Beyond the AI Hype: How TIAA Successfully Modernized Its Middle Office Operations

Harvard Business Review, citing economists who linked large-scale adoption surveys to administrative payroll records, delivers an uncomfortable verdict: two years on from large-scale AI rollouts, employers have poured capital into automated tools in the exact domains where the technology should have moved the needle — and the effect on earnings and hours is statistically indistinguishable from zero. Workers report productivity gains. The bottom line doesn't. That's not a paradox. It's the predictable outcome of treating capability as outcome.

The transformation that almost didn't

TIAA, the nonprofit financial services giant, walked into 2022 with more than 80% of its IT workloads running on fragmented, end-of-life platforms — the operational definition of an unsecured attack surface. Chief operating, information, and digital officer Sastry Durvasula, who joined in early 2022, pitched the board on a multi-phase modernization called TETRIS. Phase one, launched in 2023, replaced the core stack with 10 new enterprise platforms. Phase two, launched in late 2024, enabled 87 use cases across major business lines, cut tech debt nearly in half, consolidated 17 design systems, accelerated product launches by 35%, and extended multilingual capabilities to more than 185,000 non-English-speaking customers. TETRIS earned CIO 100 Awards in 2025 and 2026.

The framing is honest. "As you expect in a company of our size, we have problems of yesterday, today, and tomorrow being solved at the same time," Durvasula says. That is the operational reality of any enterprise trying to modernize without breaking the revenue engine.

The 95% clause

Large digital transformation projects fail at rates as high as 95%. TIAA's TETRIS sits on the right side of that statistic, so far. The wider pattern is more concerning: organizations layering AI capabilities onto infrastructure that was never designed for the workload, decommissioning nothing cleanly, and calling the result "transformation." The middle office — the operational machinery between customer-facing systems and back-office accounting — is precisely where model failures, data lineage gaps, and integration debt compound silently before they surface in an audit or a breach post-mortem.

Adoption is not outcome. Rollout is not return. The organizations that will pull ahead treat AI as a tool inside a disciplined modernization program — not a replacement for one. The rest will keep filing press releases about productivity gains while the financials stay flat.