Purposeful AI™ – humans and technology working together

The Chief Digital Officer of a prominent retailer was on the phone with me catching up after his visit to our labs along with his business leaders. He sounded excited. “We are ready to dip our toes into this AI stuff,” he said. He had two big business problems to solve – How to dramatically automate his customer service operations and make it more customer centric? And then, how to make his customer targeting more precise and effective? Now these were both areas with lots of historical data, clearly identifiable hypotheses and sponsorship from business leaders. This has not been the only time I have received such a request; he is not alone in his desire to take advantage of Artificial Intelligence (AI) on behalf of his enterprise to drive purposeful outcomes.

AI is here in the enterprise

Another innovation leader from a large CPG company is making machine learning a strong basis for executing on their company strategy. The first few problems were identified using a design thinking workshop – one of them included the ability to better predict the outcomes of clinical testing for new skincare product launches. A large industrial company with fairly complex labor contracts is working with Infosys leveraging natural language programming to digitize their labor contracts and create a much more user-friendly and accurate HR and employee experience in complex circumstances. A large energy major is leveraging machine learning to predict fraud in their purchasing processes which are significant and material for them. A global fashion major is working with a few of my more inspired team members to pinpoint and predict the specific factory, bill of materials and the cost that will be involved to build a new piece of apparel, compressing or crashing traditional concept-to-shelf apparel cycle times. We at Infosys are in the middle of at least 60 exciting machine learning projects with the top enterprises around the world, across their business and IT landscapes, attempting to solve the toughest challenges they have. AI has significantly dominated the conversation in recent times, but now it is becoming a mainstream strategy with several early projects in practically all enterprises we work with.

AI is Hard

So did the prominent retailer or the CPG company or the energy giant solve their problems with AI? Yes, and no. Some problems have clearly been easy to identify and solve. Others have required more sophisticated approaches. For example, in the instance where we are crashing apparel industry cycle times the treatment needs a lot more contextualization – we need machine learning experts steeped in the company’s unique business model, unique problems and unique data/KPI reporting structures for the effort to be truly effective. Finding the right data from the right data sources has sometimes taken weeks instead of days. Most large global organizations also have “data empires” guarded fiercely by different functional owners, all passionately trying to drive great optimization within their silos. Having the right clear thinking business leaders working with machine learning engineers asking the right questions, driving the right hypothesis is always tough. The accuracy of any machine learning algorithm even after some work is usually 60-70% in the first pass – this has to be followed by fine tuning and further work. But most of our clients are getting there. They realize that this is an important and powerful tool and initial problems will likely be tougher to find and solve. Like for anything new in the enterprise they often face teething problems when introducing a new AI system, but these aren’t anything that can’t be solved with the steadying influence of a seasoned partner.

AI is easy

The good news is that almost all exercises that we have experienced have been quick. For every use case, each iteration has been delivered under 12 weeks thanks to some great thinking by our Infosys Nia™ architects who can today ingest data from practically any source and drive quick intelligent patterns, knowledge ontologies, insights, decisions, and then automate the actions needed. So what this means is corporations don’t have to “wait and watch” until AI technology becomes mainstream, possibly after months and years. They can take advantage of AI now. The learnings, the early failures, the likely successes are all happening now – in a matter of weeks and days. This approach to AI taken by Infosys has been the posterchild illustration of failing fast and cheap.

AI is ultimately human

So where lies the magic? It’s not the technology alone. Yes, Infosys Nia makes it easier because it is a unified, flexible, modular platform with wide-ranging capabilities. But the use of such a platform can only be completely successful when it is combined with the right actions, and in the right environment, led by dynamic leadership. It’s the business leader finding that new problem to solve (hint: use design thinking), it’s about asking the right question. It’s about driving a culture that allows access to different types of data, driving a culture of cleaner data, and driving the right hypothesis. It’s about being patient with the outcomes and iterating till we make the prediction, the insight, truly useful. These human acts of patience, of constant experimentation, of curiosity and of creativity are perhaps needed in far greater measure with AI than with other projects. AI is ultimately more human than we have thought it to be.

What experiences are you seeking with AI? What are the use cases that have you concerned today? Come join the conversation here or simply reach out.

@SandeepDadlani

Infosys Nia™ – the artificial intelligence platform with knowledge at its core

If one were to believe the buzz it would appear that artificial intelligence (AI) has recently become that silver bullet most organizations have been looking for to achieve a competitive edge. For those of us who have followed and participated in technology trends over the years, it is clear that while it has now come to occupy attention that is front and center, the notion of AI and related techniques have been around for years. In my current role, I am asked fairly frequently about AI, related technologies and how best it can make a difference; often I am also asked whether one really needs a full-fledged platform when needing to leverage these technologies. So let me attempt to break it down in the context of what we are doing at Infosys.

Why a platform at all?

First, with 35 years’ experience across many industries and geographies, we know services get easily commoditized so carving out differentiation is critical. This is a step forward on behalf of all our clients. How so? If we can bring to the table a set of technologies in a coherent and cohesive manner that have been tried and tested to address the issues of the day for a client, that gives them a tremendous boost toward gaining that competitive edge they have been seeking. This is very different from the old approach of a one-off effort to cobble together relevant technologies and capabilities in an attempt to solve the problem.

Second, a smart next gen platform not only drives unprecedented efficiencies but also becomes an enabler for new areas. Let me explain with a couple of examples. For instance, in the realm of managing application services, we are familiar with the notion of L1, L2, L3 support. In our experience, there are many inefficiencies in how this is handled traditionally. As an issue goes through the 3 steps there is significant loss of context in each step which leads to inefficiencies. Moreover, often companies outsource one or more steps to third parties which compounds the problem further. Thus while the need for a proper handshake here is critical, there is unfortunately often some duplication of effort involved. A platform, with unified capabilities, that spans across this set of steps can eliminate redundancies and most inefficiencies. Another example that we commonly encounter is the inefficient processing of transactions in a, say, business process outsourcing type of scenario. At one client we found that a single individual had to go between 70 – yes, 70 – different screens to complete one transaction. Often there are manual steps involved as well that are error prone. Again, a platform that can enable end to end processes eliminating the redundant steps can alleviate such situations.

The Infosys platform – Infosys Nia

Data processing and analytics, automation, and AI are not capabilities unique to our platform. The uniqueness is in how we approach the matter of unifying these capabilities and bringing to bear for our clients the full force of the platform.

To begin with, the heart of our platform is knowledge. Yes, knowledge. One definition (Merriam-Webster) of knowledge states: “knowing something with familiarity gained through experience or association, or apprehending truth or fact through reasoning or cognition (where cognition is defined as: of, relating to, being, or involving conscious intellectual activity such as thinking, reasoning, or remembering)”. We have found that to deploy intelligent systems in an organization there is a need to harness and leverage all that is known about the processes, existing solutions, and thoughts about future growth. Since this knowledge in organizations is often scattered across systems and individuals we need a systemic way to capture it and then use it to create new solutions or renew existing ones. This is where we begin.

The challenge is the ability to represent knowledge in a structured manner. We use a comprehensive ontology model (OWL based), which makes the knowledge model readable by machines to perform correlation and inferences. Using a big data repository to store all the information generated from various fragmented tools in a knowledge base, we then apply machine learning algorithms, artificial intelligence techniques such as natural language processing and clustering to extract knowledge models to apply on various domains and industries. Knowledge models are constructed for various complex scenarios, such as business domain, software, machines, inventory, operational process, system events, design and development, etc. The models form the basis for correlation, providing structured and insightful ways for analysis, diagnostic and prognostic, and recommendations on operation, business, production, and product lifecycle management. The knowledge can then be consumed within various channels through chatbots, APIs and automation tools. As patterns within and across models emerge, it leads to learning that needs to be internalized by the organization. This makes it possible to automate smartly – whether it is simple robotic automation or automation that is more cognitive and predictive in nature. As this cycle continues, we find that new facts and recurring events come to light which provide additional opportunities in areas that are unprecedented. AI capabilities are employed in learning and updating/enhancing the knowledge in order to further improve efficiencies and build new solutions. This cycle of processing data for knowledge, automation, and learning with AI is not one that is of a stop-and-go nature. It must be ongoing and must have a pervasive reach into all aspects of knowledge gained, stored, and learned.

This is the primary focus of our AI Platform, Infosys Nia.

How is it put together?

Since our approach to the platform employs knowledge as a foundational pillar, our approach is one that seeks to combine the power of software with the ingenuity of people. To give this shape we rely on core techniques and technologies that include knowledge engineering, automation, machine learning, natural language processing, reasoning, user and data interaction, and learning.

Capabilities include OCR, Advanced Machine learning, Infrastructure Management, and Robotic Process Automation. The platform is also flexible and modular in nature. In other words, instead of a loose collection of capabilities, our clients can take advantage of the functionality as a well-integrated set of capabilities within a platform leading to better interoperability. Of course, the platform gives you the freedom to employ only those aspects that are applicable to your context.

Netting it out

The capabilities and technologies to address the challenges of our times exist. With the proliferation of data and the ability to compute and process it cost effectively, we can now have increased sample sizes. These have helped us get better at learning and employing smart algorithms. But true differentiated value with greater efficiency comes from employing the strength of a platform to address these needs. Infosys Nia, our next gen AI platform, is built with a conscious design principle that knowledge – deep understanding with context – is fundamental in solving problems. Working with it in tandem, it has state of the art data processing and data analytics capabilities and the tools necessary to smartly automate processes in the most relevant manner. It is important to note that deep understanding of context and applying systems in the most relevant manner are not mere platitudes. They represent how we get to intended outcomes. Infosys Nia is the platform that strives to deliver Purposeful AI™ to an enterprise.

Please join the conversation here and let’s work together for a more creative and purposeful future.

Infosys Nia™ – What’s beneath the covers? A conversation with Sudipto Dasgupta and Ganapathy Subramanian

A couple of weeks ago I shared with you what my first few days revealed to me about Infosys Nia™. I was fortunate to get time with two of my engineering colleagues, Sudipto Shankar Dasgupta and Ganapathy Subramanian, who have worked extensively on Infosys Nia. Sudipto is Vice President and Head of Engineering for Platforms at Infosys responsible for the development of Infosys Nia. Ganapathy is Vice President and Solutions Head for Platforms at Infosys responsible for solutions and customer implementations built on top of Infosys Nia. What materialized was a fascinating and informative conversation about the platform, the underlying technology, and the impact on clients. Here’s how the conversation unfolded:

Sudipto Shankar Dasgupta’s thoughts on Infosys Nia’s engineering scope

Me: What does it take to bring an artificial intelligence (AI)-based automation platform like Infosys Nia to life?
Sudipto: Well, Infosys Nia comprises of several critical components. At the core is knowledge and underlying it is a data-based platform powered by purposeful AI. To bring Infosys Nia to life, we seamlessly integrated automation, data and knowledge platforms. This was the first step – the amalgamation. The second step was to gain a strong understanding of our customers’ business problems that Infosys Nia had to handle. This is dynamic and will keep evolving, from one phase to another.

Me: What is unique about how AI technology has been integrated in Infosys Nia?
Sudipto: When we at Infosys talk about AI, it is always purposeful. We try to understand how AI innovations can be instantiated for business processes and be a differentiator for customers. We also consider the massive amounts of data customers generate together with the knowledge derived from it along with machine learning techniques. In short, AI is not considered in isolation, it is looked at comprehensively, as a way to handle knowledge in order to offer customers differentiated value.

Me: How do you work with data scientists for Infosys Nia?
Sudipto: For any organization, data scientists are critical resources because there are very few of them. Our opportunity was to understand data scientists’ roles and responsibilities and automate the tasks they would traditionally perform. This way, even if there are complex issues that data scientists would typically handle, Infosys Nia will be able to manage it. This will reduce the organization’s dependence on data scientists.

Me: How do you see Infosys Nia advancing in the future?
Sudipto: We are constantly asking questions that will help us keep a finger on the pulse of AI and focus on innovation – “How can software and hardware overlap?” “How can impact analysis or root-cause analysis be carried out better?” “How can traditional dashboards be improved with Augmented Reality (AR) and Virtual Reality (VR)?” We are also thinking about video and image processing advancements to uncover customer landscapes and understand their work. This is the future we are looking at.

Me: What does it take to maintain the momentum and keep your customers ahead in a continuously evolving AI technology landscape?
Sudipto: Yes, AI is a dynamic space with innovations taking place constantly. It’s driven by different ecosystem players like academia and open AI community, where developers and industry leaders contribute and share. Infosys ensures that we stay ahead of these innovations, keep customers hidden from the burden of the changes and help them reap the benefits rapidly. To achieve this, the AI platform should be able to seamlessly integrate primary components like automation and analytics with new technologies.

Ganapathy Subramanian’s views on customer Proofs of Concept (PoCs) as a way of fostering confidence in Infosys Nia

Me: What in your opinion is the benefit of an Infosys Nia PoC to a customer?
Ganapathy: With Infosys Nia we ensure that our clients get to experience, in their own context, the art of the possible with AI. This approach has been developed such that customers can solve business problems and see the value their data provides within short time periods. PoCs help us review what is, and is not, working for clients which is addressed by rapid engineering capabilities as needed. In essence, the ability to gauge outcomes quickly has been our inspiration for developing PoCs to test real-world situations.

Me: How do you help clients take the AI route with Infosys Nia-based PoCs?
Ganapathy: In the past three quarters, we have conducted over 20 proofs of value with respect to Infosys Nia which are now going into production. Across the board, Infosys Nia has offered good results. Take for instance, a PoC deployment we did for a European apparel and retail manufacturing company that wanted the capability to forecast design cost, a critical factor impacting a time-dependent final product – that season’s collection. In order to drive value, Infosys Nia based its recommendation on historic data across a wide category of apparels. Similarly, we have done transformative implementations for various industries and sectors like railroad and banking. It is important to note that most were done in just 6 to 8 weeks.

Me: What is the importance of doing PoCs, for both Infosys and clients?
Ganapathy: From an engineering point of view, Infosys Nia is more than a platform. We can leverage it to engage meaningfully and iterate fast with clients – which PoCs allow us to do. It helps engineers rapidly change the way they create a roadmap for the client. Next, from the client’s point of view – they just want to see value, fast. A PoC lets us offer this iteratively with real data. It also gives clients the opportunity to collaborate with engineers to define the roadmap the way they want it.

Me: How ready were customers to adopt Infosys Nia?
Ganapathy: For a while now, our customers have been hearing about AI and cognitive technologies like machine learning. Our customers are eager to capitalize on the potential that it brings. What’s more, our customers are now getting good at collecting, storing and leveraging data because they understand that is where true value lies. Combine this with their drive to be more efficient, improve business processes and stand out from the competition, and they are all geared to leverage AI.

Me: Can you share any key learnings arising from your PoC work with Infosys Nia?
Ganapathy: I’ll touch on three things. First, the space of machine learning is dependent on data. Though the platform is positioned to learn and amplify, a lot depends on incoming data. This helps the platform re-calibrate. Secondly, as I mentioned earlier, different clients are in different spaces related to AI. Understanding where the client is and what they have (in terms of data) are key to creating a plan that will align best with their needs. Thirdly, doing PoCs has helped us look at necessary and critical elements to be integrated into the platform in order to offer the best possible value to our clients.

It was clear as I walked away from my conversations with Sudipto and Ganapathy that Infosys Nia isn’t just cool technology. It is grounded in the foundational principle of being deliberate to provide businesses the differentiating value they seek in an ever-changing dynamic business landscape. It was abundantly clear that making available the option of PoCs is a strong way to reach out to clients so they can recognize value quickly and iterate fast to realize it completely.

As final questions to both Sudipto and Ganapathy, I did ask them why working on Infosys Nia was exciting to them. Here is what they had to say:

Sudipto: Knowing that the world is becoming more accepting of AI is truly exciting. In the coming years, we will witness automation taking on several roles that humans do. This will free up humans to try out other tasks. The world is moving more to AI, the cutting-edge, disruptive technology of the future. And since we are working on this with Infosys Nia, it is thrilling and exciting.

Ganapathy: Today, we are at an inflection point – technology-wise and research-wise. The underlying factors enabling the development of AI – like compute power – are not new but they have evolved by leaps and bounds recently. We can now do everything in this realm cheaper, faster and better. Add to this the fact that software is evolving thanks to the focus that academia, the open source community and top IT companies are placing on it. On top of this, improved data collection and analysis tools have matured significantly to allow companies to maximize data. All this makes for a very exciting time to be an AI engineer.

Our engineers are clearly excited. Businesses need to step forward to embrace this technology, and this approach of bringing Purposeful AI™ to the enterprise. I happen to know we will be discussing business use cases and Infosys Nia at Confluence 2017. If you are not going to be there join the conversation here or at: @PurposefulAI.

@puneetsuppal