Sastry Durvasula says he relishes the challenges legacy companies face implementing new technologies like artificial intelligence and cloud computing.
“My ‘brand’ is that I love transforming these 100-year-old iconic companies with the power of technology,” says Durvasula, chief operating, information and digital officer at 107-year-old financial giant TIAA.
During his career, which also included stints at American Express, insurance brokerage and consulting company Marsh McLennan, and consulting firm McKinsey, Durvasula sometimes contemplated switching to a pure-play tech company. The feeling was especially strong during Silicon Valley boom cycles, so he could learn more about innovation like blockchain and social media. But ultimately, Durvasula decided to stay with legacy, multinational businesses to help them modernize their payments processing and bring their businesses into the digital age.
He joined TIAA, ranked 96th on the Fortune 500, a little over three years ago initially to oversee the company’s client services and technology organization. Within a couple months, it underwent a reorganization in which business-facing CIOs were also given responsibility for all of the technology supporting retirement services, wealth management, and asset management. Previously, there was no single tech leader with such broad oversight.
The new CIO leadership team includes a mix of internal candidates who were promoted to those roles and external recruits. Durvasula also lured talent externally to oversee shared services within the company like cybersecurity and AI.
Early in his tenure at TIAA, Durvasula forged a partnership with Google to bring AI to his company's customer service centers, deployed new machine learning techniques in asset management, and built a more centralized data foundation. “When it comes to data and AI, there’s disproportionate attention towards generative AI,” says Durvasula.
That includes the creation of TIAA gAIt, a proprietary AI tech platform that lets users match a variety of large language models to the desired use case. LLMs from OpenAI, Anthropic, and Meta are among the options. TIAA has also deployed an AI assistant, which Durvasula calls a “research buddy,” that helps employees in asset management research and comb through quarterly earnings and investor relations documents.
The TIAA gAIt also has an external use case, letting clients ask questions about their different retirement accounts and receive generative AI-crafted responses that have less financial jargon. The assistant pulls information from multiple company sources, including the think tank TIAA Institute.
Humans are in the loop to test for bias, privacy, security, and prevent hallucinations. TIAA deploys hallucination mitigation agents, which cross-validate generated content against trusted datasets to filter out fabricated responses. The company also uses retrieval-augmented generation, which integrates both external and internal data sources using pre-approved retrieval pipelines, in an effort to enhance accuracy and reduce hallucinations.
Durvasula’s next rollout is what he calls “my gAIt,” which will give all TIAA employees access to generative AI. Testing is already underway, and the hope is that workers will use it to explore new ways to be productive.
That measured rollout also lets TIAA keep an eye on the expenses that come with generative AI deployment. “CIOs have to be very diligent to understand the total cost,” says Durvasula. “Because there is so much sizzle and less steak people can get carried away.”
He anticipates that immense competition between AI vendors will help lower the cost of AI and that by 2027, businesses like TIAA will have a realistic view of budgeting, similar to how companies track spending for cloud computing, software licenses, and other ongoing expenses.
One more pressing concern for Durvasula is bringing the workforce along as AI proliferates. He says about a third of all workers take a utopian view of AI's future, another third fear a dystopian, Terminator-esque one, with the rest somewhere in the middle. “Culturally, this is the most complex challenge that any C-suite has to deal with,” says Durvasula.
He offers his own hypothesis about the work AI will ultimately take over and what it will mean to employees. Eighty percent of today’s jobs will see 20% of their workflows disrupted by AI. The remaining 20% of jobs will see a much-more dramatic 80% impact from AI. With that expected turbulence ahead, TIAA offers workers AI training to promote upskilling.
“Whether you like it or not, AI is here to stay,” says Durvasula. “You might as well make the maximum use of it for your job.”
John Kell
Send thoughts or suggestions to CIO Intelligence here.
