Unlike many other previous technologies that never lived up to their hype, Generative AI already is adding enterprise value. But we don’t yet know its true potential or its risks.
Darlene Newman, DTCC Executive Director, Head of Internal Research and Innovation, joined other technology experts for a virtual fireside chat - Demystifying Generative AI and the Impact It’s Having on Businesses - sponsored by Enable. The panel members included Charlie Serotoff, Product and AI Leader, ex-Senior Director, Capital One, and Niels Meersschaert, Global Technology Leader, ex-McKinsey CTO, Global Banking Solutions. Enable Chief Strategy Officer Paul Clarke served as host and moderator.
Related: Why AI is a double-edged sword
The panel explored Generative AI’s impact on the workforce, avoiding bias and balancing innovation with safety.
Speed of Adoption Will Vary
According to Newman, more than 1,000 Generative AI tools launched in April 2023 alone. But the speed of adoption – and further innovation – will depend on the industry. “The level of adoption of Generative AI has varied by industry,” she said. “Those that are highly regulated [such as financial services] are taking a more calculated approach versus those that might be jumping in headfirst.”
As a designated Systemically Important Financial Market Utility (SIFMU), “The security and resiliency of [DTCC’s] platforms is paramount,” Newman explained.
Newman advised firms in regulated industries to take four steps to prepare for Generative AI in the enterprise:
- Establish proper governance. Only 21% of companies that are testing Generative AI solutions have AI policies in place. Address risks such as data leakage upfront.
- Understand the technology and solution landscape. It’s almost impossible to keep up with all the new solutions. Don’t be distracted by shiny objects – focus on how the technology can help your firm.
- Identify high-value use cases and develop a strategy. It’s important to have a plan and to stay focused on adding value.
- Pilot. Many firms already are at the pilot stage for at least basic use cases.
Building AI Literacy
Unlike many previous emerging technologies that did not live up to their hype, Generative AI already is proving its practical value. In addition to personal productivity applications, many firms, including DTCC, are using Generative AI to automate aspects of software development, legacy code modernization and data cleansing. It’s not sexy, Newman said, but it will have a big impact on productivity.
A lot of the early value in Generative AI will be in automating repetitive processes, according to Meersschaert. “That’s not super exciting, but it’s the stuff that will actually improve a lot of productivity.”
Serotoff pointed to two categories of Generative AI applications in the enterprise: operational efficiencies, such as faster data analysis and content generation; and customer-facing solutions, such as chatbots and personalization, which are arguably more complex.
“The more that people can understand how AI works and what it can accomplish, the more it will help them apply those principals to much more complex customer-facing solutions in the future,” said Serotoff.
The real value of Generative AI, though, will be in applying Large Language Models to your own corporate data, Newman noted. That requires consumable data but “most organizations’ data isn’t ready,” she added. DTCC is getting ready now, assessing how to use using machine learning to clean its data for consumption in Large Language Models.
Culture Shift
There are two cultural obstacles that need to be overcome, according to Serotoff: human nature’s fear of change, and the very real harms that could result if we do not establish the proper precautions. People have always been afraid of rapid technological progress, he said. “This technology is here. We need to figure out how to harness it in the right way.”
While Generative AI may democratize coding skills and technology capabilities, for example, people will need to understand how it works so they can use it to build their own tools and outcomes. Training at all levels will be required. “What kind of skillsets do we need to teach and train in the future?” Newman asked. “People are going to be able to do more things themselves than ever before – how do we get them ready?”
Specialization in the software engineering space is likely to give way to generalists, according to Meersschaert, and work will become more outcome based rather than code based. The question will be, “‘What do I want this application to do?’ Not: ‘What is the specific code that I have to write?’” he said. “A lot of the stuff that we used to take for granted as a necessary evil is eliminated for us – it is automated away.”
The Path Forward
In the end, trust that Generative AI’s outputs are truthful and accurate will be a critical driver of adoption, DTCC’s Newman said. If you want to eliminate biased outcomes, examine your inputs. In addition to the model itself, “Is your dataset objective? The teams that are building the models – do they have a diverse perspective?” Newman asked.
Serotoff compared Generative AI’s power to the discovery of fire. “It can cook your food, but what else can fire do?” he asked. “It can burn a city to the ground.” He stressed the role of regulation in ensuring fair and safe innovation but highlighted the need to improve regulators’ AI literacy to make sure they implement the right regulations - an incredibly complex challenge, he acknowledged.