Leadership Insights: AI in Enterprise IT Environments

Transcript

Mike Vizard: Hey, guys, welcome to the latest edition of the Digital CXO videocast. I’m your host Mike Vizard. Today we’re with Anand Mahurkar, who’s the CEO from Findability Sciences. They are a provider of technologies that are used to build various types of AI applications in the enterprise. Anand, welcome to the show.

Anand Mahurkar: Thank you, Mike. Thank you for having me, and thank you for the listeners tuning into this podcast.

Mike Vizard: One of the things that you just did, recently, is you created a way for organizations to white-label AI technologies, and that’s kind of an interesting idea, ’cause a lot of organizations are spending a lot of time and effort in hiring data scientists to go build all this stuff. So, if I white-label that stuff, can I skip all the heartache and just enjoy the benefits?

Anand Mahurkar: Yes, Mike, and just to answer your question a little bit, let me step back and tell you that Findability Science _____ traditional enterprises through its enterprise AI product line, and for a number of years, we offer those products and solutions for their inhouse usage. But over the last couple of years, we’ve found that most of the end customers are looking for AI-powered products, and there are thousands of legacy technology software and hardware products, which are not AI-enabled. And if those legacy or traditional enterprises want to power their applications through AI, they have to undergo a lot of changes within their organization to power those with AI. So, we found our technology very suitable, where we can plug in our already prebuilt technology into legacy and traditional products, and power them with AI. So, yeah, you can call it as white-label, Mike, in one way, but the second way is to really powering the traditional enterprise products with the modern latest AI, and serve customers with better increased revenues, efficiency, and lots of benefits with _____ _____ not really putting all the effort which typically you have to put when you want to build AI products.

Mike Vizard: So, how does that work? Are we inserting AI models into existing legacy applications? And give me an example of that.

Anand Mahurkar: So, for example, I’ll give you one software example and one hardware example. One of the largest ad tech companies out of Silicon Valley, which is about 13-, 14-year-old company, they have a very good advertising technology platform whereby they distribute advertisements to the mobile users or mobile platforms. But that is not AI-powered. So what we have done is we have integrated our predictive analytics models into the platform, whereby, now they have improved their click-through rate, because they know, through predictions, who is likely to click and who are the likely targets of those advertisers, so they can efficiently target onto that. They also can predict effectively, because we give them big prices predictions on the platform.

So, the click rate and price prediction have improved both their engagement with their end customers or end users and their profits, because of increased their revenue through their bidding process. So that’s a typical example of an enterprise software platform integrating predictive or AI algorithms to predict or forecast something. The second example we are working on is one of the largest scanner manufacturer, and scanner, as you can imagine, the technology has matured, the hardware is sitting on every table at house or in home offices or in businesses. But what we have done is we have creatively integrated our natural language programming tool into now the scanner. So when you scan _____ a 50-page document or a 100-page document, not only you will get the scanned copy of that document, but the natural language programming or processing will automatically summarize, read the document and summarize, maybe in two paragraphs or one page for you, so that you don’t need to go through the entire document unless you are required to do that.

Now, that increases the interesting feature on scanner, it increases revenue for the company who manufactures scanners, and the end customer gets a very cool feature on hardware. So these are the two different examples of a software and a hardware, how they can enable their products with AI.

Mike Vizard: Do you think a lot of organizations are investing in data science teams that are, essentially, inventing the same wheel over and over again, and need to start distinguishing between things that are gonna be fairly commonplace and things that truly differentiate one business form the next.

Anand Mahurkar: Right, and, Mike, this is a very interesting question, because I keep talking to our customer, if we just take predictive analytics, there are a number of tools available, ranging from opensource to paid programs. But there is a finite thing you can do with predictive analytics. Statistics is not a new subject. Statistics is in nature, it’s available all around us. So, what humans have really tried doing is that teaching computer statistics and doing the predictive modeling, et cetera. But to some extent, we are maxed out on that, so when you’re talking about data science, it’s not necessarily just about developing the same programs and doing the same things. Those should be _____.

What data sciences team should focus on is really the data innovation, how they can get the data into their systems, use these ready-made models or technologies, and move faster in terms of what will happen. Because staying with the predictive as an example, typically, these technologies will tell you what will happen. Where the businesses need to go is what to do, and that what-to-do part is where the data science should focus, and not what will happen. So, yes, what you are saying is right, that no need to reinvent the wheels. There’s a lot available, you can power your businesses much quicker, and then do, most important, domain-centric things for your organization with your data science team.

Mike Vizard: How do I maintain that exactly, if you’re gonna provide the core model or whatever it is that does the deliverable, a lot of these things drift over time, so how does that get updated and how is that continuously being maintained as – once I insert it into some sort of production environment?

Anand Mahurkar: So, these technologies are also now very modern, so the tools and the technologies we help build, Mike, they are self-learning in their nature, they are quite dynamic, it’s not just a single model or a single natural language program working on it. It’s a continuous learning process, so it needs very less maintenance going forward. So, gone are those days where you are taking one model and deploying it, and then after every six months or one year, you are updating that with your change data. The data, in terms of its variety, is increasing, as you know, so these models also have now come to a level where they adjust themselves automatically with the dynamism of the data and the variety of the data.

So, Mike, now the customers don’t need to really babysit these models anymore, or they don’t need to update. Yes, there is the requirement of refreshing the data and data learning itself to the models, and those are the technologies and advanced algorithms we have developed.

Mike Vizard: What do you see as the mistakes that organizations are making? If you had to offer some advice to some digital CXOs about how to think about AI or approach it, what would it be?

Anand Mahurkar: So, one of the most important things I am preaching to my customers is really about data, because as we just discussed, the algorithms, the models, the programs, they’re fairly, now, standard, and there is enough you can do which has already been done. But data is the most important asset for all the CXOs. And what we are now talking to our customers with our experience, that we talked about big data a lot over the last 10-15 years. Most of the organization shave big-data implementations. Big data, by its nature, has three Vs. V as in velocity, variety, and volume.

But for the AI use cases, you really don’t need volume, and you do not need velocity, as well. What you need is the variety, and we have now started calling it as a wide data. So, my request to all the CXOs is that you should spend energy in understanding the data in your organization and outside your organization, and create a capability for wide data. Once that is done, the rest of the things are fairly easy. That’s the most challenging area we are seeing, over years, the organizations struggle is with their data assets.

Mike Vizard: How much does the quality of the data matter? ‘Cause a lot of the organizations that I know wouldn’t get a good housekeeping seal of approval for the way they manage data. Do we really need to clean that up to drive the AI model? Or is there some forgiveness in there?

Anand Mahurkar: There is a little bit of forgiveness and lots of tools now are coming to _____, but, yes, quality of the data is important. So, when I define wide data, in addition to the variety I just talked about, I also talk about veracity or the accuracy of the data. So, it is important, but there’s a lot of help to, now, data exploration tools and automatic quality detection and correction and et cetera. And then there’s a little bit of forgiveness because the models do understand if there are some fields that are missing or there is some empty or zero or a null field, it can take care of those. So, there is a little bit of forgiveness, but I still want to stress that accuracy is important for eventually getting you good results.

Mike Vizard: Do you think that maybe digital CXOs have too big an expectation for AI and they think it’s some sort of magic silver bullet, and maybe they need to have more realistic expectations?

Anand Mahurkar: You’re right, and most of the CXOs I have met in my career, they somehow feel that AI should be a magic, and it’s not a magic. We want, we want it like a Hollywood movie as something to happen, but in a real enterprise AI implementation, what we talk about to the CXOs is that they should not be jumping the guns and look for financial return on investment. That _____ for any organization or business, financial or economic investment is that main aim, but for AI, there are two return on investment they should first look for, is the capability ROI and strategic ROI. If the CXOs focus on capability and strategic ROI, without expecting magic and, as I mentioned, build their wide data strategies, set up their data platforms, make their data fields available to different algorithms to test. Once that capability and strategic ROI comes, the financial ROI will follow. And so, it will eventually turn into magic, but there’s quite a heavy lifting required before you go there.

Mike Vizard: There was a lot of skepticism about AI, early on, and then the pandemic came and it seemed like investments went up because people were trying to transform their businesses. And now, I think we hear more about the Great Resignation and it’s getting harder to find people. So what is the appetite for AI, right now?

Anand Mahurkar: AI appetite has just increased dramatically, Mike, and post-pandemic, actually, I am now saying we were three to five years ahead of the curve offering these solutions. Because the same customers I met a couple of years ago, they’re saying, “Oh, you were talking about this? We want that.” So, post-pandemic, what has happened _____ _____ whether it is a resignation, whether it is just, this whole supply chain disruption happened during the pandemic and lockdown, the businesses are disrupted so heavily that all are now looking at in terms of becoming resilient, and how they can be self-dependent on their own businesses. So, there is a definitely big appetite, in fact, most of the gurus and pundits of the AI are saying that the revolution has just begun.

And we are also experiencing, as Findability Sciences, we have tremendous customer interest. The skepticism is gone, because we have shown tangible financial return on investment, going through that capability and strategic return on investment, companies have started getting the financial return on investment, and therefore, they want to do more. And Mike, traditional enterprises, they are now getting onto this bandwagon of AI, and they will be benefited for sure.

Mike Vizard: Last question, do you think that we need to put some governance around how AI is employed? ‘Cause a lot of concerns from the government, well, I think a lot of digital CXOs, for that matter, are kind of waiting to understand how and when they can apply AI. So, in some ways, do we need more regulation _____ keep the AI _____ at the rate it is?

Anand Mahurkar: So, Mike, our work is in the traditional enterprises, and when an organization is deploying artificial intelligence for their own usage, yes, governance for their internal things is very important, and _____ are following that. Because not only implementation of AI but _____ of the outcome of the AI, how that is used. But typically, if you see the wide spectrum of the usage is in manufacturing or retail or financial services. It’s not really impacting on something outside or individuals or personal lives; it’s really enterprises using the AI for their business decisions. So, having said that, governance is important. I’m not really talking about the public domain, but within the organization, the companies are putting their internal data governance, security, how that AI output is used, explainability. And we, as an organization, support that.

Mike Vizard: Hey, Anand, thanks for being on the show and sharing your insights.

Anand Mahurkar: Thank you, Mike. Thank you for having me.

Mike Vizard: All right, and thank you all for spending some time with us. You can check out this podcast on Digital CXO, along with a host of others. We look forward to seeing you again next time.