The machines check in at Indian BPO companies

Sunny Sen March 12, 2018 13 min

Story Highlights

  • Robot process automation, or RPA, where artificial intelligence and other automation tech is deployed in BPO operations, is beginning to mainstream in the Indian back office support industry.
  • If the task can be broken down into discrete parts, like in KYC verification, RPA tech can eliminate as much as 90% staff. Even in complex tasks, a 10% reduction is the norm.

At last month’s Nasscom India Leadership Forum, the biggest event of the Indian information technology and business processing outsourcing companies, Mohit Thukral bumps into Rohit Kapoor. Thukral is a senior vice-president at Genpact, India’s largest independent BPO company, responsible for its banking, financial services and insurance business. Kapoor is CEO of Gurugram-based EXL Service, a smaller peer.

The mood is collegial and friendly. When Thukral wishes Kapoor luck on EXL’s upcoming results, the CEO’s reply is sharp and witty. “You must be relieved… Genpact is a big company,” Kapoor says with a broad grin. Unsaid is that poor performance in certain quarters can stay hidden in large companies – a week earlier, on February 12, Genpact has posted a modest 6% jump in annual revenues to $2.74 billion for 2017 (it follows the calendar year in its financial results).

Rohit Kapoor, vice chairman and CEO, EXL Service
Rohit Kapoor, vice chairman and CEO, EXL Service. | Photo: Rajesh Subramanian

Five weeks later, EXL announces an 11.1% jump and $762.3 million in revenue. Kapoor can remain smug that EXL’s price-to-earnings ratio continues at nearly double that of for Genpact. Still, he is anything but complacent. “The business is changing as clients are moving towards better business outcomes,” he says.

At the cusp of this change are technologies like artificial intelligence, machine learning, robotics and data analytics, which in turn are driving higher amounts of automation with machines taking over a lot of the human work. The industry calls the trend ‘robot process automation’, or RPA, which came up as a concept in the early to mid Noughties and in recent years has become mainstream. It is only now that it has gained traction and yielding tangible results for clients in the Indian BPO sector.

As much as it promises massive benefits, RPA has always been looked at with suspicion – for the jobs it could eat up and for the difficulty in getting it to yield results. According to consultancy McKinsey, 1.2 million technology jobs will be eaten up because of automation. US-based HfS Research pegs seven lakh low skill job losses in the IT and BPO industry in India over automation and AI by 2022.

American computer scientist Andrew Ng gives a ring-side view of what the Indian BPO industry is living through: he says the demand for AI-based solutions is mainstreaming. “AI is shaping up corporate strategy. CEOs are wishing that they came up with AI strategies earlier, and the way you build businesses in the AI era is different,” the co-founder and co-chairman of online courses firm Coursera says during his session at the Nasscom event.

Kapoor has examples of RPA at his shop though he won’t name clients over confidentiality agreements. A large client bank, headquartered in the US, he says, had close to 1,000 people doing know-your-customer verification for every new customer. They would do background checks on a new customer against about 100 databases, half of them internal and the other half public. This would be a long – and fairly dreary – exercise. For example, a name alone can have different variations: initials versus initials expanded, last name first, first name first, misspellings and so on.

“We use AI and ML to do it with only 100 people. You still need people to prepare the case and present it. But the big work is done in an automated fashion,” says Kapoor.

The bank is happy as it reduces the operational expense of hiring 1,000 people. The work gets done faster because it is done my machines, the probability of error is lower, and the bank can be quicker in making decisions.

Paradigm Shift: Man to machine

Traditionally, clients in the US and Europe have outsourced back-office work to BPO firms in India that were part of the non-core business functions. The outsourcing services provider would take on work around accounts, claim settlements, customer data management, customer acquisition and support, among other tasks with the promise of doing them cheaper and faster. More than 50% of client engagements were of this nature – replicate client processes offshore but do it faster and cheaper.

The BPO units typically charged based on the number of hours of human labour put into getting the work done. And the work was pretty non-technical compared to what IT companies did.

That is changing. “Clients want to transform their business using domain expertise, and then digital and analytics expertise, and they want us to come and consult as experts, vis-a-vis in the past they use to come and ask us to run the process, improve the quality and lean the process,” says Thukral.

His colleague, Sanjay Srivastava, Genpact’s chief digital officer feels that back offices are swinging from an over-dependence on humans to digital entities. For Genpact, he says, more than 60% of the revenue is impacted by digital interventions, and about 20% of the revenue comes from completely automated processes. Genpact counts Mondelez and many among Fortune 500 companies as its clients.

Srivastava insists that the contact centre business is completely moving to AI-based chatbots. Unlike humans, bots have data of other queries and can preempt the next question that the person on the other side of the chat might ask.

The infusion of AI and related tech into the 25-years-old BPO industry is opening up opportunities for back-office businesses like never earlier, experts say. “The state of the BPO industry today is no different from the IT industry 15 years ago. We are just scratching the surface,” says Anil Doradla, chief financial officer of wireless optimization company Airgain, who tracked tech companies for 15 years in an earlier avatar as a Wall Street analyst.

The standards vary on how much work will be done by machines. In the past, BPO managers used to apply algorithms for high volume, simple repetitive tasks, but now new technologies can be applied to complex tasks as well. Kapoor says, “For KYC-like processes, can eliminate 90% of the people. But for other processes, it can eliminate 10-15% of people.”

BPO 3.0: A different game

The BPO in India of today has progressed through stages the last 20-25 years. The first phase was pushing non-core work to cheaper destinations over high-speed communication links. This was so successful that not only did domestic companies such as Tata Consultancy Services, Wipro, and Infosys dive into it to take strong positions, their international rivals IBM, Cap Gemini, Accenture, Dell and a host of others set up and scaled shop in India.

The second phase was using analytics and engage domain experts to solve problems for clients. BPO shops are now in their third phase of evolution leveraging AI, ML and related solutions.

Keshav Murugesh, CEO, WNS
Keshav Murugesh, CEO, WNS. | Photo: Rajesh Subramanian

While most people say that automation and AI will cut jobs, Keshav Murugesh, group CEO of WNS says that track record has shown new technologies have only led to the expansion of the market. This, he says, will result in people moving to higher level roles with machines taking over commoditized tasks – something that is already happening.

For WNS, the journey started seven to eight years ago. Clients were comfortable with India and Indians doing their jobs. “Most of the larger BPO companies like WNS, EXL and Genpact can come and run everything for large multinationals because of domain knowledge that they have built over years,” says Murugesh.

But, where would earnings growth come from? That goal would be achieved only by de-linking headcount and revenue. Revenue had to grow faster than costs and that was possible only by using technology.

For a large UK-based energy company that is a WNS client, there were a lot of emails coming in from customers. Some 600 people were looking at these emails and it was taking a long time to put them into different categories and responding to them. WNS started using natural language processing-based algorithms to sort emails into different buckets. Responses were drafted depending on the category of each email. Only the final responses were cleared by a human with the rest of the process entirely automated.

“It dropped time by 80% and reduced cost by 52,000 pounds. If we were using 600 humans to do this earlier, we needed only 200 people now,” says Murugesh. The remaining 400 people were moved to analytics side and helped the client in product development. And, yes, WNS increased revenue from the utilities company.

“We suggested a separate pricing for newly married people or a student, based on usage. As they (the company) went to smart meter technology, they partnered with us,” Murugesh says. It kept growing the account with WNS generating leads and managing customer relations for the client’s water supply division and also running roofing contracts.

For another client – an airline company – WNS created a platform for fare audit. The airline company was facing revenue leakages in its agent network. WNS asked the company to provide the data dump of tickets it had sold. Within a few hours using automation it provided the airline company with a reconciled view of what revenue is lost from which agent, and went on to issue the debit notes on behalf of the airline company to the agents. “What used to take six months now takes two hours,” says Murugesh. The icing on the cake: the airline is charged on a variable cost basis. “We take 10% of the revenue got from the reconciliation,” says the WNS CEO.

A lot of this is possible, Genpact’s Thukral explains by focusing on the client’s customer, which was almost impossible to do in the previous era. He says clients are not asking to reduce people, but to increase productivity. For a financial services company, he explains, Genpact first put the customer in the centre and layered technology, capabilities and processes around it so that the customer experience impact is not 10% but it is 30-40%. “So, we say, you don’t pay us by people, but pay by outcome,” Thukral says.

He sees the use of technologies like computational linguistics for neural chat programs, where bots run transactions instead of humans. That requires a change in hiring as well. Genpact, Thukral says has hired dozens of design engineers, data scientists, digital engineers. To be sure, humans still do a lot of the intervention and design handling.

Opening the floodgates

Systems used by BPO companies have become intelligent over time. With the increasing use of the internet and data-based services, clients were throwing up a lot of data but the problem was to make sense of the data, which was mostly unstructured. Handling it was a problem even with transferring it from one platform to another. That part was addressed by using IT tools to automate processes and have clients move to templates.

What was tougher was dealing with data that was not easily structured that computer programs could recognise and interpret. For example, a court order or a doctor’s prescription. In case of a medical record, the doctor may have summarised the problem, the symptoms of a disease, the action that was taken for the entire medication and treatment procedure – all is written down by hand or, often, recorded on a dictaphone. Someone needs to transcribe it and then look at the data to tell, say, how much of an insurance claim can be paid.

EXL prepared a machine learning algorithm to pick up specific fields. There is no one standard template, but the algorithm is smart enough to look for the doctor’s name, the patient’s name, the name of the disease, the code of the disease, the value of the treatment, and then input the data in an automated fashion. That allows EXL not only to process the claim or to make the summarisation chart, but it also makes the BPO company a partner for the insurance company on outcomes.

Mohit Thukral, Senior VP and Global Business Leader, Genpact
Mohit Thukral, Senior VP and Global Business Leader, Genpact. | Photo: Rajesh Subramanian

Now EXL doesn’t need a person to read 200 pages of medical reports, prescriptions and diagnosis. All that information that needs to be extracted and summarized in one page is done by the machine. “It used to take us 2-4 hours per person to read one document, summarize it and put it in one page. Now it takes a minute,” says Kapoor, listing such work among the “completely new business opportunities” before India-based BPO units.

The RPA shift, still, is slow and will likely be bloody. “Robotic process automation breathes new life into legacy systems. But the shift is not happening overnight. Once you have the trust of your clients, you can remove the bodies and replace them with automation,” says Phil Fersht, founder and CEO of HfS Research. “The service providers are shrinking in size – 7.5% of the jobs will go down.”

There are other areas, too, where AI is opening up opportunities like in vision analytics or image analytics. In the past, insurance companies would send a person do a property survey. The way it would do it is take the permission of the owner, visit the property, inspect it for damages, and provide a report on it. “Now we are able to send a drone, and we can identify if there is a damage caused by hail storm, rain or flooding, and also what type of damage it is, and all this using image analytics,” says Kapoor.

What’s in the future

For almost two decades, the BPO industry has been viewed as the stepchild of the IT industry – an industry where both the growth and margins are lower. The valuations of BPO stocks – Genpact is valued at $6.23 billion, WNS at $2.3 billion and EXL at $2.02 billion – could never match those of IT companies like TCS ($89.4 billion), Infosys ($38.8 billion), and Wipro ($20 billion). Neither could revenue.

That could change in future. BPO outfits are not only helping their clients to improve efficiency and productivity, but they are also sharing risk and building intellectual property that is scalable. If a bank has 5,000 small and medium businesses taking loans from it, it will have to access the risk of these loans from time to time. Earlier the work was done by a certified finance professional, who would go through the balance sheets and income statement and make a risk score. “Now we just ingest all the documents and use analytics and AI to interpret them and give a risk score. This is fully automated,” says Srivastava of Genpact.

Companies such as ] 7 that focus on customer acquisition and engagement have AI-based virtual customer assistants for self service for its clients’ customers. The company lists American Express, Citibank, BestBuy, Ebay, United Airways, Target, and others on its client list. Its voice automation solutions, it says, have reduced repeat customer calls by 30% and saved $80 million.

In the past, BPO operations had limited ability to take risk and make changes to processes. With an increase in confidence in their own capabilities – and that of their clients – they are much more open to make investments to make changes in technologies, applications and tools.

Higher levels of automation also open up the ability of BPO companies to take on more. EXL, for example, in the US has data for 250 million consumers – their credit history, their demographic, their past purchase history, their linkage with social media, and all forms of data. “We can marry this information with the client’s data and tell what all to sell,” says Kapoor.


Updated at 10:13 am on March 12, 2018  to give credits to pictures and graphics.

Disclosure: FactorDaily is owned by SourceCode Media, which counts Accel Partners, Blume Ventures and Vijay Shekhar Sharma among its investors. Accel Partners is an early investor in Flipkart. Vijay Shekhar Sharma is the founder of Paytm. None of FactorDaily’s investors have any influence on its reporting about India’s technology and startup ecosystem.