Maximizing ROI through AI-enabled Digital Transformation with Shobha Patil

Organizations often struggle to achieve high ROI from digital transformation due to traditional strategies falling short of dynamic market demands. This can result in inefficiencies, poor customer experiences, and missed growth opportunities. To address these issues, businesses are increasingly adopting AI. AI automates repetitive tasks, improves decision-making, and personalizes customer interactions, boosting operational efficiency and satisfaction. By integrating AI into their digital strategies, companies can unlock new revenue streams, optimize resources, and maintain a competitive edge in the digital era.

For this week’s CDO Insights series, we have with us Shobha Patil, a seasoned IT leader with two decades of expertise in Information Technology, Digital Transformation, and Consulting. With a proven track record in Banking, Finance, Manufacturing, Process Management, and Automation, Shobha excels in Practice Management, Large-Scale Transformations, Delivery, and Operations Management. Currently, she serves as the Chief Executive Officer at Sankey Solutions, where she continues to drive innovation and excellence in the IT sector.

For this interview, Shobha discusses the evolution of ROI discussions in the context of AI and digital transformation. She highlights how technology, particularly AI, plays a crucial role in maximizing ROI by automating tasks, enhancing decision-making, and personalizing customer interactions. Shobha emphasizes the importance of aligning AI projects with business goals, testing small-scale projects before full implementation, and ensuring data quality. She also discusses the need for human oversight in AI to maintain accuracy and ethical standards and the importance of continuously adapting to emerging AI techniques. She shares real-world examples of AI projects, such as fraud detection and customer insights, and underscores the significance of ongoing learning and adaptation to leverage AI effectively.

AIM Media House: You’ve been with the company since 2021 and have a diverse background in IT solutions. How has the conversation around the ROI of AI evolved over time? Has the vocabulary or focus shifted, given the industry’s ongoing goal to cut costs and improve efficiency? Could you share some insights on this?

“Technology has always been a driving enabler across horizontals.”

Shobha Patil: Technology has always been a driving enabler across horizontals. One of the important parameters that any organization works on is increasing revenue or decreasing cost. Technology made a significant impact in defining how the cost leverage can be given using automation and digital transformation. Now, we are talking about the AI age, which is again focused on maximizing return on investments. When we specifically talk about ROI, it’s about the technology costs that are put to leverage and how that makes it equal or reduces costs. It’s not just about reducing costs but maximizing ROI. The decision to buy versus build, the right tools and technologies, whether to use open source versus licensed ones, and the factors of managed services or annual maintenance costs—all of these have become important and fundamental in maximizing ROI. Evolving into AI-enabled digital transformations, a lot of these initiatives are also percolating in key areas, whether you have an improved CRM, faster content generation, reduced customer attrition, streamlined operations, predictive analytics, or process automation. All of these are key areas that influence how AI tools and technologies are leveraged to maximize ROI.

AIM Media House: How have various technologies always been enablers, and with AI being a buzzword, what is its impact? With Generative AI, implementation is more complex than traditional IT solutions. How do you leverage AI tools for maximizing ROI? Given the hype, how do you educate organizations on good AI use cases and realistic expectations?

“I do agree that AI is a buzzword, but as a technologist, I see the full potential that AI has.”

Shobha Patil: I do agree that AI is a buzzword, but as a technologist, I see the full potential that AI has. As a pragmatist, I would also say that we have to be strategically aligned in whatever we do. This applies to AI and Gen AI projects too; they need to be aligned with our overall business goals. That is the key parameter to judge whether we should undertake such projects or not. Secondly, it is important to understand the measurable ROI on these projects—whether we will be able to achieve it, the kind of technology infrastructure required, and whether we have the data foundation. Ensuring we have high-quality structured or unstructured data, the insights needed to build this engine, the right leverage of technology, a skilled workforce, and the investments to scale further are crucial. It is justified to do prototyping or pilot projects first to test them before going for a full-scale implementation. These are the key things anyone should keep in mind before venturing into AI. There is definitely a lot of potential in that area, but we should try, test, and then go wide on it.

AIM Media House: AI implementation requires many factors, including infrastructure. Focusing on data, many organizations aren’t ready for AI due to data readiness issues. What roles should domain experts play in curating and validating data, and how can organizations balance human and machine intelligence?

“The AI models are as good as the data they are based on.”

Shobha Patil: The AI models are as good as the data they are based on. I was reading an interesting article the other day, and it mentioned human empathy in AI. It is very important that while AI has progressed and you work with whatever rules and preset engine we build around it, humans are very much required to curate that data to see whether the data is relevant, accurate, comprehensive, and touching all the outliers, exceptions, and other things. It is very imperative to have that check, and I don’t think that is irreplaceable. Secondly, it is also important to curate the quality around that data to ensure that AI is able to give you valuable insights. Once you build that engine, it is very important to validate the AI-generated insights to ensure that it is as good as an expert giving an opinion or insight and that these are as equivalently actionable, not something random, misaligned with business goals, or not doable. Humans have the power to make a judgment, which AI cannot take. Humans will still need to intervene in the validation of those insights.

Third is how we make it more ethical and ensure that data is prevailing, secure, and has proper nuances on how the AI system’s guiding engine is based. It is very important that we define the objectives and roles of AI systems and what they need to align with. This human intervention will always keep them in check, monitor the performance of AI regularly, and ensure that it is in line with strategic goals and objectives within the ethical standards, ensuring we are not going off course. 

About striking a right balance we need to ensure proper collaboration; it’s not like we are introducing AI and replacing humans. Humans are well adapted to take judgment calls better than AI, and that has to be leveraged in proper strength. For example, AI’s strength is speed and scalability, while humans’ strength is judgment, expectations, and experience. We have to marry those benefits and strike the right balance.

Secondly, there is always a feedback loop of what is changing. There is continuous change in behavior, whether it’s customer trends, clothing trends, or fashion trends, and the data and insights will change accordingly. Who will keep control of this behavioral pattern? It has to be looked at, and the right time to loop in those feedbacks has to be a human judgment. Regular review and refinement of AI inputs are required. All these models must have transparency and understandable outputs. They cannot be something humans cannot comprehend or make decisions on. Human intervention will always monitor whether the model is understandable. It is key to ensuring that it does not give any vague answers and is something relatable. We need to collaborate on the human and AI aspects, working with their respective strengths.

AIM Media House: Given the wide array of AI use cases across various industries, the answer to calculating ROI isn’t straightforward. When clients ask how much implementing AI will cost and how much money it will make, what framework do you use to calculate and demonstrate the ROI and real impact? How does Sankey Solutions approach this?

“We would never advise our customers to use a big bang approach and make heavy investments in AI at an early stage.”

Shobha Patil: So I think the golden principle is always that you test the waters before you enter them. We would never advise our customers to use a big bang approach and make heavy investments in AI at an early stage. While we understand the merits of AI, we also need to test it pertaining to the value it could create for their use cases and ecosystem. We first ask them to identify quick wins and scope out smaller projects that can showcase immediate value. With a cost-benefit analysis, we are able to demonstrate the value generated with a certain volume of work and project what could be achieved at a larger scale. A phased implementation and quick wins help generate interest and determine if AI will deliver the desired ROI and meet short-term goals.

If the smaller projects don’t work, there’s no point in continuing. These small, three-to-four-week projects help generate and showcase immediate value before scaling. Scaling involves workshops aligned with strategic objectives, such as cost reduction or operational efficiency, to set realistic, achievable targets. This approach enables a transparent and judicious decision-making process, making it easier to convince management of the project’s value.

The next step involves fine-tuning the models and projects to reach the target outcomes. For example, if our target accuracy rate was 95% but we achieved only 80%, we need to analyze the data, insights, outliers and finer points required to enhance the model and give the right output. Fine tuning and continuous adaptation is the next step which comes into play once we are able to scale. We work with our customers through these three-pronged steps at Sankey too. 

AIM Media House: Can you provide real-world examples where you identified short-term gains and long-term benefits in AI projects? How did you navigate these conversations with your clients to ensure successful engagement? Also, how did you manage to drive the engagement and ensure continuous adaptation after scaling?

“We saw how the calculation of ROI has evolved, and with the evolving nature of AI itself, especially with generative AI, the ROI calculation framework needs to evolve.”

Shobha Patil: For shorter-term value creation, a simple thing would be like a chatbot or a search engine that gets you the data faster than anything, which improves your response time. So, you could implement it as a smaller project within the current ecosystem with minimal dependencies, allowing for easy testing. With that, all the immediate parameters that you could look for in value addition could include reduction of workload and improved response times, which could be the key factors you could look to calculate your immediate short-term gains.

For longer-term expansion, the goal would be to build it into a more intelligent engine capable of using more models-driven decision-making processes and further enhancing it. To achieve that, it will have to work on a set of data, past data, and even futuristic information which will take some time. We saw how the calculation of ROI has evolved, and with the evolving nature of AI itself, especially with generative AI, the ROI calculation framework needs to evolve.

For instance, for a financial firm, we also did one fraud detection engine, which was a short-term win. From a longer term perspective we were able to expand that fraud detection to an even bigger platform, which was expanded to do predictive analytics of different customer insights and how to use that to enhance even before the fraud happens. Just by looking at the patterns, we were able to predict that something was going to happen. So that is a longer-term goal.

AIM Media House: Any closing thoughts on how the entire space of Gen AI is going to shape up and with the evolving nature of Gen AI how does your framework of ROI evolve? How is your model and framework going to evolve with this?

“We need to continuously keep ourselves aware of upcoming developments and adopt them.”

Shobha Patil: With large language models, there are multiple techniques coming in. First of all, we have to be tuned with it and understand how the nuances of different models are going to work for which kind of use cases. As we are talking, there are many learnings and immediate tools. The wrappers are getting built on top of these things and can be easily used. Whether to build versus buy and what to leverage is very, very key. For example, large language models have a large impact in terms of making enhanced decision-making, improving customer service, and content creation.

When there is a focus on customer interaction, how the potential of content generation can be automated versus multimodal learning, which is more with respect to integrating insights from text, images, and audio, can be leveraged into many other applications such as healthcare or manufacturing, or automotive for that matter. From the frontiers of emerging AI capabilities, we need to be abreast of evolving techniques and focus on investing in keeping ourselves up to date on these evolving techniques. At Sankeys we have an ethos of continuously learning and relearn; it has become part of our DNA. We need to continuously keep ourselves aware of upcoming developments and adopt them. Implementing innovations on a small scale to see what really makes sense, applying them to real-world problem statements, and checking whether they really give the kind of value to the businesses or not. If they do provide that value, then we should go ahead and implement them deeply within the industry paradigms and further develop those skills. This way, we can address real-world problems with the right talent and workforce to implement them.