In this Digital CxO Leadership Insights series video, Mike Vizard speaks with Ajay Khanna of Tellius about managing decision intelligence.
Mike Vizard: Hey guys, welcome to the latest Digital CxO Leadership videocast. I’m your host Mike Vizard. Today, we’re with Ajay Khanna, who is CEO of Tellius. They are a provider of a platform that helps you manage decision intelligence. Ajay, welcome to the show.
Ajay Khanna: Hey Mike. Great to be here with you today.
Mike Vizard: I think most organizations today are trying to get more insight out of the data they have and maybe even collect more data to drive more insights. And yet we struggle with the data. We have loads of it, but we don’t know how to manage it or organize it. Very few organizations I know honestly, would get a Good Housekeeping seal of approval for the way they manage their data. So my question to you is: what today is the fundamental challenge with managing data that we’re facing in order to drive some more informed decisions off of that data?
Ajay Khanna: Yeah, no, that’s a great question. The way we look at that, Mike, is that there’s a massive insights gap created by the silos between the BI dashboarding tools and some of these machine learning and AI tools. Now, when we look at the BI tools, and they have been around for some time, that most of the organizations are using that for reporting and dashboarding, but they want to go beyond reporting and dashboarding. They want to uncover those insights, but they don’t have the capabilities within these tools to be able to do that. Now, when we look at the other side of the spectrum, the machine learning and AI tools are fully capable to uncover insights and use those machine learning models to scan through all the data. But those tools are limited to very technical elites. You know, someone who has a masters or PhD in statistics or data science. So this whole gap is leading to, you know, this huge issue with insights not being generated and that’s leading to delayed decision making and that’s kind of what’s, you know, costing a lot of huge issues with the inefficiencies and lost revenue for the businesses at this point of time. So the way we look at that is, how do we bridge this gap now? How do we bring the power of machine learning and AI to the masses, and the masses being the business teams and the analytics teams, but, but still keep it easy to use for those business teams, because most of them may not understand SQL or Python or the, you know, advanced machine learning concepts.
Mike Vizard: Well, how do we do that? Because in theory in the machine learning algorithms should be to discover and organize our data in a way that doesn’t require us to do such manual effort that we have been doing in the past. So what is the state of the art right now?
Ajay Khanna: Yeah, so at Tellius, we built this decision intelligence platform and the way we did it, we bring in the ease of use of a modern interface. We can provide like a Google like search interface where you can ask a question in plain English. So for example, if I’m head of marketing in a big organization, and I want to know, you know, what happened to my campaign conversions for a certain campaign last week. And I, instead of going, you know, going through writing some SQL logic, I can ask this question in plain English and the system will scan through all the data, do the right joints, whatever needs to be done and get you those answer pretty quickly. But that’s only what we call the what part of the question, like what happened, but then most of the businesses want to know why things changed. So providing a seamless experience where now from you can go from what happened to why things changed, where now you can, you know, click a button or see a drop in your campaign conversion and be able to say, okay, and tell me what’s driving this change. And, the system actually goes and looks through all the data, finds those key drivers, but more importantly, presents it in a consumable form for the business user, because if you were to bring it back to a machine learning model, which will have, you know, a black box machine learning model, the business user wouldn’t know what to do with that.
So, the ability for a decision intelligence platform to translate that complex logic into something which a business user can understand; what we call using a natural language narrative, and then be able to provide that and say, you know what, because your campaign conversion went down because you know, a certain set of people who used to like your product, you know, aren’t resonating with maybe the new messaging you have, right. And maybe a certain, and also your marketing budget, may have shifted that you’re spending more money in another area. So these are the ones who are driving a change in that marketing campaign.
So that’s the, I call it the why part. So finding that why, and then you also go to how, how meaning, so what, how, how can I improve it now? And that’s kind of where you can actually ask this question and say, okay, what are the customer or the prospects I should be targeting, who will resonate better with the messaging I have? And the system can actually, again, use some machine learning, find those key segments, and then again, present it in a form which the business user can understand. So what decision intelligence does is short circuits this process from data to decisions by putting all this, what, why, how within a single unified platform.
Mike Vizard: Are we going to get to the point where I don’t even need to say, type? Can I just talk to the machine and say, “Hey tell me the three things that are likely to get me fired today, or tell me, you know, what happened to these four campaigns compared to these three other campaigns?” And we’re just going to talk to it.
Ajay Khanna: No, that that’s a great point. We do provide the ability to speak like we have a as language interface where you can speak. Right now, we support in English and Spanish and in couple other languages, but you can certainly speak into that machine and the machine will say and provide and create the chart on the fly. But there’s another area we are really excited about what we call as the proactive intelligence is what you are referring to, is proactive intelligences, where you can think of like a feed, like you have a social media feed, right? Like we, everyone has some kind of feed, you know, Instagram or Facebook or whatever. So you will subscribe to certain, you will be following certain key topics, for example, or following certain people. Now you have feed becomes curated to the things you’re normally, you know, more interested in. And I think the similar concept is already emerging in decision intelligence, where you can subscribe to those metrics, the metrics being, let’s say, I am in a salesperson for the northeast region. Like, I really don’t care what happened in the west, right. I just want to care about what’s happening to the sales performance in Northeast region. So I can go and say, you know, tell me and pick some of the sales metrics for a northeast region and just kind of subscribe to that. And the system will then keep monitoring these metrics going up and down and then say, okay, these are the metrics which are going to going outside the range and then do the key driver analysis and kind of push it to you through whatever channel you feel convenient, like could be Slack or, you know, your email, or whatever the channel would be. So, that’s going to, you know, things we already starting to see. And there, there is when we talk to our customers, you know, they, they certainly want those kind of capabilities to be able to a push. I call it a push and start of the pull mode mechanism of you know, analyzing data.
Mike Vizard: Do you think that business executives trust the data? I mean, historically a lot of the times they always were suspect of a BI tool because, you know, the analysis that would come back wouldn’t line up with their own personal experiences, or then sometimes it’s hard to know whether or not that is just merely indigestion that they’re having when they’re thinking about that, or whether or not there’s actually something wrong with the data itself, but it always feels like the business side’s a little suspect of the whole IT relationship. So how do we get to the point where we can have confidence in the data?
Ajay Khanna: I think there’s two parts. I mean, one thing is that the trust in the data is certainly an ongoing topic, right? Like you remind any discussion, like five years back, or 10 years back, I think this, you know, continue to be a, be a hard topic. And this is a very valid point. You know, businesses need to be able to trust the data there. I think there is certainly some progress has been made in terms of how can we detect some of those issues in a more autonomous manner. But I don’t think that problem, you know, that’s going to completely disappear. You know, it’s certainly improving, improving where you know, some of the detection happens automatically, but I think there’s another aspect emerging, which is a second point is providing the transparency into it. I think that’s something we believe big in that, that it’s not about just providing the results or the answers, but also providing transparency into how we got those results. And also maybe if we had some missing data, or in certain cases, you know, some of the columns did not have the records information. So providing that transparency along with that, because there is going to be, you know, cases where, you know, people who are entering that information in the systems isn’t right. Or, I mean, all these things will continue, but you have to provide the transparency. I think that’s an area which we are certainly, you know, seeing a lot of progress and customers and the users really expect that.
Mike Vizard: As we go along, do you think we’ll get to the point where, you mentioned the word proactive, but can I just get to a point where I think every business leader has like four or five KPIs that they’re tracking, and I just want to say, “Hey, these are my KPIs and give me a shout when one of them gets violated or somehow or other that a KPI starts falling off in a way that I’m going to view as maybe suboptimal.” But, how smart can the machines and the systems get?
Ajay Khanna: Yeah, I think there’s – and that’s the direction we are heading. I would say there are things you know, and the things you don’t know, right. So, you know, you want to track certain metrics but there could be other metrics you want to track and in, you know, how do you bring that intelligence? So, one thing we are working on is we can have customers or users specify that these are the four or five metrics they care about. But then one area of we are working on is like, but the users then goes and start asking other questions, like, okay, they will come and say, okay, I care about campaign conversion. I care about, you know, revenue, I care about profit, right? So they’ll subscribe to that. But then, then they go to a search interface. We have a search interface where they start asking questions about some other metrics, like they could be asking metrics about sales, effectiveness, maybe, you know, you know, how much time it takes to close a deal. Like there’s going to be other metrics they start asking.
And the way we are working on is called the feedback loop where they have an explicit specification and the implicit learning. So we can actually use that information to that then say, you know what, even though you didn’t put that into the list, you’ve been asking, you know, a lot of questions around that. And let me just add that to your feed. So it’s not only that I’m doing a root cause analysis around other metrics, but also the metrics you care about. That’s an area which is something we are working on, but then there’s another interesting part to the questions you ask. You know, at 8:00 a.m. , when you walk into the office or whatever, or whatever your home office is these days, that are different than what you worry about at 3:00 p.m. in the afternoon. Right?
So at 8:00 a.m., you come in, and you want to look at, you know, okay; let’s say if I’m the head of sales, and I want to look at, you know, what did my sales reps do? You know, yesterday, last week, right? You know, how many calls they made, or, you know, how many people they talked to, whatever the metrics they’re looking at in the morning? But the questions in the afternoon could be very entirely different because you may be, you know, flipping that and maybe going and talking to the other teams around the help you need from the marketing team or your engineering team, or whatever the case may be. Right? So how do you personalize this information which has time context? Like, you know, your Google maps on a weekend, you know, you go into the car and your system will say, you know, it, it’s ten minutes to the grocery store. Like, the Google maps knows that, you know, you typically on a Sunday morning, you go and, you know, go shopping for milk in the morning. So Google knows.
And I think you bringing that kind of information in there is; we are still very, at the least stages of that. But I think once we crack that, I think that’s kind where it becomes more contextual to what people are asking. And sometimes without even them explicitly specifying it, because we are all lazy in our own way. And then the system just figures this out and then push this information to you, but that’s kind of the direction we are heading.
Mike Vizard: So the system may know more about me than I do on any given day. And it’s going to help me to figure out the parts that I do care about. It may be a little scary on a certain level, but honestly, there’s a lot of things you got to think about and you can’t remember it all, right?
Ajay Khanna: Yep. You can’t remember it all. Yeah. Yeah.
Mike Vizard: What’s your best advice to folks then about how to get started down this path? I mean, other than the fact that they can go buy something and set it up. But are there things that they should be doing with their data today that will pay off tomorrow?
Ajay Khanna: Yeah, I would say there’s three things I’ll say. And, I guess, the concept I call, Three Ps, and the first being the problem and then people and the process. So the problem we are going to start with is the problem of the analytics. I think the bigger issue everyone has is that they start with the technology first and then go to the problem. As an example, I think the ML is a very you know, classic case where there will be a typical CEO level initiative where CEOs say we got to go, you know, the ML data science route, and it comes down to the, you know, the VPs level. And then they start you know, hiring a bunch of data scientists and, you know, get a bunch of tools. And then they said like, let’s, you know, use ML to do this.
So they’re approaching it from the technology first and the problem second. And then they say, what can I solve? I think that’s really, really wrong. And leading to, I think like the failure rate on some of the machine learning AI projects is close to 80%, which isn’t good because there’s a lot of investment going into these initiatives. So you have to start with the problem. You’ve got to figure out what is the problem, and, you know, we typically advise our customers to, you know, create this effort and kind of the value on this and create that map to figure out okay; which are the problems which have the best effectiveness or the value you can create and then figure out like how much effort you need. And then from there, figure out, okay, what kind of ML technologies you need to make this happen? So, the problem.
I think the second piece is the people. I think there’s certainly a lot of concern where people don’t really fully understand what ML would do, you know, to their jobs, or like, you know, is it my job at jeopardy? And I think it’s just providing that the education knowledge on like that, this is all good, right? It’s going to create benefits and, you know, create more efficiencies. And, you know, someone can, you know, finish their work at, you know, five o’clock instead of finishing at a seven o’clock and they, you know, can play baseball with their kids. Right? I mean, so, providing that aspect to people and educating them on what that could do is the second part, and also a little bit on the analytics majority for those people, because where you are in the journey; because there are organizations who are still at the early stages in the organization, a little bit more advanced.
And the last piece is the process, the processes, where, how do you reduce the friction between the data and people? I think there is a lot of talk about the governance, and those are all good things to put governance around data, but at the same time, how do you balance the governance with the accessibility and self-service aspect there? So, how do you look at your processes and say what processes need to be improved to reduce the friction, so you can make all that, that stuff happen, where you can now use the technologies to solve a problem and improve your business outcomes?
Mike Vizard: All right. Ajay, I think you just said that, no matter how smart that AI hammer is, everything is a nail, right?
Ajay Khanna: Yep. Absolutely.
Mike Vizard: Hey Ajay, thanks for being on the show.
Ajay Khanna: Great talking to you, Mike. Great to be on the show today.
Mike Vizard: All right. Thank you all for listening to our show. You can find this episode on digitalcxo.com, along with our other episodes. We invite you to check them all out. And once again, thank you all for the watch.