Bhaskar Ghosh, Accenture’s chief strategy officer, handles all aspects of the company's strategy and investments including ventures and acquisitions and Accenture Research. In his new book "The Automation Advantage", Ghosh explains how automation works, the challenges therein, and how to seek return on investment from automation. More importantly, he explains why automation should not be treated as just a technology problem.
In a fireside chat, he talks about his book and also explains how intelligent automation, which involves the application of smart machines that leverage various artificial intelligence (AI) technologies, can also evolve their own capabilities to recognize problems and figure out how to solve them. Edited excerpts:
Estimates show that enterprises have automated only 15-20% of what can be automated. Though enterprises are experimenting with intelligent automation, why are they not deriving the full value?
When we talk about automation, we talk about driving the four aspects--cost, scale, quality, and productivity. But in today's context, the fundamental expectation is different. What we are expecting in the future from automation is to drive differentiated experiences for the users, better decision making, and most importantly to help grow business. Now, this is possible and this is driven fundamentally by some of the new technologies which have significantly developed over the last few years, such as cloud, data, artificial intelligence (AI), and machine learning (ML).
When we started talking to our different clients, to our surprise, we learned that they already understand the power of data and artificial intelligence. Most of them have done some prototypes that have also been successful. The challenge is that they are getting stuck when it comes to moving from prototype to scale at the enterprise level. Once you do that, then you will find that the percentage that you have referred to will significantly improve.
Let's talk about some common barriers to implementing automation as discussed in the book such as talent and skills shortage, cultural resistance in organizations, fears of job loss, outdated policies, unclear metrics, and no roadmap or strategic plan.
I think it is very important to understand the barriers and what stops the implementation even when the management is committed and wants to invest. Once you use an artificial intelligence-based system, the data acts as a backbone. One of the barriers comes from legacy systems -- a lot of the time the data sits in different systems and don't talk to each other. The other big barrier is the culture and the human aspect of it. A lot of time, people think that automation is a technology project, when it is actually a change management project. You can bring lots of technology, but people need to embrace that in the organization to be successful.
The third thing is the clarity in the scope. Any technology project has to align with the business goal because that whole paradigm has shifted and it is no longer just about cost-cutting. It is about transforming the business, better decision-making that will impact the business. So, one needs to understand the areas where investment is required and the right project that can get the quick result that is measurable. To overcome some of these barriers one needs to understand and take proactive steps, otherwise, you will find there is a passive resistance sometimes active resistance.
What are your thoughts on myths about automation such as not wanting to be the first or thinking of automation as a one-time project?
A lot of the time people think automation is a one-time project. Today's automation is driven by artificial intelligence where the system learns and becomes more intelligent the more you use it. It is a continuous project. You need to make sure your data is not biased and your system continues to behave ethically. These barriers are important to understand and implement. We at Accenture had an opportunity to implement this at scale within our organization. What we realized is that when you try to implement at scale, it is not just technology, you need to consider all aspects of the implementations, otherwise, it will never be successful.
In your book, you point out that intelligent automation is only as good as its data because of the problem of dirty data or bad data. What are the common mistakes that companies make when they want to implement automation but the data is not of good quality?
When you implement this type of system, one needs to capture the right data and in today's context right data is not only the structured data that is available in the firm. You are talking about a lot of unstructured data from social media and different other fields. Garbage in and garbage out was a very practical term in the past when the data source was all cleaned data. But now when we develop a system there will be garbage data because I'm taking it from social media and different feeds.
Two things are very important. Once you've captured this type of data, you need to be very clear with your business objective as it will help you focus on the type of data you need to pick up. The second part is that these systems learn from use as they are driven by machine learning, one needs to continuously monitor and fine-tune the system so no negative bias enters into the system. When we look at the various examples of automation we have implemented, we found that the companies are empowering their employees with data that is not humanly possible for them to keep track of when they are talking to a customer. This makes it different from earlier automation.
How should companies evaluate return on investment (ROI) from intelligent automation?
When you start an automation project, you need to make sure that the ROI is clear. It is not the implementation for the sake of it. Every organization has a different mechanism. Sometimes ROI is not measured strictly if it's an internal project. We need to put some checks and balances even in internal projects like what we do for our customers for the external project. So, the ROI is locked in the beginning and we try to deliver that ROI at the end of the project. We first try out a lot of projects internally.
For every application, we believe there should be a clearly-defined objective that will drive the outcome. In the past, it was just the cost. But now we are not talking about productivity improvement only. We are talking about growing the business. Since the whole paradigm of automation is different now, the business outcome is the first thing one should define before investing in automation projects.
How do you build an enterprise automation architecture? Are all aspects required from the beginning of the implementation or can they be added later (modular)?
If you move everything to the cloud, you get a lot of flexibility. What is important in architecture is modularity, because what is relevant today may not be relevant in five or 10 years. The question is that you should create the architecture in such a way so that you can adapt and add new functionality in the future. Your system should be modular, so that in the future when a better product is available, you should be able to take the old tool out and plug in the new tool, without changing the whole architecture.
You have also talked about the impact of automation on jobs and human-machine collaboration. Please tell a little bit about this process and how it can be achieved?
When we implement this type of system, some repetitive manual jobs will be eliminated. At the same time, new jobs and business models will be created. That is the story of automation for the last hundreds of years, starting from the Industrial Revolution. When you talk about Intelligent Automation with data and artificial intelligence, it is all about empowering people.
Things people cannot do humanly today can be done tomorrow with it. That is the principle of the book and that is what we are trying to achieve. Also, we need to back this up with the right training capabilities. So that the people who will get displaced, can get trained for the new type of work, which is getting generated. So, every time we are implementing this automation, parallelly we run a program called reskilling and deskilling people.
In the book, we mentioned 3Rs--relevance, resilience, and responsibility. The whole premise is that automation is relevant for the people and not for the sake of automation. It has to be resilient, so if the system breaks down your production is not hampered. The last thing is the responsibility of the organization. Once you implement this type of system, along with that make proper training and capability building initiatives so that the people learn the new technology, new type of job and stay relevant for their work in the future.