Beyond Faster Work

Every AI conversation should be an operating model conversation

What happens when every enterprise gets AI agents, but nobody rethinks the work?

Most leaders aren't asking that question. They should be. Because the window to get this right is shorter than they think, and the upside for those who do is enormous.

The substrate is that every AI conversation should be an operating model conversation.

Today, Anthropic and Infosys announced a collaboration to build the next enterprise agentic layer. Between Claude Cowork, their legal tool, and the broader wave of recent releases, Anthropic is positioning itself as the operating system for how knowledge work gets done at scale.

I’ve spent the past year talking to leaders across financial services, healthcare, and manufacturing, and I keep running into the same pattern. A company deploys an agent on a narrow task like invoicing, contract review, or ticket routing. They bolt it onto the existing process, measure time saved, and declare victory.

They’ve strapped a jet engine onto a horse-drawn carriage. And they’re celebrating the extra two miles per hour.

In McKinsey’s 2025 State of AI survey, only about one in four organizations has actually scaled an agentic system. The ones that have are nearly three times more likely to redesign workflows rather than just speed up existing steps.

Agentic AI doesn’t optimize tasks. It makes entire workflows obsolete. When an agent can orchestrate a full procurement cycle end-to-end within clear guardrails, “how do we speed up each step?” becomes the wrong question entirely. You start asking why the steps exist. And you can’t get there by thinking incrementally about a technology that moves exponentially. The leaders who haven’t internalized that are about to be caught off guard.

Infosys sits inside the operational backbone of nearly 1,900 enterprises. Core banking systems across hundreds of banks. Tech stacks powering much of the Fortune 500. They are now standing up a Center of Excellence to build industry‑specific AI agents in telecom, financial services, and manufacturing.

When a firm with that reach teams up with the fastest-moving enterprise AI company to build agentic infrastructure, the effects multiply across the network.

The companies that grasp this early will strip out entire layers of process that only existed because humans needed handoffs to coordinate. They’ll design around outcomes, not activities. The gap between them and everyone else will look a lot like what happened when digital-native companies ran circles around incumbents still “digitizing” paper processes. Except this time, the gap will open much faster.

Now here’s the harder question, and I think it’s the most important one. When agents handle judgment, analysis, and synthesis, the temptation is to outsource thinking itself. Let the agent summarize. Let it propose a strategy. Let it decide what matters. Every time you do that without engaging your own mind, you might save some time, but you’re eroding the one capability that actually compounds.

The most powerful technology in human history is still the human mind at full capacity, augmented by tools that extend its reach. Human judgment amplified by agentic intelligence. That’s where the real leverage sits, and the scale of it is massive.

The leaders who define the next decade will be the ones who used agents to think harder, explore wider, and make decisions they couldn’t have made alone.

Every serious enterprise will have agents. That part is settled. What’s unsettled is whether you’ll use them to do the same work faster or to do work that wasn’t even possible twelve months ago.

That’s a leadership decision, and the leaders who make it well are going to build something extraordinary.

Until next time,

Ram