How businesses can find and prioritize AI opportunities
Most organizations are fairly well acquainted with the basics of generative AI, but few have put together a vision and road map for executing the still-fledging technology at enterprise scale.
Société Générale, the sixth-largest bank in Europe, is ahead of the pack, according to chief digital strategy officer Noémie Ellezam. The 150-year-old company supports 25 million clients daily, providing retail, corporate, and investment banking services as well as those specific to insurance, private banking, and securities. Innovation is central to Société Générale’s charter, and AI technologies, now including generative AI, have long been a pillar of the company’s digital transformation strategy.
“AI is an accelerator of our digital strategy, with potential impact across all of our businesses and business areas,” Ellezam said during a webinar hosted by MIT Sloan Management Review and sponsored by Skillsoft. Banks have been using AI and advanced analytics for years, with use cases such as real-time payments and fraud and overdraft management. “Generative AI is a very big step on the AI journey, but it’s just a step, not a complete shift,” she said.
Financial services institutions could see a 3% to 5% productivity boost from generative AI, but those gains won’t be reached for another three to five years, Ellezam said. Even though the technology is rapidly maturing, business outcomes are slower to manifest due to challenges related to ideation and prioritization, organizational structure, employee skills, and risk management. “There’s huge potential, but it’s a bit of work to mobilize it,” Ellezam said.
In a conversation with MIT Sloan lecturer George Westerman, Ellezam shared some of the company’s best practices for finding and prioritizing AI, recommendations for preparing an organization for AI, and ways to structure initiatives so they deliver measurable business outcomes.
Some key takeaways from Société Générale’s generative AI implementation road map include the following:
Experiment and find successful use cases
Ellezam’s group has gathered more than 100 qualified generative AI use cases in less than three months, representing all areas of the business.
The group determined that generative AI might be able to outperform traditional AI in four primary categories:
- Virtual experts to help people find information in complex, heterogeneous documents for uses such as agent or legal assistance.
- Content generation for creating things such as requests for proposals or marketing campaigns.
- Client assistance for interacting with customers in personalized ways via chatbots or callbots.
- Code generation and software optimization.
Take a value-driven approach to turn ideas into reality
Since there are many potential ways to apply generative AI, it’s important to formalize a prioritization and risk management process to keep resources focused, Ellezam said.
Société Générale has established governance in a number of ways, including regular communication to investors on global value targets for AI use cases, and formal studies to determine feasibility, risk, and reusability potential. Business units must register all AI use cases in a central portal where frameworks are provided to deliver a value assessment. Once the use case is in production, units must report the effective realized value to see whether original estimates were accurate — a closed-loop process that helps evolve assessment methodologies.
“When you go from number of use cases to value, you realize that in the end, there are a couple of use cases in the portfolio which tick the box of AI but have no real relationship to the strategy,” Ellezam said. “It’s better to have [fewer] use cases with bigger value at stake than a broad range of use cases.”
Think globally when building generative AI competencies
Other AI initiatives are highly specialized and require specialized skills. Because generative AI is more accessible, it requires a more global approach to skilling workers, Ellezam said. Société Générale business and service units focus on framing use cases and assessing impact and feasibility, while generative AI methodologies and technical competencies are nurtured through a global center of excellence.
Poor data quality and rogue modeling are among the greatest AI-related risks, so being intentional and creating a culture of excellence is critical. “You have to make sure people have the right risk culture and know the way they can use AI in order for it to remain in the firm’s best interest,” Ellezam said. “It’s really a matter of people, culture, and skills, and for this, you need to be very clear and you need to be global.”
Source: GWFM Research & Study