Adopting RPA reduces cycle time to deliver client services for AT&T

Robotics would instantly conjure up an image of mechanical humanoids prancing alongside human beings, doing tasks beyond human capabilities. The truth of robots in an enterprise setting is, however, somewhat different.

Software robots or robotic process automation (RPA) helps automate several rule-based tasks and business processes that frees up precious capital, time, and manpower for more critical functions in network management.

Implementing software robots still relies on significant upfront human effort to identify processes, create business roles, and test software. The upside, however, is compelling.

At AT&T, software robots have taken over several routine and customer-facing tasks such as locating and tracking a complaint ticket, or for generating invoices. We have over 1,000 such software robots deployed throughout the organization.

The experiment began in last five years when we started exploring how humans and robots can work together. In 2016, we launched an online training program to teach employees how to build bots.

We then started experimenting with artificial intelligence, integrating it with software robots, to make them even more efficient and useful. The bots that scan phone calls to AT&T’s customer service division and compiles network traffic reports, has been particularly useful over the past few years.

Tending to the basics
We have been on a transformational journey for last 5-10 years. The first step was a strict application rationalisation program that dropped the number of applications we work with by 60%. While we were doing that, we also put in place a digital transformation strategy that was holistic, starting from front end and then going deeper into the technology infrastructure layer.

Skilling and re-skilling are important spokes of the wheel for AT&T here. Skills transformation and culture are after all critical to any digital transformation journey. We invest about $250 million annually in re-skilling. Last year 1.8 million hours of training was taken up by our workforce in areas spanning from big data, to agile networks, software delivery lifecycles, API platforms, and so on.

This was also an opportunity to liberate the knowledge-base of our people in front-end and taking everyone along the transformational journey. We trained our resources in the front-end team to build bots independently on the bot’s platform. The results were instant. We could immediately cut the cycle time of developing solutions, with results visible in days, or even in real-time.

It’s a tiered approach we follow in re-skilling and reusing the talent base. If a resource has done very good bots, we train them to automate business processes using bots. If they do well there, we train them to be process engineers, who then go on to redesign the base processes itself. We have also been thoughtful about how far we take the application of bots. Once a bot is mature enough, we think of how to integrate it into the base system and free up resources to build other bots for other functions.

Borderless innovation
One of the other key tenets of our transformation journey is our focus on open source. Because historically, we were getting black boxed by network equipment. We didn’t know what was going wrong or what was happening inside. We wanted to turn it into a white-box strategy to liberate the network functions trapped in the black box. So, we moved from a hardware-defined world of network management to a more software-centric world. We put our most impressive IPs in open source to align our global software intellect.

The skills transformation internally and engagement with the open source was a big pivot for us. Apart from re-skilling, we also changed our processes and policies to open this avenue of co-creation.

The focussed approach in re-skilling employees has helped us create an aggressive platform-based roadmap. With the right talent around, we felt it was time to put in place an open source API platform on top of the existing systems to consolidate functions. API platform allowed us to be more rational about the number of interfaces we have, dramatically reducing it, increasing re-use, and improving cycle time for delivery of services to our clients. Our API platform today does over 390 billion transactions a year.

Adopting a platform approach dramatically reduces the cycle time for simple tasks, that would otherwise require its own interface, localised tech skillsets, more human power, and it will still not be free of human errors. In case of a network overhaul, the disruption will be negligent, since the hardware is using the same customisable software layer.

The next bet
Our next threshold was middleware. Our API, data, Artificial Intelligence, Machine Learning platforms are the middleware that help us transform legacy systems in telecom networks. It allows for a federated development, in the sense that the platform itself offers you tool kits to quickly build solutions independently. It allows us to optimise human resources, train them easily, and they don’t even need to be expert developers, yet still create software.

Applying the AI and ML layers to software bots has helps us make our solutions more effective in achieving highly complex results that are multidimensional and still deliver an agile framework in a way which is cost-effective. Our outage team, for instance, receives real-time information, manages and dispatches tickets, and provides call information for customers.

A common pitfall of bots and enterprises is that it often becomes an IT project rather than a business project. At AT&T, we prefer leveraging our existing workforce as knowledge to implement software bots with a vision to eventually use to overhaul end-end processes, from front-end to the back-end.

Authored by Sorabh Saxena, President — Global Operations & Services, AT&T Business