The Next Phase of Work Is Agentic — And Most Teams Aren't Ready

AI has made knowledge workers measurably faster. Drafts come together in minutes, code gets reviewed in seconds, summaries write themselves. By any reasonable measure, we are doing more than ever.

And yet, when I talk to leaders across the enterprise, I keep hearing a version of the same admission: we know we're not getting close to the real value. The productivity gains are real but bounded. They sit on top of workflows that were designed for humans doing every step. We've made the existing motion faster. We haven't questioned whether the motion should exist.

That's the gap. And closing it requires something most teams haven't done yet: a hard audit of the work itself.

Start by asking what humans should actually be doing

The instinctive framing — "where can AI help my team?" — is the wrong starting point. It anchors on the current workflow and asks where to bolt on assistance. The better question is inverted: which steps in this workflow genuinely require human judgment, creativity, or accountability — and which are there because, until recently, only a human could do them?

That second category is larger than most leaders want to admit. Triage. Routing. Initial drafts. Status synthesis. Data reconciliation. Most enterprise workflows are stitched together with these connective tasks, and most of them exist because no other option was available.

Agentic systems change the available option set. An agent can pick up an incoming signal, gather context, take a bounded action, and hand off to a human only at the point where human judgment actually matters. The workflow gets shorter. The human touchpoint moves later, and the touchpoint itself becomes higher-leverage.

This is elevation, not replacement

I want to be direct about the concern that sits underneath this conversation. When you redesign workflows around agents, fewer humans touch each step. That sounds like labor reduction. In practice, what I've seen — and what I'd argue any team going through this transition should aim for — is something different.

The work that remains for humans is the work that was always supposed to be most valuable: the judgment calls, the creative leaps, the ambiguous situations, the relationships, the decisions about what to build and why. Those are the things knowledge workers signed up to do. They are also the things that have been systematically crowded out by the administrative scaffolding of modern enterprise work.

Pulling that scaffolding out doesn't shrink the role. It restores it.

What this looks like in practice

Three things tend to separate teams that get this right from teams that stall:

First, they audit honestly. They map the actual flow — not the documented one — and they're willing to name the steps that exist out of habit. This is uncomfortable. It surfaces process debt that someone built and someone owns.

Second, they invest in the interface and orchestration layer, not just the model. The model is rarely the bottleneck. The bottleneck is how work enters the system, how context gets assembled, how handoffs to humans are designed, and how the output gets back into the business. Teams that obsess over model selection and ignore orchestration tend to produce demos, not durable change.

Third, they redesign roles deliberately. If you're freeing your team from connective work, you owe them clarity about what they should be doing instead. "Focus on higher-value work" is not a plan. The plan is naming the specific judgment, creative, and customer-facing work you now expect them to own — and giving them the time and authority to actually own it.

The honest framing for leaders

The teams that will compound advantage over the next few years are not the ones with the best models. They are the ones willing to look at their workflows with fresh eyes and ask whether the humans in them are spending their time on what humans are uniquely good at.

That's a leadership question, not a technology question. And it's the question worth sitting with this quarter.