According to Forbes, MassMutual is accelerating its AI plans with a clear target: a fully agentic software development lifecycle (SDLC) by the end of Q1 2026. Sears Merritt, the insurer’s head of enterprise tech, revealed that early agent tests for tasks like drafting requirements and refining code have cut work that used to take a full sprint down to just hours. Those tests showed a consistent 35% productivity boost per team, a figure that held whether teams were reduced by 35% or kept the same size to increase output. The company is now expanding from virtual assistants to autonomous agents, with early IT help desk deployments slashing call times from 5-10 minutes to about two. This push is supported by a massive backend modernization that consolidated 19 core systems down to six.
From Assistants to Agents
Here’s the thing: there’s a big difference between an AI that fetches information and one that takes coordinated action. MassMutual’s early wins with virtual assistants for advisors were helpful, but the shift to agents is where the real transformation happens. An agent that doesn’t just find a policy clause but automatically starts a Slack thread to get a ticket approved? That’s a different beast. It’s moving from being a fancy search bar to being a semi-autonomous team member. And the impact is tangible—those faster resolutions directly led to a double-digit jump in employee satisfaction scores. Now, they’re aiming this at call centers for 2026. The goal is multimodal agents that handle voice and complex interactions, not just text. That’s a huge leap.
The Foundation Matters
You can’t run this new, fast AI software on old, clunky infrastructure. MassMutual’s “warp speed” claim isn’t just about buying the latest LLM API. It’s possible because they spent years doing the unglamorous work: modernizing the core tech stack. Consolidating 19 systems into six and migrating over 300,000 policies off legacy platforms is a monumental task. But it creates clean, accessible data and flexible architecture. That’s what lets them experiment quickly. Merritt’s point about build vs. buy is crucial in this context. With a modern foundation, you’re not locked in. You can build a prototype, buy a solution that gets better, or switch strategies entirely. That flexibility is a superpower in a market changing daily.
The Trust Equation
So, they have the speed and the infrastructure. But will people trust these agents with real work? MassMutual is tackling the classic AI hurdles head-on. Hallucinations are “largely addressed” for their use cases, which is a bold claim. More importantly, they’re focusing on audit trails and “chain of thought” reasoning—knowing exactly why an agent made a decision. That’s non-negotiable in insurance and finance. I think their approach to trust is smart: they’re actually measuring it with “trust scores” and asking employees for feedback. When given a choice between faster, sloppier outputs and slower, higher-quality ones, employees chose quality. That tells you a lot about what’s needed for real adoption. It’s not just about raw speed; it’s about reliable speed.
Rethinking the Manager’s Role
This is where it gets really interesting for the future of work. Merritt hints that agentic AI will change management, especially in people-centric functions. Imagine a manager whose direct reports are primarily AI agents. Their job becomes monitoring KPIs, tweaking prompts, and guiding continuous improvement—it’s a more quantitative, data-driven role. That’s a significant shift. It also highlights why MassMutual is spending so much time on education and mindset. The tech is a tool, but the superpower comes from teams knowing how to lean into it. The 35% productivity gain isn’t automatic; it’s the result of rethinking processes around what the agents can do. Basically, the organization itself has to become more agentic.
