Readying business for the age of AI

Rapid advancements in AI technology offer unprecedented opportunities to enhance business operations, customer and employee engagement, and decision-making. Executives are eager to see the potential of AI realized. Among 100 c-suite respondents polled in WNS Analytics’ “The Future of Enterprise Data & AI” report, 76% say they are already implementing or planning to implement generative AI solutions. Among those same leaders, however, 67% report struggling with data migration, and others cite grappling with data quality, talent shortages, and data democratization issues. 

MIT Technology Review Insights recently had a conversation with Alex Sidgreaves, chief data officer at Zurich Insurance; Bogdan Szostek, chief data officer at Animal Friends; Shan Lodh, director of data platforms at Shawbrook Bank; and Gautam Singh, head of data, analytics, and AI at WNS Analytics, to discuss how enterprises can navigate the burgeoning era of AI.

AI across industries

There is no shortage of AI use cases across sectors. Retailers are tailoring shopping experiences to individual preferences by leveraging customer behavior data and advanced machine learning models. Traditional AI models can deliver personalized offerings. However, with generative AI, these personalized offerings are elevated by incorporating tailored communication that considers the customer’s persona, behavior, and past interactions. In insurance, by leveraging generative AI, companies can identify subrogation recovery opportunities that a manual handler might overlook, enhancing efficiency and maximizing recovery potential. Banking and financial services institutions are leveraging AI to bolster customer due diligence and enhance anti-money laundering efforts by leveraging AI-driven credit risk management practices. AI technologies are enhancing diagnostic accuracy through sophisticated image recognition in radiology, allowing for earlier and more precise detection of diseases while predictive analytics enable personalized treatment plans.

The core of successful AI implementation lies in understanding its business value, building a robust data foundation, aligning with the strategic goals of the organization, and infusing skilled expertise across every level of an enterprise.

• “I think we should also be asking ourselves, if we do succeed, what are we going to stop doing? Because when we empower colleagues through AI, we are giving them new capabilities [and] faster, quicker, leaner ways of doing things. So we need to be true to even thinking about the org design. Oftentimes, an AI program doesn’t work, not because the technology doesn’t work, but the downstream business processes or the organizational structures are still kept as before.”  —Shan Lodh, director of data platforms, Shawbrook Bank

Whether automating routine tasks, enhancing customer experiences, or providing deeper insights through data analysis, it’s essential to define what AI can do for an enterprise in specific terms. AI’s popularity and broad promises are not good enough reasons to jump headfirst into enterprise-wide adoption. 

“AI projects should come from a value-led position rather than being led by technology,” says Sidgreaves. “The key is to always ensure you know what value you’re bringing to the business or to the customer with the AI. And actually always ask yourself the question, do we even need AI to solve that problem?”

Having a good technology partner is crucial to ensure that value is realized. Gautam Singh, head of data, analytics, and AI at WNS, says, “At WNS Analytics, we keep clients’ organizational goals at the center. We have focused and strengthened around core productized services that go deep in generating value for our clients.” Singh explains their approach, “We do this by leveraging our unique AI and human interaction approach to develop custom services and deliver differentiated outcomes.”

The foundation of any advanced technology adoption is data and AI is no exception. Singh explains, “Advanced technologies like AI and generative AI may not always be the right choice, and hence we work with our clients to understand the need, to develop the right solution for each situation.” With increasingly large and complex data volumes, effectively managing and modernizing data infrastructure is essential to provide the basis for AI tools. 

This means breaking down silos and maximizing AI’s impact involves regular communication and collaboration across departments from marketing teams working with data scientists to understand customer behavior patterns to IT teams ensuring their infrastructure supports AI initiatives. 

• “I would emphasize the growing customer’s expectations in terms of what they expect our businesses to offer them and to provide us a quality and speed of service. At Animal Friends, we see the generative AI potential to be the biggest with sophisticated chatbots and voice bots that can serve our customers 24/7 and deliver the right level of service, and being cost effective for our customers. — Bogdan Szostek, chief data officer, Animal Friends

Investing in domain experts with insight into the regulations, operations, and industry practices is just as necessary in the success of deploying AI systems as the right data foundations and strategy. Continuous training and upskilling are essential to keep pace with evolving AI technologies.

Ensuring AI trust and transparency

Creating trust in generative AI implementation requires the same mechanisms employed for all emerging technologies: accountability, security, and ethical standards. Being transparent about how AI systems are used, the data they rely on, and the decision-making processes they employ can go a long way in forging trust among stakeholders. In fact, The Future of Enterprise Data & AI report cites 55% of organizations identify “building trust in AI systems among stakeholders” as the biggest challenge when scaling AI initiatives. 

“We need talent, we need communication, we need the ethical framework, we need very good data, and so on,” says Lodh. “Those things don’t really go away. In fact, they become even more necessary for generative AI, but of course the usages are more varied.” 

AI should augment human decision-making and business workflows. Guardrails with human oversight ensure that enterprise teams have access to AI tools but are in control of high-risk and high-value decisions.

“Bias in AI can creep in from almost anywhere and will do so unless you’re extremely careful. Challenges come into three buckets. You’ve got privacy challenges, data quality, completeness challenges, and then really training AI systems on data that’s biased, which is easily done,” says Sidgreaves. She emphasizes it is vital to ensure that data is up-to-date, accurate, and clean. High-quality data enhances the reliability and performance of AI models. Regular audits and data quality checks can help maintain the integrity of data.

An agile approach to AI implementation

ROI is always top of mind for business leaders looking to cash in on the promised potential of AI systems. As technology continues to evolve rapidly and the potential use cases of AI grow, starting small, creating measurable benchmarks, and adopting an agile approach can ensure success in scaling solutions. By starting with pilot projects and scaling successful initiatives, companies can manage risks and optimize resources. Sidgreaves, Szostek, and Lodh stress that while it may be tempting to throw everything at the wall and see what sticks, accessing the greatest returns from expanding AI tools means remaining flexible, strategic, and iterative. 

In insurance, two areas where AI has a significant ROI impact are risk and operational efficiency. Sidgreaves underscores that reducing manual processes is essential for large, heritage organizations, and generative AI and large language models (LLMs) are revolutionizing this aspect by significantly diminishing the need for manual activities.

To illustrate her point, she cites a specific example: “Consider the task of reviewing and drafting policy wording. Traditionally, this process would take an individual up to four weeks. However, with LLMs, this same task can now be completed in a matter of seconds.”  

Lodh adds that establishing ROI at the project’s onset and implementing cross-functional metrics are crucial for capturing a comprehensive view of a project’s impact. For instance, using LLMs for writing code is a great example of how IT and information security teams can collaborate. By assessing the quality of static code analysis generated by LLMs, these teams can ensure that the code meets security and performance standards.

“It’s very hard because technology is changing so quickly,” says Szostek. “We need to truly apply an agile approach, do not try to prescribe all the elements of the future deliveries in 12, 18, 24 months. We have to test and learn and iterate, and also fail fast if that’s needed.” 

Navigating the future of the AI era 

The rapid evolution of the digital age continues to bring immense opportunities for enterprises globally from the c-suite to the factory floor. With no shortage of use cases and promises to boost efficiencies, drive innovation, and improve customer and employee experiences, few business leaders dismiss the proliferation of AI as mere hype. However, the successful and responsible implementation of AI requires a careful balance of strategy, transparency, and robust data privacy and security measures.

• “It’s really easy as technology people to be driven by the next core thing, but we would have to be solving a business problem. So the key is to always ensure you know what value you’re bringing to the business or to the customer with the AI. And actually always ask yourself the question, do we even need AI to solve that problem?” — Alex Sidgreaves, chief data officer, Zurich Insurance

Fully harnessing the power of AI while maintaining trust means defining clear business values, ensuring accountability, managing data privacy, balancing innovation with ethical use, and staying ahead of future trends. Enterprises must remain vigilant and adaptable, committed to ethical practices and an agile approach to thrive in this rapidly changing business landscape.

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff.