Interview: Samir Sinha

Tipping point

Print edition : June 23, 2017

Samir Sinha. Photo: By Special Arrangement

AILA, or Artificial Intelligence Lightweight Android, presses switches on a panel it recognises during a demonstration at the German Research Centre for Artificial Intelligence stand at the 2013 CeBIT technology trade fair in March 2013 in Hanover, Germany. Photo: CARSTEN KOALL/AFP

Interview with Samir Sinha, founder, Robonomics AI, Sydney, Australia.

As the Indian IT industry wades through an unprecedented maelstrom, triggered by a unique confluence of circumstances, it finds itself in an existential crisis that threatens to end its two-decades-long dream run. Samir Sinha, founder of Robonomics AI, a Sydney-based start-up, is uniquely positioned to provide a bird’s eye view of the impending “perfect storm” that threatens the hold of Indian companies in the global offshoring business. His experience in the industry (he has worked for three of the top four Indian IT companies in India and overseas in the last 15 years) enables him to situate the ongoing crisis in a historical context. More importantly, his recent venture into artificial intelligence (AI)-based technologies enables him to provide a ringside view of what may be in store for this industry.

In a five-hour interview to Frontline, he speaks about the Indian software services industry’s journey over the last two decades, how it matured, and why it is in jeopardy now. He also explains how and why new technologies, broadly classed under AI, are sounding the death knell of offshoring as we know it. Excerpts:

What are the key stages in the Indian software industry’s journey? And why is it now under stress?

The Indian IT industry grew out of the Y2K opportunity at the end of the last century. The prime mode of engagement with clients was through staff augmentation, which is basically a time-and-material-based commercial model. The client decided who did what. Indian companies would assign staff, depending on what skills the clients wanted and billed on a daily rate. But the client would still implement the project. There was no expectation whatsoever that the team would deliver the outcomes the client sought.

Large projects such as SAP implementation, which were driven by global consulting majors (like Ernst and Young, Pricewaterhouse (now PricewaterhouseCoopers) and Anderson Consulting) commanded a premium because they were large, multi-year projects worth billions of dollars. The Indian IT services industry saw, in the late 1990s, an opportunity to disrupt the existing model that was driven by the likes of IBM and the consulting majors.

Interestingly—and this is often lost in the narrative—the first few projects that Indian IT services companies did on a fixed-price model were for clients in India, large corporates, including many PSUs [public sector undertakings]. The overseas engagements of the Indian service companies were still mostly for staff augmentation and for support functions. We did not have the brand, the skill and the depth to pick up a large end-to-end implementation and be able to deliver it on our own steam. We still needed to be working for someone to be able to deliver those projects. But within India there was a different game altogether. Indian clients thought: ‘If these guys can be doing significant work with overseas clients, maybe I can access the phoren capabilities here.’ India thus became a training ground where companies could deliver end-to-end projects on a fixed-price basis. Because the projects were priced in rupees, the risk was also lower.

But there was another thing about Indian companies that was truly unique. Typically, an Indian company does not execute a project based on the contract it signs with its client. Once a contract is signed, the company goes overboard to deliver whatever the client wants, even if it means slightly lower profits.

The Indian companies’ emphasis on relationships was the biggest differentiator vis-a-vis the famed multinational consultancy companies. This is often missed out in most analyses of the Indian software industry’s evolution. Of course, price mattered, but price was usually what a client used to justify a decision they wanted to take anyway.

Companies like TCS realised early that they needed to move out of staff augmentation mode and go after fixed-price projects. In fixed-price projects, the client would define the project and its scope and float tenders to get vendors who had done similar work to implement the project. The decision on who would come in to actually work on the project would depend on the vendor. Indian companies venturing into fixed-price contracts signified a maturing of the capabilities.

Staff augmentation was still reigning in the 1990s, but the shift towards fixed price became more widespread in the late 1990s and early 2000s.

An important aspect of the fixed-price project business was that big multinational consultancies dominated the scene. The key to this lay in the manner in which they tightened the legal framework of the contracts. They actually ended up making more money overall by structuring the “change request” in such a manner that clients would have to pay for every little deviation from the original scope of the contract. It is not often the case that the client would have full visibility of how his business was going to evolve as the project was being implemented. The projects, which were implemented over several years and were worth millions of dollars, followed what is called the ‘waterfall’ methodology, in which the entire project was defined in stages—you get a clear sign-off from the client before you move to the next stage. What this implied was that the client had to have a very clear idea of the scope and design (business as well as technical) of the project, which would be enshrined in a contract. In most cases, over, say, a three-year period, the requirements of the client were bound to change, and this was where the change request from clients would be extremely profitable for the multinationals. This implied a ballooning of project costs for the clients. Of course, the IT service company also ran the risk of underestimating costs as well as technology risks in a fixed-price contract. In fact, companies that do not have a mature delivery system even now lose money when they do this.

But are costs not a significant factor?

Costs are not the sole determinant in this industry. The risks to a client are far greater than the dollar value of the projects they are implementing. Moreover, efficiency gains that arise from a transformational project are a significant multiple of the cost of the contract. What the client is actually looking for is a comfort level that the project would be implemented, and very often, he is willing to pay marginally more (among bidding contractors) for the level of comfort he is seeking. To the client, risk always comes first, cost is lower in his mind. Price comes significantly into play only when all the prospective contractors have a similar risk profile.





CULTURAL BAGGAGE



What was the key differentiator for Indian companies, if it was not price?

It was this relationship thing that Indian companies offered. The Indian worker would be technically quite competent, having moved across several projects. Multinationals like Accenture would operate differently: they would offshore lower-level work to their Indian branches. The offshoring would happen over a fence; the Indian entity would bid for that portion of the work within the overall company. The Indian companies had a more centralised approach. The cost for the Indian worker in overseas project was generally the Indian salary plus a mark-up. Companies like Accenture thus outpriced themselves.

But the more important reason why clients preferred Indian companies was that the Indian workforce did not have a sense of entitlement that was often the case with the native workforce in foreign markets. For instance, an IT job in Australia is not one of the most flashy ones. But an Indian IT consultant who is flown in is grateful for the opportunity to go on an on-site posting. An Indian IT consultant would never leave work at 5 p.m., and an Aussie would never stay beyond 5 p.m.! Our cultural upbringing also fostered this mindset. But this same cultural baggage was a hindrance when it came to telling the client: ‘Mister client, you are wrong, this cannot be done this way, there is a better way.’ While the humility and the service-minded attitude may be useful in a support function, the qualities of being able to be forthright with the client are vital when you are dealing with a transformational project.

Almost uniformly, the Indian IT industry grew on the back of a strategy that can be called Land and Expand. Indian companies overseas would start with a small staff augmentation engagement or a support engagement, and then win the client over with their work. Overlaying this engagement was the commercial construct in which the Indian company would treat the client as God. Indian companies treated this as the cost of establishing a long-term relationship.

How did Indian companies mature as a result of these engagements?

The delivery processes of Indian IT companies improved significantly. I will explain this with an example, that of TCS. One of its biggest clients in the early 2000s was GE, which had evolved the 6Sigma code, a set of techniques for improving process quality. TCS actually learnt these processes from its client. And then it applied these techniques and principles to its own delivery processes. It saw other clients doing other things, and it incorporated these too; all this resulted in a continuous maturing of delivery processes at TCS. It built its methodology on the back of multiple client engagements.

How did the incumbent multinationals who were sitting pretty react to all this?

The first reaction of companies like IBM, Accenture, Deloitte and KPMG was to raise the risk profile of projects that were being implemented by Indian companies. That worked initially, but clients started seeing through the selling strategy of the MNCs, especially after they gained first-hand experience from working with Indian companies. Clients also started breaking up billion-dollar contracts into smaller pieces, which reduced their risks. This played into the hands of the Indian IT companies. Indian companies started getting more complex projects to be executed on a fixed-price basis. There was also an expansion in scale; Y2K, for instance, was all about scale. As the methodology of implementation improved, profit margins on the fixed-price contracts also improved. The MNCs were too number- and contract- driven to be able to counter this.

How did offshoring come into the picture?

The offshoring story did not happen overnight. Initially, the clients were very sceptical. The commercial realisation that they could get the job done for a third of the cost forced them to consider offshoring. Offshoring by U.S. companies shot up in the early 2000s, but in Australia, a much smaller market, it came after a lag of about five years. The importance of fixed-price contracts, accompanied by the breaking up of contracts into smaller bits, became much more important after the global financial crisis of 2008. I suppose, as clients’ visibility of the future became less clear, following the extreme uncertainty caused by the crisis, fragmentation was a way of reducing risk in an uncertain world. I think Indian companies have managed to move up the value chain, but this has also increased costs of operation. But the catch is that the second-generation offshoring projects could not result in the same magnitude of increase in profits for the IT companies that the first wave did.

How are start-ups with new technologies like automation acting as disrupters?

A lot of the business for the big banks is from their big clients—what they call High Net Worth individuals. Typically, banks would allocate a financial planner to these millionaires, who would do planning for all the clients’ assets. Typically, this would cost the client about A$3,000-4,000 per year. So, a person earning, say A$10,000 a month, would not even consider this option because it would be too expensive. What the financial planner would do was to run an algorithm based on the client’s asset holdings, the targeted return on investment and other criteria, and tell the client what they need to do with their money. What some bright people have done is to code the algorithm in a software and publish it on the Internet. So, all that a person has to do is answer the questions at the portal and the software generates results suggesting what you ought to be doing with your money. The cost of using this software has reduced costs from A$4,000 to just A$300 per annum! You no longer have to be a millionaire to get advice. This is what is meant by a robot adviser.

Moreover, the algorithm can be tweaked further, depending on how reality corresponds to expectations, all of which is currently done only intuitively by humans. This is what happens with machine learning, which improves the accuracy of the solution. The machine learns from each transaction that it handles, improving its accuracy over time.

Now, what do the big banks do with all the financial planners they have? Their legacy systems are not geared to learn the algorithms that are now possible. They are married to monolithic systems that are difficult to change because they were designed for scale. There is the Singapore-based DBS Bank, with has operations in India, that operates as a mobile-only bank. A client transacts with the bank using a chatbot, which simulates a conversation with users in everyday language.

The model that the financial technology companies (fintechs) follow—which is what we also follow—is that unlike the waterfall methodology, they do not force the client to think of all the business scenarios upfront before signing off. With what is called the “agile” methodology, we only need to provide a small proof-of-concept to the client, which will be available in something like three weeks instead of three years. If the client likes the solution, we move ahead; if we fail, we fail fast, junk that idea and try something new. At the core of the model is the understanding that you can change the requirements as you go along. Facebook is a good example of what I am talking about. Facebook did not build the product you are using today out of one waterfall methodology. Instead, it builds a set of requirements every two weeks and then pushes out an update to its users. They build little layers of functionality on their product, depending on what users want or are using more. They are thus able to do a product-market fit at every stage of the development cycle every two weeks maybe. This is not possible with monolithic systems. This is how start-ups are disrupting the market.

I get the point that clients possibly find the business model attractive in an environment of greater economic uncertainty, but what technologies have made this possible?

The availability of cloud technologies brought down barriers to entry and allowed for unlimited expansion in terms of scale without having to sink capital initially. Moreover, the availability of open source technologies is having a huge impact. In AI, for instance, there is the case of IBM Watson. IBM started this several years ago; it used to play chess. IBM wanted to recoup the investment it had made in it, but the uptake of this massive solution was very slow because the price point at which they pitched it was outside the reach of most of the market. Today, the first 1,000 transactions with the supercomputer are free for a user. Beyond this threshold the user pays IBM 0.02 cents per transaction, a transaction being defined as a process request. But free software does not come with free support; that is where the companies make their money.

There is so much hype about robotics process automation. How does it work and why does it constitute a threat to the offshoring model?

Microsoft Excel has a feature called macros. A macro basically records a set of actions the user performs. Once it has recorded this, it can keep doing this even after you have gone away; the set of steps remains the same even if you change the data on which it does the operations. Now, take that concept and build more functionality on it and harden it for use in an enterprise environment. That is robotics. Whatever a human can do in terms of interacting with a computer, a robot can do it repetitively over multiple iterations. Anything that is repetitive, routine and grunt work can be recorded into a piece of software that is a robot. If you take that same concept to a computer application, as long as the decision-making in the process is definable and unambiguous—it does not matter if it is complex—it can very easily be executed by a software.

The cost of an IT consultant in Australia or the U.S. may be about $100,000 per annum. The cost of doing it offshore would be about $30,000. The cost of a robot doing the same thing is $3,000. But a human being working for $30,000 would need to be away from work for two shifts a day; a robot would work for three shifts and year-long without a break. Training humans takes months; training a robot takes a few minutes! Scaling a robot up or down is easy, but it is painful in the case of human beings.

Currently, to what extent is robotics a potential replacement to the offshoring model?

The steps that can be automated are generally the steps that are offshored. This is very profound for the Indian IT industry because it makes its money from work done offshore, not on site. In that sense, this technology is hitting where it hurts. If you look at it from a process perspective, there are certain kinds of decisions that are fuzzy. This kind of software is not ready for that kind of decision-making. A robot cannot work on the basis of gut feeling! Decisions that require a deeper understanding of the client’s domain are more difficult to automate. That is where you would need senior people or subject matter experts to drive, predict, control and manage things.

There is a lot of buzz about automation, but what is the state of the business?

The market leader in the robotic process automation space is Silicon Valley-based Automation Anywhere, founded by Mihir Shukla. They have been working on the assumption that they were competing with the Indian IT companies by cannibalising their business. But about three years ago they ended up breaking into companies like Wipro and Cognizant. The approach that Indian companies took was to look at this as a replacement for their fixed-price contracts. Indian companies had the option of using automation to cut costs, even if they did not pass on the savings to their clients. That using robotics could generate a richer data trail meant that the IT companies could use it to optimise and improve the use of robotics progressively.

But clients would have got wind of this. Surely it is not possible for the IT companies to fool them…

Yes and no. We are still in the early stages where not everyone knows everything. Typically, an incumbent IT company would never talk automation to their clients because it would cannibalise its business. Clients would only learn about automation when they are issuing an RFP [request for proposal] or a tender because it is only the non-incumbents who would bid. The incumbent would try to protect its turf, which is based on people, not automation. The growth of the IT industry happened by companies picking up students from colleges and training them. That is how they kept costs low and fostered a sense of loyalty and built their capability. It is at the lower end where the automation story is hitting the hardest. The foundation of the Indian IT industry, which is based on a people-centric approach, is being shaken. It is at the base where it hits companies’ margins the most. I think it is reaching the tipping point.

So, it is hitting at the large base of the pyramid that constitutes the Indian IT workforce. But anecdotal evidence seems to suggest that there are large-scale sackings at the middle level.

In terms of how companies would be reacting, it is useful to look at the companies’ cost structure. Wipro had a concept called Bulge, which looked at the ‘fat’ in the middle management level. This section would be paid much more than those at the lower end of the pyramid. Thus, getting rid of one mid-tier manager would be equivalent to two or three persons at the lower level. It would be perfectly feasible for a company to push up a college graduate to a more senior position at a lower pay compared with that of the person who has been sacked. By doing this, the company is lowering its overall costs. Ironically, it is quite possible that there are some who are getting promotions even as many are losing jobs. The sackings are a reflection of how the markets are reacting. And, they will cascade.

So, is there a risk of the IT companies fading away?

In fact, all the IT majors have seen this coming. And, they are preparing for it. I see the sackings as part of a reaction plan. These industry trends have not happened overnight; they have been around for a couple of years at least. What we are seeing are the symptoms of an underlying structural shift that is being driven by market needs and technologies that are available.

In response to this, almost every IT company has an automation strategy. I think, very soon, just like these companies were evaluated on the fixed-price versus the time and material metrics years ago, they may soon be evaluated by the market in terms of how much automation they are doing. Those who move first may enjoy some initial advantage, but sooner than later it will become the norm.

But would it not result in a bloodbath if the incumbents were to move towards automation in a big way? Would it not cannibalise their revenues?

I have a feeling of deja vu. Offshoring was a disruptor 10 to 15 years ago. Automation is the next. But I think each player will use automation differently. The size of the Indian IT market is about $150 billion. The bottom of the pyramid will be about 40-50 per cent of this—that is, about $60-75 billion. But the number of people at risk will be even more because people at the bottom are paid less. This would mean about 75 per cent of the people will be under threat.

How does your model work?

As a start-up I don’t have unlimited resources. Nor do I have the capability to cover all the clients. I have signed up partnerships with several large companies. What I tell them is that they are good at the pricing game, but what I can do is to add a thin layer of automation for a robust and scalable outsourcing strategy. I believe the business at this stage is still going to be people-centric, not automation-heavy. But automation is going to be the thin end of the wedge that is going to prise open markets for companies to get their foot in the door.

So, there will be pressure on the margins as well as volumes of these IT companies…

Volumes will actually go down first. Because it is still the first generation in automation of this kind, companies may still be able to hold their margins. But there is a lot of smoke and mirrors around automation. The ones who have a reputation to protect are playing it cautiously. They are saying they would only try automation with their existing clients and would only try them with others if they work. For the IT companies it is a Catch-22 situation. For a company employing two or three lakh people, it is very difficult to tell its workforce: ‘We are changing course from a people-based strategy to an automation-based strategy.’

How are the IT companies gearing to face this?

The companies have different approaches. TCS wants to build its automation platforms in-house. Infosys has acquired a company called EdgeVerve—well known for its product AssistEdge—and appears to be taking a more holistic approach under its CEO Vishal Sikka. Wipro was an early mover in automation through Wipro HOLMES, an AI platform. It also has an incubation centre working from Silicon Valley. But we need to understand that the ongoing lay-offs are not because automation has already made major inroads. This round is merely because of the fear of lower revenues and margins. I would not say that automation has already had a large-scale impact. It may have only percolated to the BPO space now; outside of it, there is only talk about it. My own reading is that the train wreck has not quite happened yet.

A letter from the Editor


Dear reader,

The COVID-19-induced lockdown and the absolute necessity for human beings to maintain a physical distance from one another in order to contain the pandemic has changed our lives in unimaginable ways. The print medium all over the world is no exception.

As the distribution of printed copies is unlikely to resume any time soon, Frontline will come to you only through the digital platform until the return of normality. The resources needed to keep up the good work that Frontline has been doing for the past 35 years and more are immense. It is a long journey indeed. Readers who have been part of this journey are our source of strength.

Subscribing to the online edition, I am confident, will make it mutually beneficial.

Sincerely,

R. Vijaya Sankar

Editor, Frontline

Support Quality Journalism
This article is closed for comments.
Please Email the Editor
×