Sunil Senan is a business/IT leader with 24 years of experience. Over the years, they have worked closely with global clients in their digital and data/analytics transformation journeys across industries - Financial services, Retail/CPG, Healthcare/Insurance, Manufacturing etc. He is the Senior Vice President and Business Head, Data and Analytics at Infosys. 

 This interview features Mr. Sunil Senan’s insights on the role of data governance, regulations privacy and many more in ensuring successful AI practices. 

What is Infosys Data and AI Radar 2022? How can it aid organisations across the country?

Infosys Knowledge Institute’s Data+AI Radar 2022 studies how businesses apply artificial intelligence to their data. Our analysis identifies three key actions that will lead to better results:   

1. Share data extensively.  
2. Build explainability into advanced forms of AI  
3. Build a multidisciplinary AI team and keep it focused on business outcomes.  

Our survey findings provide details of functions and use cases that have been satisfied with the application of AI in various industries and those that have not. For a firm looking to invest in AI, this can provide a good pointer of use cases about others in the same industry getting value from AI and those that may be riskier.

How can organizations decide which AI approach will aid the smooth functioning of their company?

Companies must first assess what AI capabilities they are achieving and then gauge the utilization and satisfaction with the specific business use they are pursuing. 

A good starting point is to understand that almost any process in an organization has scope to become smarter. Now the trick is to prioritize those that are most feasible, least risky, and provide the highest bang for the buck. The study helps provide a view into how others have approached this and the results at a broad level. 

How can organizations encourage data sharing by ensuring data privacy?  

Most of the data that an organization utilizes to perform analytics are first-party data, but more is needed for organizations today. To understand the market, better access to other organizations' first-party data has become essential. Data could be sensitive and personal, so organizations need to take all the appropriate precautionary measures to protect such data, as any kind of breach could attract penalties and fines and could impact the brand image/goodwill of the company.  

But with new privacy technologies, concepts, and best practices, supported by a strong data policy framework to manage data across the data lifecycle from the collection, storage, processing, and consumption to purging/ deletion, it is not a difficult task to share data without impacting overall privacy compliance level.  

"Differential privacy", "Data Clean Room," "Automated data transfer approvals," "Secure Analytics", and "Tunnel Encryption" are a few of the privacy-preserving methods/ideas/technologies that have gained wide industry acceptance for sharing data without compromising levels of privacy compliance.  

Some of the examples are –  

1. An organization wanted to launch radio marketing campaigns but didn't have enough information on which radio station its targeted customer listened to. To understand that, they used clean data rooms to collaborate with another organization that could provide the required data/insights. The data was shared, so the purpose was met with full privacy compliance.  
2. Retail firms that provide insights at a collective level rather than individual and yet retain demographic profiles through clusters. Thereby providing CPG firms with a good view into buying behaviour without divulging private data.  
3. An insurance firm leveraged data science teams worldwide without fear due to the values and identities having been changed in the data while maintaining the variance levels and patterns.  
4. One of the biggest tech companies in the world used differential privacy to share data between third-party vendors and the organization in a secure way which helped them in improving the overall user experience.  

Are there any rules organizations should follow to derive the best results from their AI practices? 

By studying the behaviors of AI practitioners and interviewing experts in the field, we identified three critical data+AI action areas – data processes, advanced AI, and AI team structure – that can help companies deliver on the great expectations they hold for data and AI. 

Please elaborate on the importance of having good business leaders for the successful implementation of AI in an organization. 

Business leaders ensure that AI efforts remain relevant to the business question. Take a retail example: Many companies have used AI to build virtual “try-on” tools for clothes and accessories. Our research showed that this tool did not lead to high satisfaction. Retail AI that helps with checkout speed and inventory management has been much more effective, our survey found. 

What is a good business leader? Some would do or invest in AI for sound bytes but not contemplate the sight to value, i.e., how will this work in the daily life of a business user and generate better business? Thereby leading to AI projects that never scale since they were only supposed to be POCs to be talked about. A good business leader will look at the long term, and in AI, more than anything else, the long term is when a firm will reap the value of AI. If a business leader seeks a case that will provide returns in less than a year, he will be misled. When the initiative is thought through, explaining it to the ranks and getting buy-in becomes that much easier. And can be implemented in phases, with learnings incorporated along the way, which will lead to building well and driving adoption.  

The year 2023 is around the corner. What, according to you, might be game-changing AI trends in the coming year?  

Cloud-based AI will continue to democratize data science and swell the ranks of citizen data scientists. This will lead to breakthroughs in novel AI uses, provided companies stick to high data process standards and AI ethics. 

AI used in data quality will achieve game-changing potential. Today, we work on algorithms that streamline and auto-correct data based on past patterns. But this is expected to become much more prevalent than it is today. And even essential for downstream AI. Another game-changing trend will be the acceptance that AI uses probabilistic models as we all do for learning. And so, errors will be there, but the question will be: are these errors lower than human errors? Once that is the criteria, AI adoption will increase. 

Could the year 2023 be a milestone for data governance and regulations? 

Data privacy rules will require more compliance requirements for global companies as more nations set their standards. 

Data regulations will increase, given the knowledge about the potential misuse of such data. As mentioned above, the role of AI in data quality will be game-changing. 

What would be the impact of real-time data on businesses? 

The impact is already here. Most companies refresh their data in near real-time, our survey finds. But it’s not about how fast you process your data, and it’s about finding the most instructive or critical data for your particular challenge. Moreover, that data often resides outside your tech estate, which reinforces the value of data-sharing – importing third-party data and sharing data with third parties. 

Firstly, most strategic decisions do not need real-time data. Automation is a big area for the application of AI, and many of these use cases require real-time data. Of course, without real time-data where the streaming is reliable, automation that depends on it cannot work. Organizations that get it right will start seeing better profitability and customer satisfaction than others. And that will propel others to adopt the same. Reliable real-time data will be a game changer in business automation.