“AI algorithms help automate the automation itself”
In an interview, Mihir Shukla, CEO and co-founder of California-based enterprise automation company Automation Anywhere, which is also planning a public listing, said AI algorithms are not only capable of automate mundane tasks, but also predict and suggest processes that need to be automated. Edited excerpts:
How is automation helping to fill the talent gap in the industry?
There is a large group of people doing (both) mundane and high-value work. By automating mundane work, you can increase the ability to perform, for example, customer-facing work.
Using our employees as human beings rather than robots and allowing them to add value at a higher level changes the equation for many large Indian companies. Some of them are among the biggest companies in India.
What happens to the talent doing the mundane work if it’s automated? Are they ready for other tasks?
Absoutely. I have seen this not only in India but all over the world. It wasn’t the people that were the problem. People were amazing everywhere and it was us who didn’t tap into the capacity they have to offer. We needed some work to do and we hired people to do it.
In all Indian companies and GICs (Global Internal Centers) I have visited, we have repeatedly said that when you deploy automation, people will be able to undertake more meaningful work.
Why is there a sudden talent shortage? Is it just because of the additional digital transformation?
My view is that the world has changed very significantly over the past 10 years. First, the working-age population has shrunk around the world, while demand is increasing every day.
Additionally, our businesses have grown from millions of interactions to trillions. Imagine the number of interactions you have with your bank on your mobile device. There were none five years ago. Where is the manpower to manage those trillions of transactions across so many systems and create the customer experience? What organizations are realizing is that we’re taking trillions of interactions and applying intelligent automation and bots to them to simplify them and create specific insights and actions, then push them to people to go further. This is the new operating model everywhere.
How is intelligent automation different from what we used to call robotic process automation (RPA)?
When RPA first started, it automated repetitive or mundane tasks. Intelligent automation takes the category to a different level. You are able to make decisions, including suggesting and guiding companies on processes to automate. By using different AI algorithms, it is able to automate the automation itself.
It used to be that we had to shop around and understand what processes people were doing to document and automate them. Now the AI automatically tells you what processes are going on and if you automate this process you can save 2,000 hours, while if you automate the other you can save 10,000 hours of mundane work.
How are we sure that the algos are able to make such decisions and do so with precision?
Let’s take an example of invoice processing, where you might have thousands of invoices. You may decide that below a certain amount you are comfortable with AI decision making, which is one way to draw a line.
However, even in invoices below this set amount, the AI algorithms are able to show you by learning that its confidence in a decision is say 30% after processing 100 invoices.
At some point, he will tell you that his confidence is 98%. At a certain level of trust, you can allow the computer to make decisions in a controlled and compliant manner.
At the beginning of the automation, we had the problem of not having our own data. Has this problem been alleviated now?
I would say he has made significant progress. Over the past eight years, for example, our product that processes invoice documents has seen millions of invoices and in 18 different languages, and learned from that experience. There is not a single human being on the planet with so much exposure and it is remarkable the precision he is able to achieve.
How far have bots evolved, given that the promise was that they will eventually replace human support operators?
He has improved considerably over the past three years. For example, our contact center solution, which uses Google DialogFlow technology, is quite sophisticated. If I tell you I’m married and somewhere later I say “she,” he can recognize that you’re talking about your wife. It wasn’t there until a few years ago.
The goal isn’t to make it look the same as a human, but to make sure it can take 80% of the stuff humans handle, so the 20% of the times you need people they’re available and you don’t. You don’t have to wait 20 minutes to get a call answered.
How is the IPO project going?
We focus on growth. Our cloud platform is the fastest growing today and the intelligent automation platform is doing extremely well. We are definitely heading in that direction.
Download the app to get 14 days of unlimited access to Mint Premium absolutely free!