Stop Hiring Measurers. Build Agents Instead.

Cloudflare laid off a chunk of its workforce last week. The pattern wasn't random. The roles that disappeared were the ones built around tracking what others do, reporting it upward, and coordinating across teams.

Not bad roles. Roles that made sense when organizations ran on information friction — when moving data between teams required a human in the loop.

That friction is disappearing. And with it, the roles that existed to manage it.

I manage over 50 engineers across multiple teams. I've watched this pattern play out firsthand. Not as a headline, but as a series of quiet decisions about which roles to open, which to redesign, and which to replace with workflows.

The Question That Changed How I Hire

The old question was: "Can AI help this person?"

The better question is: "Should this be an agent problem from day one?"

That shift sounds subtle. It isn't. The first question assumes the role exists and asks how to augment it. The second question challenges whether the role should exist at all.

When I started asking the second question, my hiring decisions changed. Not dramatically at first. But over the past year, I've declined to open at least four roles that would have been automatic hires two years ago. In every case, the work got done — faster and more reliably — by a combination of workflows and existing team members.

A 6-Question Rubric for Every New Role

I've been using a scoring framework to evaluate roles before opening them. Six questions, each scored 1 to 5. The total tells you what to do.

1. Making vs. Measuring

Is this person building, selling, or deciding? Or are they mostly tracking and reporting on what others do? The more a role is about measuring rather than making, the more automatable it is.

2. Repeatability

Are the outputs genuinely one-off, or are they recurring reports, status updates, and approval flows? Recurring outputs are workflow problems. One-off outputs require human judgment.

3. Data Availability

Is everything they need already in your tools — logs, CRM, tickets, analytics? Or is it locked in conversations, relationships, and institutional knowledge? If the data is already structured, an agent can access it.

4. Judgment Complexity

Are they resolving genuinely messy trade-offs with incomplete information? Or applying clear thresholds to structured data? The latter is exactly what rule engines and AI agents excel at.

5. Error Risk

If AI gets this wrong, is it catastrophic — legal liability, safety risk, damaged key relationships? Or is it cheap to catch and fix? Low-stakes errors are fine for agents with human review.

6. Oversight Ease

Can a human review the AI's output quickly and spot mistakes? If the output is a report, a dashboard update, or a status summary, review takes seconds. If it's a nuanced client communication, review takes as long as doing it yourself.

What the Scores Mean

24–30: Build an agent. Don't open the role. The work is measurable, repeatable, data-available, and low-risk. Build a workflow.

18–23: Hybrid. Let AI do the measuring. Humans handle the deciding. Redesign the role around the judgment parts.

12–17: Keep and streamline. The person stays. But strip out the automatable parts so they can focus on what actually requires them.

6–11: Protect and amplify. High-judgment work. This is where your people create irreplaceable value. Give them AI tools, don't replace them with AI tools.

I built an interactive version of this rubric — try scoring a role on your team.

What This Looks Like at Scale

When I ran this rubric across the roles I've hired for in the past three years, a pattern emerged immediately.

Roles that score high (24+):

• Product ops and internal reporting
• Cross-team coordination and status aggregation
• Analyst-style "glue" roles that synthesize data from multiple systems
• Approval flow management and compliance tracking

Roles that score low (6–11):

• Senior engineers with system-level ownership
• Product owners with direct customer accountability
• Technical leads making architectural trade-offs
• Anyone whose job requires reading a room, not a dashboard

The people in the second group don't get replaced by AI. They get accelerated by it. They ship faster, make better-informed decisions, and spend less time on the measurement work that used to eat half their week.

The Org Design Lesson

The takeaway isn't "fire measurers later." It's "don't build those roles in the first place."

Hire builders and owners. Give them agents who do the measuring.

The people who get hurt when companies restructure aren't bad hires. They're people hired into roles that AI was quietly making redundant for years. The more humane thing is not to create those roles at all.

This is an org design problem, not a people problem. If you're building a team right now, the shape of that team should assume AI handles coordination, reporting, and status aggregation. Your humans should be doing the work that requires judgment, creativity, relationships, and accountability.

What to Do Monday Morning

If you're writing a job description right now:

• Run it through the rubric. If it scores above 24, build the workflow before you post the role.
• If it scores 18–23, redesign the role. Keep the judgment, automate the measurement.
• If it scores below 12, hire fast. That person will be more valuable with AI tools, not less.

And if you're in one of those measurer roles yourself? The answer isn't panic. It's migration. Start owning outcomes instead of tracking them. Move toward the work that scores low on this rubric. That's where the leverage is.

What's a role on your team that scores high on this rubric — and what would the agent version of it actually look like? I'm genuinely curious.