Applying AI Across Business and IT Operations – Kanti Prabha, SirionLabs

This is texturong TV. Hey guys. Thanks to the throw.

We're here with Conti prava who's co-founder for Syria. They are specialist in the realm of AI and it automation. We're going to talk about where we are on the maturity curve for AI these days Conti.

Welcome the show. It's a very interesting, you know line of you know conversation. Yeah, I have started to become real if into this, you know, 10 years ago.

There were limited use cases. That's so a real outcome coming out of AI. However today and I speak from the domain of contract management.

AI does so much that earlier was either not possible or just was extremely time taking mundane boring activities that you know, generally humans should not be doing so from my perspective AI forms that forms that technology layer that helps you do so much more than what you could have. Otherwise not achieved. Do you think we've moved from this point where there was initially a lot of skepticism?

A lot of people didn't believe in Ai, and now maybe we'd get into the point where people are going. You know, I just don't want to do this job without some help from some sort of machine learning algorithm somewhere because it's just getting too darn hard. Very much.

I think we are making that transition into folks. One understanding water. What is AI capable of and what is AI not capable of right second being able to see real life use cases of efficient uses of Automation and AI.

And third immediately understanding that if that particular part of you know, boring mundane activity was done by Tech. What time gets freed up for me to do other judgment you waste activities, which I'd rather enjoy right? So with people, you know going through that that three step Journey the faith that AI is very useful in, you know, certain types of activities leaving, you know good time for, you know, folks to do what they enjoy best.

I think has allowed to make that transition from AI is really a gimmick to is truly useful in corporate business processes. I think a lot of the challenges organizations are now wrestling with is they're not quite sure to what degree AI is going to be infused into their existing platforms versus whether they will require a whole new platform. So what's your sense on whether we need to kind of rip and replace what we have to get something that is quote unquote AI native or do we think over time?

We'll see more of the stuff just add it to our existing platforms. I think it's a combination of both there are independent AI technologies that can be used and there are then AI technologies that can be infused into your existing technical ecosystem as well. For example in contract management, we provide AI Technologies when customers are working on our platform.

But then we also provide AI as a bolt-on app to other ecosystem platforms that already exist within our customers so that the user experience does not need users to jump, you know to different solutions for Automation and Ai and they can enjoy the benefit of automation within the ecosystems and technologies that they're already comfortable with and and so depending on resources, you know budgets available one can take you know, one or both of these approaches. What is your sense of the level of patients required because one of the things I hear from folks all the time is, you know, they bring in something that is AI infused and then it doesn't immediately provide, you know, massive benefits and they lose patience, but it seems to me that maybe these AI platforms are really like a new employee and it just takes them a while to get up to speed and just like anything else they have to be trained. So are we not exhibiting their procreate attitudes and level of patience for these AI systems?

You've called it right Mike. I think AI takes time to learn there are systems that need to learn from scratch and then there are systems that complete trained to a particular level and it's only the Delta remaining which makes the system conducive to your particular organization that needs to be done. For but both of these require patients AI yet is not equal to Magic.

And therefore there is time that AIT takes to become useful. My recommendation is always to start with not just blank systems where you need to do the entire hard of training but take AI platforms become without the Box capabilities and then hone them to your needs and your internal requirements in that way you get started the wait time is not you know several years but gets reduced to weeks and months. But but the value outcome is similar and it takes away from the frustration of why is AI not working, you know, very quickly for me.

Are we getting better at training these systems using less data, it seems like initially required a massive amount of data. But as we go along I happen to wonder if you know, we can start training these models with less data. And as a result, we can update them and kind of move the whole process along a little faster.

Is that fair? Um, it is and not just volumes. I think the data Centric model of AI training allows you to keep making the data set better and better and more conducive to the outcome that you're looking for.

Therefore allowing this to become less of volumes game and therefore time to how do you intelligently train? How do you pick the right data set which works for the right outcome or the problem you're trying to solve and Technologies are you know, rapidly enhancing into human driven training is getting replaced my tech driven training. So the AI trains itself in many scenarios, make sure that it reaches the right level of quality thresholds.

And so therefore there is quite a lot of innovation The training space and data Centric approaches where the quality and precision of the type of data you use is helping AI produce results so much faster and much more accurate as well. What is your sense of the Fear and Loathing of AI these days the people still concern that their jobs are going to be replaced or do you think that people are starting to understand that perhaps my job can be more fun as I go along. It's one he has not.

You know. It's not to replace jobs. It's to you know, a person can do really boring data entry work while they're, you know, trained to be a lawyers and professionals who can negotiate really complex deals or they could be folks who need to manage.

Let's say a contract with all they're doing is reading the contract and manually extracting information one. Has the ability to completely transform roles that you know, folks are currently working in for the better with the use of AI and AI gives you that leg gives you that leg up where and that efficiency which takes away stuff which you've done once you've done twice. You know, how it's done.

You want Tech to take over and you want to focus on what's value adding. What's Revenue generating? What's what drives savings?

How can you reduce risk for your organization? There are so many higher order problems that humans are designed to solve which you know without AI takes away from the time you have to do that. So yeah is definitely I see it as a huge efficiency booster for humans who can then focus on, you know things that they will enjoy and that will matter so much more to their company.

And in terms of scale we should be able to do a lot more than we currently do because whether it's a customer service representative handling more calls or it's an IT person handling more complex. It environments. We shouldn't be able to get to something where the things that we're doing are not only more valuable, but we're doing him at a level of scale.

It's fairly unprecedented. Very much because you know the world is becoming so unpredictable. The climates changing regulations are you know, keep coming up around the world in different countries for different reasons the whole world of business.

Is you know so dependent on rapidly changing technology, we need, you know, folks in different roles to have time to react to these different, you know, golf balls being thrown at them and you need tech to take care of the day-to-day management. You need, you know, folks who are contract managers folks who are running businesses Finance professionals to be able to have the mental space and bandwidth to be able to do those, you know value judgment activities that are needed in this, you know Ever Changing environment today, so A scale is a big big part of you know, how much more can you do while all of these new elements keep hitting you in your day-to-day jobs? So absolutely, you know Tech plays a very valuable role there.

Now on the opposite end of that Spectrum, how do I know that the AI model is doing the right thing or is one wag one said, you know, it's one thing to be wrong. It's quite another thing to be wrong at scale. So how do we kind of make sure that we're not making the world's biggest mistake?

So I think it's it's a it's how you use AI it's you know, one can go all in and say oh I you know, we've got this AI let's just, you know go all in and throw a million documents at it and or we could take step by step, you know, two phase approach saying we're gonna do it for a bit. We're gonna see our results. We're gonna hone the model on the data make sure that we are very confident that we're getting 95% plus results and at that particular point in time use the model for much larger volumes.

This will allow for one to you know, fail faster, correct the mistakes, you know train more as needed make sure that who's using the right models to solve the right problems before you get into you know the high volume. We also know that bias is an issue and we really not talking about it in a social context. It's just that the people programming the model have certain assumptions that they make.

So how do you test for those assumptions to make sure they're valid in the first place even before we wind up deploying in a production environment? So that's that's a very important. Student at Syrian we've taken two pound approach AI is not just about tech and data scientists and you know folks who are you know, writing the bikes and core?

It is also about understanding the problem statement and having subject matter experts available alongside to understand. What is the direction? What is the problem being sort?

What is the output required and a combination of subject matter experts who understand the problem statements last tech experts who understand? Okay. This is my problem statement.

And this is what you want to achieve. This is a model that works on it. This combination is what makes one more successful as opposed to treating this just as a, you know, very complex rocket sciencey Tech problem.

I think we've done very well when we've taken you will approach to it. We have about 50 lawyers dedicated to Eating new models understanding new contractual needs and that in combination with our data Sciences machine learning team is able to provide results that are much more conducive to the outcomes that our customers are looking for. One of the things too that people seem to be struggling with this.

The AI models are built on their own Cadence by data science teams, and then they have to be inserted into some sort of application somewhere along the way and generally speaking those are built by devops teams that are upgrading the application, you know anywhere from twice a day to twice a month. But how do I bring these cultures together? Because a lot of times it's hard to align when the model is going to be ready as an artifact to be injected in the application and City have any best advice for folks.

Do I just throw all these people in a room and lock the door and hope for the best there? Is there some other way to think about see, I think now that AI you know is becoming Commonplace. It's something that you know, everybody's using on a day on day basis.

We've at Syrian made AI a part of the Sprint process, right? So you have requirements of All Sorts AI requirements are yet, another a new model is just a new requirement. We use the agile method and given our training times our you know within days and weeks we are able to produce new models every time we release every three weeks and does AI is not something where you know, one day you will see a result you're using the agile method to make sure that the results are produced.

They're tested they if they need to be enhanced. That's the next Sprint. So in a very short time period you start to see our results and it's not a black box anymore, which it used to and I think the the lawyers working alongside the technology teams.

In that agile scrum model driven by Sprints and short Sprints gives you you know, that comfort that this is going to be you know, one day something will come out of this and one has to make AI re you know much more available or to end users it cannot, you know remain that magical thing anymore. No, well speaking of magical things will be calling what we currently refer to as ai ai in the future or is it going to be like every other Innovation that came along and we called it AI until we got used to it. And then we just basically called it something else because you know, you're kind of treated as we understand it now.

So I guess I forgot what the exact phrase was that somebody says, you know every new innovation that you don't understand looks like magic. So when will it no longer be magic? You know, there is so much more companies around the world are innovating in so many different ways.

I think there is some shelf life for AI for sure, but I'm sure that you know, given the investment that is going in and the real world impact of AI already that we are seeing. Um, you know around the world in so many different problems AI will become you know something which is quite common and I'm sure something new will come up the human race has, you know been extremely feisty in such matters, but but AI in itself has so many applications and so many varieties and variations that there is a lot of work still to be done. I think we've kind of scratched this surface and Made it, you know a genuine valid technology lever, but it's just that now there is so much more to be done.

But you know who's seen the future. All right. So what's your best advice to folks as they contemplate all this?

What's that? One thing that you see folks doing over and over again? That just makes you shake your head and go I can't believe we're still doing it.

I think um, you know pulling off rules and calling it AI is something that I think the one needs to stop doing. Yeah, it has to be Ai and you can't just write a bunch of if then I'll statements and say oh, this is AI that gives a bad name to you know, everybody else is trying to do really I work out there. So I think that's the one thing and it is also counterproductive.

It is not provide results. It provides wrong results. It has the ability to you know, have false positives all the time their contradictory results.

So there's nothing coming out of rules. Anyway, so I think the one thing I would say is, you know, folks to understand the power of really I and put in more Revenue dollars. Sorry more, you know, spend dollars there.

Um, and the outcomes a fascinating with the user really I You know, the the ability to truly replace manual effort is very high and that's not the same when it comes to rules. It actually ends up increasing human effort at the end of the day plus a lot of frustration along with it because you're late in your deadlines. All right, folks.

You heard it here beware of AI washing Kathy. Thanks for being on the show. Thank you so much Mike.

It is a pleasure. All right back to you guys. Bye.