Why enterprise-wide intelligent automation only works if it’s #BuiltOnAI
No matter where you are on your IA journey, consider cognitive solutions now
In the video below, AVP and Lead Product Manager Suresh Bharadwaj takes you through a few outcomes of #BuiltOnAI – informing value of spend, leveraging historical data to predict better outcomes and iterative transformation to future proof your business...
AI Informing value of spend
Working in the consumer goods industry, a lot of our clients are all about big brands and brand building. They spend a lot of money promoting their brands year after year. To the extent that they spend as much as 20 to 30 percent of their annual sales in promotions, trade promotions, brand promotions, and related activities.
The question has always been, do they get value for that money? Because they don't necessarily sell direct to consumers, they sell through channels some of which could be modern retail. Some could be smaller distributors in emerging markets.
The big question that has always been on their minds is, is that money well spent? Are we getting the value for that money that is spent? Are our brands seeing the uptake that we expect in the market? Are we gaining market share?
Those are questions they are constantly looking for answers to, and that's where we are looking at making investments. Those are areas that can be addressed differently from what we have done until now, with the advancement in technologies, including artificial intelligence, machine learning. That's where most of our investments are going today and where we're building solutions. Go to the #BuiltonAI home for more.
AI leveraging historical intelligence to predict better outcomes
That example is about assessing the effectiveness of trade promotion for example. When promotions are run the question always is, was that promotion successful? Would I want to run the same thing again? If something failed, why did it fail? There could be many causes and factors for a promotion to be either successful or to be a failure. When you actually go back and look at historical promotions, and see how many of them succeeded, and how many of them failed.
"All that we are trying to do is imagine there is a technology like AI and that technology can actually look at all of the data and make sense out of it."
There's probably something in all of that data that you can glean from and say, that these are the factors that actually made it work. Did I run the promotion too frequently? Did I run it for too long a time that people lost interest in buying my product? Or was there competitive activity in the market, that actually slowed down my promotion itself? Or did you know did my channels participate wholeheartedly in this promotion that I was running? Or did they buy all of this product, and this content then just did not pass it on to their customers? These are the types of questions that they want answers to.
These are expensive exercises. Looking at the historical performance of various promotions, the cause and factors. We're coming with models which will give them a good idea of whether the next promotion, at least what are the chances of the next promotion being successful? Obviously there are no guarantees, but at the same time at least give yourself a reasonable chance of success by looking at, at least not making the same mistakes that you did last time. All of the data that you've been sitting on was actually not used at all, and the latest artificial technologies help us leverage all of that intelligence. Bring it, encapsulate it in a business app, as we call it, and then be able to deliver that along with some domain expertise and consulting, that goes hand in hand with these applications. That's where we're making the investments. I think when you're talking to a business person, they don't really need to know too much about the underlying technology itself.
Explaining to them in very simple English, in a language that they understand, which is basically to tell them hey, we look at your historical performance of whatever function it might be. Whether it is sales, marketing, it could be logistics, whatever it might be. As you're running a business you're creating a lot of data, with or without your knowledge. As individual consumers we do that all the time, we leave a huge amount of data footprint and that's what companies do as well.
All that we are trying to do is imagine there is a technology like AI and that technology can actually look at all of the data and make sense out of it. To be able to see whether there is a pattern that it can recognize, that will be of use to predict how your business might turn out in the future doing a similar function. It's as simple as that.
I don't think you need to really get into more specifics on how the models are developed, as we call it. Having that level of conversation with business users is good enough. I think it's logical for them also okay, if I've been doing it as an individual I don't remember or recollect all of that, but if a system can do it of course I'm going to trust that. So that is one part.
Iterative transformation now to future proof your business
The way we look at it, we actually chunk it down into smaller business problems. Rather than taking anything that may be huge and therefore considered transformational. In today's world clients don't have an appetite for large transformational programs any more. Because it costs too much money, it takes too much time, it might run for three to five years. A lot of things change in that time frame today, people change, business processes change. What you started doing might not be relevant in three years from now – not only from a market or a client perspective, but even from our perspective we look at this as modern apps.
"If you can't build under 90 days it's not worth doing it. You build it in incremental steps, take it to your customer, try to run that proof of concept or a pilot. See if it works,"
For example, we have a trade promotion effectiveness app, which is what we call this app, the business app. I've got another app on supply chain visibility, where we try to predict the on time and in full delivery of orders that are actually made by customers.
Each one of these we actually take a 60-90 day time frame – nothing should go beyond 90 days. If you can't build under 90 days it's not worth doing it. That's the approach that we have taken. You build it in incremental steps, build something for 90 days, take it to your customer, try to run that proof of concept or a pilot. See if it works, take the feedback and then extend the program for another 90 days. That way you have a constant feedback mechanism from the clients, and you know what's working and what's not working. If something were to fail, it's better if it fails early than later.
One of the key ingredients to build anything on AI is having good quality data. In our own experience working in this space has been that our product has evolved from where it was acquiring a lot of data to the point now where we are saying, we've got all of this data what do we do with it? Therefore each of the business apps that we are building today is perfectly positioned to be built on AI, and that's exactly what we're doing today. That brings a level of trust that would have been otherwise missing from a customers' context.
The fact that they were with us in the journey to giving us the data, to the point where we are now able to leverage that data, and build AI applications on top of the data, is a complete package. That, I think, is where we are uniquely positioned and differentiated in the market.
For more insight and guidance from EdgeVerve, you can view a full on-demand presentation on how to embrace a holistic approach to automation in your enterprise, originally conducted as part of the AI Live online event series.