  SPEAKER_04: So so I think that's that's one axis you can use, and people don't know that's possible.
• 30:11
  SPEAKER_04: So what happens is the barrier becomes lack of understanding.
• 30:14
  SPEAKER_04: And then if I'm trying to sell you half a million dollars worth of services, would I come and tell you you can do this with an expert for 50 grand?
• 30:21
  SPEAKER_04: No, right?
• 30:23
  SPEAKER_04: I'm gonna do exactly the opposite, right?
• 30:25
  SPEAKER_03: Yeah, so basically, so start small, start with the start with a specific.
• 30:30
  SPEAKER_03: I like when you said start with the use case and start with the end in mind, and then go just set an initial hypothesis on what type of information or what you want to do with the information, let alone what's in there.
• 30:42
  SPEAKER_03: Look at a single system and just start with that one and then bridge out from there in terms of you can start you can start with two or three systems, don't start with 20.
• 30:52
  SPEAKER_04: You can start with two or three systems, but use the index almost as an intelligent repository to help you.
• 30:59
  SPEAKER_01: Right.
• 30:59
  SPEAKER_01: The index is centralized, the content and data is still decentralized in these different systems, sorry.
• 31:06
  SPEAKER_01: But the index becomes that that piece that uh knows where the bodies are buried, right?
• 31:11
  SPEAKER_04: It goes.
• 31:12
  SPEAKER_04: The experts can help you a lot, right?
• 31:13
  SPEAKER_04: Because if you if you do that, then all of a sudden the the experts have a consolidation of a bunch of systems.
• 31:20
  SPEAKER_04: They don't have to model it anymore, right?
• 31:21
  SPEAKER_04: They don't have to go system up and figure things out, they now have something that that really uses their expertise, which is okay, not that I know what you got, I can tell you what you need to do.
• 31:31
  SPEAKER_04: But but in the old days, you have to go to each system to figure out what I got before you can, right?
• 31:38
  SPEAKER_03: Let's pivot a bit, uh, because I want to get into you know your thoughts around you know the technologies that BA is using and stuff to help organizations do this and stuff, centered around like building, maintaining, and training and tuning indexes, right?
• 31:51
  SPEAKER_03: Because it's it's it's it can be super um, it'll make or break anything that you're looking for, but it can also um make or break an entire environment if you don't have, you know, and it's not just one index.
• 32:06
  SPEAKER_03: I think that's what other people, you know, what some others think too is I have just this master index, and that's not necessarily the case.
• 32:13
  SPEAKER_03: It's it can be multiples, depending on, again, what the outcome is, whether it's departmental, whether it's you know, a specific theme, whatever that is, that we can have multiple indexes.
• 32:22
  SPEAKER_03: So let's talk a little bit about, you know, you know, how you guys are using these indexes to help organizations and what that looks like in terms of of uh uh maybe just going through the process of of how fast or time to market it is to actually build an index now and get it used, get it, you know, up and running, you know, for usage and retrieval.
• 32:44
  SPEAKER_04: Yeah, so so let me let me let me use an example, Chris, on this.
• 32:49
  SPEAKER_04: You know, if web was a single index, can you imagine the size of that index?
• 32:55
  SPEAKER_04: And can you even imagine that it's possible to keep up with the change?
• 32:59
  SPEAKER_04: So the whole idea of the web is built on this thing, which is about merging after the the after somebody searches.
• 33:07
  SPEAKER_04: Now, if you go back to the to enterprise search, we tried to do that in I don't know, 2003, four, five, and it stunk.
• 33:18
  SPEAKER_04: It was really bad.
• 33:19
  SPEAKER_04: Okay.
• 33:21
  SPEAKER_04: And and we have lived with that legacy.
• 33:24
  SPEAKER_04: The reason it was bad though is on the web, everybody who's doing those systems is doing it in a way where you can merge.
• 33:32
  SPEAKER_04: If I go back and try to do Lotus Notes and do the same thing, it's not gonna work because at the time Lotus Notes was created, this idea didn't even exist.
• 33:40
  SPEAKER_04: Fast forward is almost 2022.
• 33:44
  SPEAKER_04: A lot of things have changed.
• 33:45
  SPEAKER_04: Lotus notes still exists, but then there are other things that can do exactly what the web did.
• 33:50
  SPEAKER_04: So our view is that long term, the singular index idea is not gonna work because the index is gonna become so big and the change rate of change is gonna be so big that when I search, I'm not gonna get fresh results.
• 34:02
  SPEAKER_04: So you have to distinguish between new modern systems that let you do what the web does, which is multiple search engines, multiple indexes, right?
• 34:11
  SPEAKER_04: Amazon uses a different one than Xpedia, but you get the results, and live with a legacy system that cannot do the same thing.
• 34:20
  SPEAKER_04: So the singular index idea will work for things that are not available or capable of doing what the web does, and multi-index idea will work for the things that are capable.
• 34:31
  SPEAKER_04: So elastic and I don't know, Affice 65 could be easily merged at index time at the search time, but other systems cannot.
• 34:43
  SPEAKER_04: So we have a hybrid approach, which is take advantage of both, it reduces your cost, it decreases time to freshness, right?
• 34:51
  SPEAKER_04: It it makes it so that you don't have latency because if I'm searching SharePoint and Elasticsearch the same time, I'm gonna be bound by their performance, and their performance is very good.
• 35:01
  SPEAKER_04: But if I try to do the same thing with those no sense, I'm gonna slow down.
• 35:05
  SPEAKER_04: So so we are a hybrid, do both, power to each other, evolve it, but make sure the search experience for the user will be no different.
• 35:13
  SPEAKER_04: That's critical.
• 35:14
  SPEAKER_01: Yeah, the other thing about having those distributed or those separate indices is that you're retaining context of the source, right?
• 35:23
  SPEAKER_01: Because you're you're saying, well, this is from the CRM.
• 35:25
  SPEAKER_01: Maybe the CRM is going to use slightly different organizing principles than another repository, but you can retain that.
• 35:32
  SPEAKER_01: And really it is about context, right?
• 35:34
  SPEAKER_01: We don't just look at all the information in an enterprise at once, right?
• 35:38
  SPEAKER_01: We look in certain places, and search can help us do that by by looking at those systems and then retaining that context and but but integrating it at the same time.
• 35:47
  SPEAKER_01: So we're in a single pane in a class, but we still have the understanding of those different information sources.
• 35:53
  SPEAKER_04: Yeah, and sir, you know, that's a great point because it just made me think about a use case from one of our customers.
• 35:59
  SPEAKER_04: Because if you think about natural language query where where you can interpret people's intent, okay, and and you ask a question about some HR thing, and we already know the information about HR setting in SharePoint, okay, or or somebody has asked a question about a case thing, and we already know all the cases are in Salesforce.
• 36:22
  SPEAKER_04: Now, what you can do is search, but for the person who asked the HR question, put all the answers from SharePoint on top, right?
• 36:30
  SPEAKER_04: And search, and the person who asked about the cases, put all the Salesforce questions on top, even though SharePoint or Salesforce may have other things that are related.
• 36:39
  SPEAKER_04: So combining sources with intent, which you know, nothing is perfect, but 80% of the time, right?
• 36:48
  SPEAKER_04: You are gonna know what purpose of the systems are actually can improve the search experience.
• 36:53
  SPEAKER_04: It can actually give people the answers, right?
• 36:56
  SPEAKER_04: That Chris was looking for, right?
• 36:58
  SPEAKER_04: I'm gonna give you the answer because I know what you asked about the chart.
• 37:01
  SPEAKER_04: I'm gonna give the answer from the shot.
• 37:02
  SPEAKER_04: And you end up with something that's gonna be right and it's gonna be on top 90, 95, 98% of the time.
• 37:08
  SPEAKER_03: Well, the point, yeah, you're nailing the the key aspect too of this is you know, does the work, you know, the upfront work needs to go into those indexes, but they can be managed and maintained and enhanced and stuff in a mutually exclusive way from the applications that are actually talking to those environments.
• 37:26
  SPEAKER_03: And I feel like, you know, it may be somebody that goes and actually, you know, we always you know think about somebody just typing into a search bar on key information they're looking for, but what happens now when we wire that into a conversational AI bot?
• 37:39
  SPEAKER_03: And that bot is actually looking, and we're mapping those indexes to intents, and those intents are actually searching for content for whatever application for self-service or for for all those things.
• 37:49
  SPEAKER_03: And now what we've done is enabled that time to value to be so short because we've done the upfront work to make sure that the information is there, and now it's just a matter of, like you said, you know, uh Seth, I want to just get my information to the web.
• 38:02
  SPEAKER_03: Well, but now that web is how many different um, you know, channels.
• 38:08
  SPEAKER_03: Yep.
• 38:08
  SPEAKER_04: So you know, you know, just just building on this, Chris.
• 38:11
  SPEAKER_04: So so if you had that chat bot and you did and you decided, okay, I'm gonna do everything, I'm gonna do a customer portal, I'm gonna do a partner portal, I'm gonna do an internal portal, and I'm gonna do a website.
• 38:22
  SPEAKER_04: Yeah, that same thing, right?
• 38:24
  SPEAKER_04: Based on where is it's starting from, can figure out what system to look for, right?
• 38:31
  SPEAKER_04: What information to provide, what security apply, and you just remove all the burden that IT has of okay, for this, I have to set up security that way.
• 38:41
  SPEAKER_04: I have to do this, I have to install this.
• 38:43
  SPEAKER_04: You just you just remove tremendous burden, right?
• 38:46
  SPEAKER_04: And I'm simplifying some degree, but it's a lot easier than used to be.
• 38:50
  SPEAKER_04: That's all.
• 38:51
  SPEAKER_03: Well, I mean, it's it's literally like like we talked about, it's it's where I actually point and create an index, you know, for and maybe that's into a SharePoint environment, or maybe it's into a Jira environment, or maybe it's into you know, you name it file system, you know, that's sitting there.
• 39:07
  SPEAKER_03: And again, guess what?
• 39:08
  SPEAKER_03: Guess who doesn't care about where that information is?
• 39:11
  SPEAKER_03: The user and or the bot or anybody else, nobody cares where that where all of that factory is.
• 39:18
  SPEAKER_03: I just want the information.
• 39:20
  SPEAKER_03: So, you know, now we have the abilities to make it super, you know, more simplistic on the uh, you know, retrieval side of things, and then let's provide that information to all these different uh, you know, application uh modalities and let it go.
• 39:36
  SPEAKER_03: And those modalities can also specify not like you know, like you said, um not only which index, but also you know, potentially, you know, uh creating the context across indexes too, right?
• 39:50
  SPEAKER_03: I mean, we're getting into you know some different, you know, the way that maybe a cognitive thinker would actually pull information, but uh, you know, I want to actually hold the context while I'm searching information, and that's important to the rest of the search.
• 40:02
  SPEAKER_03: So, you know, there's there's there's definitely some key uh you know enabl that now systems have that we never had, you know, in the past.
• 40:12
  SPEAKER_04: And and and and and and and and one of the things that's interesting in terms of technology evolution, right, is if you look at things like Kendra, right?
• 40:21
  SPEAKER_04: Kendra's index is around automating questions and answers.
• 40:26
  SPEAKER_04: Yeah, so it does it does a bunch of things that are very different than other ones.
• 40:30
  SPEAKER_04: So but but not everything fits a question-answer model, right?
• 40:36
  SPEAKER_04: Yeah, some things are being done differently, but the beauty of something like Kendra is okay, for the things that are cute, question answer, let it crawl, index it, and then when you ask the question, our test suggests uh with our customers that the answers that are extracted are more than 90% accurate.
• 40:56
  SPEAKER_04: So for no human intervention, right?
• 40:59
  SPEAKER_04: Yeah, so so now think about Kendra coexisting with Office 65, right?
• 41:06
  SPEAKER_04: Results coming right together, and when you ask something, if if a QA that Kendra should answer, you get that on the top, right?
• 41:15
  SPEAKER_04: You get other results below it, right?
• 41:17
  SPEAKER_03: Yeah, right.
• 41:17
  SPEAKER_03: Right.
• 41:18
  SPEAKER_03: So for for those listening, so Amazon, so Kendra is a service from Amazon, uh AWS service in their AI and ML stack that allows us to do exactly what we're talking about, create indexes off of non-structured data, and then surface those to the organization, whether it be through search portals or through you know chatbots, whatever.
• 41:36
  SPEAKER_03: But it is this incarnate technology that we're talking about now that can be as simple as you know an API call as a building block service that people can build on top of to build these types of products enablements.
• 41:49
  SPEAKER_03: So it's anyway, yeah.
• 41:52
  SPEAKER_04: The way we have done it is we have integrated our product completely with Kendra.
• 41:55
  SPEAKER_04: So there's no none involved, and we you point to it, we call, add the index, ask the question, get the answer.
• 42:02
  SPEAKER_01: So, quick question about a lot of this stuff you can plug in and get some value, right?
• 42:07
  SPEAKER_01: You can go in and index things uh and do uh do clustering, do entity extraction, uh, and and so on.
• 42:15
  SPEAKER_01: Uh but what is the value?
• 42:18
  SPEAKER_01: But but again, a lot of what we're trying to make up for with machine learning and AI are kind of our past sins of bad content management, right?
• 42:27
  SPEAKER_01: Bad content stewardship, you know, poor tagging, uh inconsistent architectures, and so on.
• 42:33
  SPEAKER_01: So moving forward, what what is the role of a reference architecture?
• 42:37
  SPEAKER_01: What is the role of information architecture, taxonomies, and so on?
• 42:41
  SPEAKER_01: What is your thinking there?
• 42:43
  SPEAKER_01: I know I have my my philosophy, and I think you're aligned with that, but why don't you talk a little bit about that?
• 42:49
  SPEAKER_04: So let me step back because I think the problems we've had in the past for tagging uh is not gonna go away.
• 42:59
  SPEAKER_04: Uh, we have had customers who've tried to get your users tagged things, and they this that when they looked at it, everybody tagged things with the first or second item on the drop-down, which basically said that's not possible.
• 43:13
  SPEAKER_04: So I don't think the problem of yesterday is going away.
• 43:16
  SPEAKER_04: I think it's gonna get worse.
• 43:18
  SPEAKER_01: We we actually put in the first value of uh of a tagging uh uh drop uh drop down something like zebra or something argvark or something ridiculous that would be inappropriate and not, and then checked content quality by auditing the number of tags that were with that nonsense term.
• 43:36
  SPEAKER_01: And there was a lot.
• 43:37
  SPEAKER_04: Yeah, I believe that.
• 43:39
  SPEAKER_04: So coming back to your question, I think the value is without it, you're gonna blindly use uh AI to help you.
• 43:47
  SPEAKER_04: And then five years from now, you're gonna turn around and say, ah, stuck.
• 43:52
  SPEAKER_04: So to me, the value of information architecture is for them to help combine the knowledge and expertise they have.
• 44:00
  SPEAKER_04: Have with what you need to do so that what you come up with is a sustainable, solid plan to go forward.
• 44:08
  SPEAKER_04: It costs people a little bit of money, right?
• 44:10
  SPEAKER_04: But is money well worth it?
• 44:12
  SPEAKER_04: Because what you have now is a platform to go forward.
• 44:16
  SPEAKER_04: And then my advice to people is okay, having done that, do what is right, which is every year bring it back to review what's going on, to review how they can improve it and continue on doing incremental improvement because that's the model you should think about.
• 44:33
  SPEAKER_04: If you want to make this stuff work and decrease your cost, improve your productivity, you need to create a continuous way of doing this in a reasonable fashion by combining experts with your team, right?
• 44:47
  SPEAKER_04: And data that comes from six months, eight years of analytics.
• 44:50
  SPEAKER_04: And you're gonna have the best of the world.
• 44:52
  SPEAKER_04: I mean, at the end of it, right?
• 44:53
  SPEAKER_04: That's what Google does.
• 44:54
  SPEAKER_04: But but Google has a lot of people, but they also have billions of searches.
• 45:01
  SPEAKER_02: Right, right, right.
• 45:02
  SPEAKER_04: So relatively the numbers are gonna be much smaller, but that's what you want to do.
• 45:08
  SPEAKER_03: You're getting into well, usually you're getting you're also getting into now.
• 45:11
  SPEAKER_03: It's one thing to have a service that we can tie into from, let's say, a vendor that that provides infrastructure.
• 45:18
  SPEAKER_03: It's another thing to um, because results may vary given that there's the semantic nature of the data that I'm actually putting into it, building it, training it.
• 45:28
  SPEAKER_03: So I need to make sure that I baseline my confidence scores.
• 45:32
  SPEAKER_03: And those confidence scores need to, you know, give me the ability to say, okay, like you say, in my audit, here's the confidence scores I'm getting from this corpus of data that I'm actually pulling in.
• 45:44
  SPEAKER_03: Is it giving me the ability to, you know, yeah, am I getting the right information, the right results, and those right, you know, the right, you know, confidence scores as I pull that information up, right?
• 45:54
  SPEAKER_03: And measure that.
• 45:56
  SPEAKER_03: And then, like you said, create the baseline and then continue to test it over time just to make sure that you know that we're not entering, you know, bias into this, into the the results and stuff.
• 46:04
  SPEAKER_04: Yeah, and Chris, if it if if if I mean this is not is it's not hard thinking.
• 46:10
  SPEAKER_04: If you bring an expert in a year later, and and I'm just gonna use a Kendra.
• 46:15
  SPEAKER_04: Let's assume you did things, you did things, you don't know Kendra exists.
• 46:18
  SPEAKER_04: Kendra comes in, an expert comes in and says, gee guys, what I see is you should use Kendra for this, because if you do, this is gonna happen, right?
• 46:27
  SPEAKER_04: And you you remove the burden on a bunch of people to track every technology in the world to see how they can make things better.
• 46:34
  SPEAKER_04: There's a lot of value, right?
• 46:36
  SPEAKER_04: To once in a while to take a breeder and bring somebody who knows what's going on in the marketplace to come and help you, right?
• 46:42
  SPEAKER_04: Yeah, no, we are we we are technology vendors, we're gonna continue on moving fast.
• 46:46
  SPEAKER_04: So we will look at technology from our perspective, but we don't look from a technology from perspective of a use case of a specific customer, right?
• 46:56
  SPEAKER_04: We don't go and say, okay, dear customer with 400,000 people, let us look at all use cases.
• 47:02
  SPEAKER_04: And but but but then they don't look at technology either.
• 47:05
  SPEAKER_04: So my my advocacy is things are technology enabling you to do wonderful things.
• 47:13
  SPEAKER_04: Yeah, don't stop, right?
• 47:15
  SPEAKER_04: Yeah, use use people who can help you on a continuous basis, and and what's gonna happen is five years from now, you're gonna have uh as good an experience, right, as people have with Google, it's gonna cost you a lot less money and it's gonna be a lot easier.
• 47:31
  SPEAKER_03: The easiest way to measure that is usage, right?
• 47:34
  SPEAKER_03: If I see my usage decrease, then I know that I'm not returning the results I need.
• 47:38
  SPEAKER_03: And so, you know, the the beauty is like you said, is to continually making it, you know, continually making it better and improving should see usage go up and or go up and plateau given the number of people, you know, users that are using it, things like that.
• 47:52
  SPEAKER_03: So yeah, it's great.
• 47:54
  SPEAKER_04: Yeah, we track we track every click when people use our solution.
• 47:57
  SPEAKER_04: We collect track every click, right?
• 48:01
  SPEAKER_04: We crack, we track uh not only the click, did you do something with it?
• 48:05
  SPEAKER_04: We crack so so we have a lot of data, just like you have what websites, right?
• 48:10
  SPEAKER_04: You have to track data exactly.
• 48:11
  SPEAKER_04: You need to know what happened, right?
• 48:14
  SPEAKER_04: Right.
• 48:14
  SPEAKER_04: We actually give the user a complaint button where they can click a thing, we will send the administrators the URL for the page that they resulted, and they can tell, hey, I saw this, doesn't make any sense to me.
• 48:29
  SPEAKER_04: Right, right.
• 48:29
  SPEAKER_04: So we can actually interact with the feedback also on top of back in the day.
• 48:33
  SPEAKER_01: Yeah, the the the content analytics is incredibly important, especially aligned with those use cases that you mentioned, right?
• 48:40
  SPEAKER_01: So that's a critical piece.
• 48:42
  SPEAKER_01: Uh, and and the and the people who are using it.
• 48:44
  SPEAKER_01: I want to just touch on the the tagging uh conundrum, right?
• 48:48
  SPEAKER_01: Because again, it's I think from tagging perspective, yes, it's still going to be a problem, but you have to make the punishment fit the crime, right?
• 48:55
  SPEAKER_01: If it's really high value content, then you're going to want to have more human intervention and judgment.
• 49:01
  SPEAKER_01: But you can use auto tagging, you can use classification uh algorithms based on that reference data to say here's our isness, here's our aboutness, this is what this content is, here's how we're describing it, here are the values that we've decided, you know, the names of our products or services or solutions or whatever it is, you know, troubleshooting guides, whatever it is, but you're tagging things so that the system can retrieve.
• 49:26
  SPEAKER_01: But again, I think you know, to your point, that's still going to be a challenge, but you have to make sure you're you're applying the right, if it's really high value content that people need to make decisions about or use to purchase products or or regulatory constraints, then that stuff needs human judgment, right?
• 49:43
  SPEAKER_01: The systems can help a lot.
• 49:45
  SPEAKER_04: Yeah, so to hybrid is a good.
• 49:47
  SPEAKER_04: Yeah, I'm I'm I'm not sure I agree with you.
• 49:49
  SPEAKER_04: And yeah, go ahead.
• 49:50
  SPEAKER_04: To me, to me, there is nothing high value content to me is I decide, not some administrator, right?
• 49:58
  SPEAKER_04: But I'm trying to find somebody, something I'm a third shift somebody.
• 50:02
  SPEAKER_04: I used to work third shift at computer vision, nobody around.
• 50:05
  SPEAKER_04: If I'm trying to find it, and and some administrator decides not high value, and I can't find it, and I'm gonna have to waste four hours, is high value to me.
• 50:15
  SPEAKER_04: So the reason I say I disagree is I think that tagging always needs human intervention.
• 50:20
  SPEAKER_01: Yeah.
• 50:21
  SPEAKER_01: And I think you can make some distinctions to say, you know, get rid of the draft uh documents, unless you say there is something in that draft that didn't make it to final or something like that.
• 50:31
  SPEAKER_04: But I think you do need to say, you know, this is regulatory and it's going to require a higher degree of focus versus even there, I will hire an expert to tell me as opposed to me deciding, because what I worry about is when I'm deciding running IT or something, I'm deciding based on a budget, not based on value.
• 50:48
  SPEAKER_03: But but we're also getting to the point, too, where what you know, and the the beauty of artificial intelligence and machine learning gives us the ability to do is to break the predefined norms that we've lived in for so long and give me the ability to say, listen, like you said, hey, um this is you know the information I'm searching for, but let me take into consideration more data elements that would provide me maybe a more holistic search like sentiment analysis or maybe historical context or something else, as opposed to me just saying, I want information about Massood.
• 51:24
  SPEAKER_03: Well, let's what information it's and it's got to think about what information I want instead of just giving me just a show up and throw up type style result set.
• 51:33
  SPEAKER_03: Maybe because of my role, my um, you know, my persona, my attribution, and whatever I'm doing is giving me more holistic context of of what you know what the search results should be.
• 51:46
  SPEAKER_03: Yeah, absolutely.
• 51:47
  SPEAKER_03: And that's that's the thing I love about it, is it doesn't it gives me the ability to blow up the predefinition.
• 51:53
  SPEAKER_01: And I do agree with you, Masood, that that value is in the eye of the beholder, right?
• 51:58
  SPEAKER_01: The value of that answer is in the eye of the beholder, right?
• 52:00
  SPEAKER_01: So it's it's you know, what is relevant to you may not be relevant to me.
• 52:04
  SPEAKER_01: Yep.
• 52:05
  SPEAKER_01: So yeah, yeah, that makes a lot of sense.
• 52:07
  SPEAKER_01: Well, we're coming up at the top of the hour, we've got a couple minutes left.
• 52:11
  SPEAKER_01: Um, what what else would you say, you know, organizations need to really look at and you know, to to make sure that these things are going to work effectively?
• 52:20
  SPEAKER_01: I mean, again, there's no there's no magic uh pixie dust, right?
• 52:25
  SPEAKER_01: AI is not going to, you know, be the silver bullet.
• 52:28
  SPEAKER_01: Um and we still need to pay attention to certain things and we still need to have governance and metrics and you know, as you say, that continuous improvement uh and and content uh gets created by humans and needs to be uh uh evaluated by humans and aligned with the use cases for other humans, right?
• 52:49
  SPEAKER_01: I mean, it really it's it's about interacting with with those systems.
• 52:52
  SPEAKER_01: What else would you say that, you know, if an organization is looking at this at an enterprise level, um, you know, big picture, what do they really need to to make sure they have in place?
• 53:03
  SPEAKER_01: You mentioned metrics, you mentioned use cases and scenarios, um, you know, ongoing governance and continuous improvement.
• 53:10
  SPEAKER_01: Is there anything else that comes to mind that that you would kind of advise organizations to keep top of mind as they go down this path?
• 53:19
  SPEAKER_04: Yeah, my my biggest worry about the future of our market is the perceived degree of difficulty.
• 53:27
  SPEAKER_04: What I worry about is um is three years from now.
• 53:34
  SPEAKER_04: And people say, God, to run it by search, I need six people managing it.
• 53:40
  SPEAKER_04: Right.
• 53:42
  SPEAKER_04: Uh you know, but on Salesforce, with a thousand salespeople, I have half a person doing everything.
• 53:49
  SPEAKER_04: That that I think that's what people need to understand.
• 53:53
  SPEAKER_04: That you do need to think long term, you need to look at total cost of ownership, right?
• 53:59
  SPEAKER_04: Because all of the things we talk about, if the total cost of ownership or the burden on the organization is too high, we'll fail.
• 54:07
  SPEAKER_04: And I and I'm convinced the reason why search got a bad name is because of that issue singularly.
• 54:15
  SPEAKER_04: And and yes, there were bird, there were technological impediments, right?
• 54:18
  SPEAKER_04: There were problems in terms of how we, the vendors, deliver.
• 54:23
  SPEAKER_04: But I I the thing that I most worry about is how do we get to the place where people actually look at this not as I purchase implemented, but I purchased and kept it for five years.
• 54:34
  SPEAKER_04: And what is my cost of those five years?
• 54:36
  SPEAKER_04: How much resources do I have to have it?
• 54:39
  SPEAKER_04: Because if they do that, then they will commit to it, whatever the cost is, and therefore it will survive.
• 54:45
  SPEAKER_04: If they don't do that, the year later it becomes, oh, we need two people.
• 54:48
  SPEAKER_04: I don't have the budget.
• 54:49
  SPEAKER_01: And and it brings up the point of value, and maybe we can end on this ROI value.
• 54:55
  SPEAKER_01: And again, the cost is irrelevant if we don't have the value.
• 54:58
  SPEAKER_01: So, why is value so difficult to establish here?
• 55:01
  SPEAKER_01: And what you know, what organizations need to do to kind of establish that business value.
• 55:07
  SPEAKER_04: I don't know if it's difficult to establish with visionary people or people who have forgotten or didn't go through the past seven, eight years.
• 55:23
  SPEAKER_04: You know, it's like go ahead.
• 55:25
  SPEAKER_03: Well, I was gonna say, you know, well, I mean, I want I want you to end your thought because what I wanted to do is it was basically say, listen, organizations like yours are demystifying the dark arts, right?
• 55:38
  SPEAKER_03: And helping bring bring this, you know, what what people think is like a very hard concept to get, which it really isn't.
• 55:45
  SPEAKER_03: It's it's more along lines of and it's being you know a little bit more democratized than it has in the in the past.
• 55:50
  SPEAKER_03: But I think that's the that's the key is that organizations like yours are actually doing that for other companies to help them understand that it's not a set of dark arts and mystical, you know, creatures that we're dealing with here.
• 56:03
  SPEAKER_03: It's just data and approaches to it.
• 56:06
  SPEAKER_03: But Masud, where you know, uh the and maybe this is something, you know, I'll let you you know take all the closing remarks, but um, where do people find more out more about you and about your organization?
• 56:18
  SPEAKER_04: Uh surely I'm on LinkedIn, so if you want to find me.
• 56:22
  SPEAKER_03: Well, I mean, I mean, just in terms of feel feel free to give a little commercial about the company uh the stuff you're doing.
• 56:28
  SPEAKER_04: We are we are B insight, www.bainsight.com.
• 56:33
  SPEAKER_04: I'm I'm not a promoter of our company often.
• 56:36
  SPEAKER_04: I think people who know me know that.
• 56:38
  SPEAKER_04: Uh but but you can you can Google us and find us.
• 56:41
  SPEAKER_04: And uh yeah, I do have one one final comment that that is really important.
• 56:46
  SPEAKER_04: We'll also have that information in the show notes.
• 56:48
  SPEAKER_01: Yeah.
• 56:49
  SPEAKER_04: I think it's really important for people to realize, and in today's world, right, that enterprises implementations are not multi-million dollar implementations, they're multi-hundred thousand dollar implementations.
• 57:01
  SPEAKER_04: And and that investment compared to the value brings, and coming back set to what you said, is almost nothing for most of the large organizations.
• 57:11
  SPEAKER_04: Yeah, sure.
• 57:12
  SPEAKER_04: But but you could also, right?
• 57:14
  SPEAKER_04: You could also start uh with coming back to Chris what you said about start small and grow.
• 57:20
  SPEAKER_04: You could start as little as$50,000,$75,000 a year and have a pretty good solution for a department with some connectors, experience it.
• 57:27
  SPEAKER_04: And if you don't like it, you don't lose a lot.
• 57:29
  SPEAKER_04: And if you like it, you grow it.
• 57:30
  SPEAKER_04: But the world, the world has changed.
• 57:33
  SPEAKER_04: Stop thinking million dollar, two million dollar, five million dollar year platforms and five people building something for you.
• 57:39
  SPEAKER_04: That's the wrong that that is the old approach.
• 57:42
  SPEAKER_03: Hey, listen, thanks for all the good information.
• 57:44
  SPEAKER_03: It's a pleasure to have you on and and to meet you and see all the great stuff you guys are doing as a company and and you personally.
• 57:50
  SPEAKER_03: So appreciate it.
• 57:52
  SPEAKER_01: Thank you, Mr.
• 57:53
  SPEAKER_01: Appreciate you.
• 57:54
  SPEAKER_04: Thank you very much, Chris, for for your time, guys.
• 57:56
  SPEAKER_04: And you too.
• 57:57
  SPEAKER_04: Have a great weekend, man.
• 57:58
  SPEAKER_01: Great.
• 58:00
  SPEAKER_01: See you all next time.
• 58:01
  SPEAKER_01: Okay, thanks, everyone.
• 58:07
  SPEAKER_00: Thanks for listening to the early AI podcast.