There are a lot of things cooking at the moment, but I want to finish the arc that started with the first three issues. I said next would be builders. We'll get there — a panel on Apprenticeships in NYC pulled me forward to a bigger question first. Let's get into it.

Build the HR on-ramp one redesigned role at a time.

A few weeks ago I sat in a Tech:NYC and Accenture panel on apprenticeships and early careers in the age of AI. One stat from the event has been with me since: entry-level job postings in New York City are down 37% since 2022. That drop maps almost exactly to the release of ChatGPT. The panelists noted there is probably correlation but not necessarily causation, which feels right but still curious.

Same panel, same event, a counter-stat: postings for jobs that are augmentable by AI — not automatable, augmentable — are up 3%. Same labor market. Opposite trajectories.

That gap is the story. AI isn't just absorbing tasks. It's quietly absorbing the on-ramp. The work people used to learn on — the repetitive, "watch how this gets done" work of an entry-level role — is the work AI takes first. The roles built on top of that work are the ones quietly disappearing.

For HRBPs, the recursive bit is the part that is concerning. Every great HRBP I know started somewhere. Some came up through the traditional HR career path and others (like me) started in another function and found their way to HR. Either way, we started learning in an entry-level role. Most of those roles are not being backfilled the way they used to be. We're not just the architects of how the on-ramp gets redesigned for the rest of the org. We're the people whose own profession's on-ramp is quietly closing.

So this issue is the fourth deliberate decision in a row: invest. What happens at the org level — and at the labor-market level — when every HRBP makes the first three deliberate decisions, and the entry-level work disappears as a byproduct? Who learns the craft now?

If you're new here, the first three issues set this up. Links are at the bottom.

The math tells you where the on-ramp is collapsing

Andrew Tien, VP of Strategic Partnerships at Pursuit, a tech training nonprofit in NYC, said something on the panel that I keep coming back to. His team rebuilt a workforce-readiness curriculum from scratch in two weeks using AI tooling. The previous version took ten months. They now update the curriculum daily.

Two weeks instead of ten months. Daily updates instead of an annual rewrite. The training providers are moving faster than most enterprise HR functions are moving. That's the pace problem hiding inside the on-ramp problem.

Here's the shift. The signal in the data isn't "AI is replacing entry-level workers." The signal is "AI is replacing entry-level tasks — and the roles built around those tasks are the ones disappearing." Roles built around augmentable tasks are growing. The hinge is in which tasks the role is built on.

That hinge is the design constraint we have to start using.

HRBPs own the on-ramp by design

Role design. Hiring criteria. Pipeline planning. Development paths. Internal mobility. All of it sits at the HRBP intersection. We are the one person in the system whose job covers both what's the work today? and how does someone learn to do it?

Closing the gap between those two questions for one role is a high-leverage move. It does not require apprenticeship-policy expertise. It requires a conversation with a hiring manager about what their entry-level role is actually for.

The output of that conversation isn't a workforce strategy proposal. It's one role, redesigned on paper. The augmentable/automatable distinction applied to the tasks the role contains, then the role rebuilt from the augmentable layer up. What you end up with looks something like the role Accenture has started calling an AI Workflow Generalist — a junior plus AI hybrid, where the human is doing augmentable work from day one and learning the craft through it.

That role is hireable. The "hire someone to do the automatable work" version is increasingly not.

Five questions to redesign one entry-level role

Take these into a 30-minute session with a hiring manager who has an open entry-level role they can't justify filling. Same structural pattern as the Issue 002 (codify) and Issue 003 (protect) questions, pointed at the on-ramp:

1. What entry-level role on your team is hardest to justify hiring for right now?

2. Of the tasks in that role today, which are automatable, and which are augmentable? Sort each task into one bucket — don't blend.

3. If we shed the automatable tasks, is what's left a real job? Or a hollow shell?

4. What craft does this role exist to teach? Where else does someone in our org learn that craft if this role disappears?

5. If you redesigned this as a "junior + AI" role from scratch, building up from the augmentable layer, what would it look like, and what would it cost?

Each question moves the conversation from "we can't afford the old role" to "the old role isn't the right shape — let's design the next one." The hiring manager walks out with one redesigned role on paper. You walk out with the input for naming what your org's next-generation pipeline actually looks like.

My own version of this

I sat down last weekend with the augmentable/automatable distinction and ran it on the version of my first HR role I could remember. The bulk of what I did in that seat is now automatable. The slice that wasn't — sitting in rooms, learning how people actually decide things — is where the craft lived.

But the automatable work was also where I learned. I had to figure out how to do it. How to deliver it. How to shape it for the client and what they expected. That's how I learned to direct the automation — by doing the work first. If I had been hired in 2026, the automatable layer would be gone, and so would the path I took to learn how to direct it.

That's not a memorial post. That's the shape of the role we'd have to design for the next person like me.

Before you do anything

Pick one open entry-level role on a team you support that the hiring manager can't justify filling. Take the five questions into your next 1:1. Redesign that one role.

That's the smallest version of the move. It doesn't require an apprenticeship program. It doesn't require a workforce strategy. It requires a 30-minute conversation about what the role is actually for, with someone who can change it.

What's the role on your patch that's hardest to justify filling right now? Hit reply and tell me — I read everything that comes in.

Subtract. Codify. Protect. Invest. Four deliberate decisions. That's the arc. Each one a choice an HRBP names before AI makes it for them.

Next issue, I'm picking up the thread that's been running underneath all four: HRBPs as builders. Not just users of AI tools — designers of them. The panel pulled me forward to the on-ramp question first. We'll get to builders next.

-Josh

The prompt I'm using this week

This pairs with the five questions above. Use it in a 1:1 with a hiring manager who has an entry-level role they can't justify filling.

"I'm working with [HIRING MANAGER NAME] to redesign one entry-level role on their team. The role is [ROLE TITLE]. The team's working assumption is that they can't justify filling it as currently scoped. We want to test whether the role is the wrong shape, not the wrong idea. Help me run a short discovery on the five questions below. For each one, suggest a follow-up I can ask if their answer is short or vague. After the conversation I'll come back with their answers, and we'll draft a redesigned version of the role together — built up from the augmentable tasks first, with the automatable work assumed handled by AI.

1. What entry-level role on your team is hardest to justify hiring for right now?

2. Of the tasks in that role today, which are automatable, and which are augmentable? Sort each task into one bucket — don't blend.

3. If we shed the automatable tasks, is what's left a real job? Or a hollow shell?

4. What craft does this role exist to teach? Where else does someone in our org learn that craft if this role disappears?

5. If you redesigned this as a 'junior + AI' role from scratch, building up from the augmentable layer, what would it look like, and what would it cost?"

Why it works: Hiring managers default to defending the role they already wrote the JD for, or to closing the req. The prompt forces a third option: redesign. Sorting tasks one bucket at a time is the move that does the most work — it stops people from blending "we still need a human to do this" with "we still need this to be a full-time job." Those two things have come apart, and the prompt makes the gap visible.

I've started running this on my own past roles before I run it with anyone else. Useful gut check on whether the framework holds.

If you try it, tag it #buildtogether and show me what you built. I'll feature the best ones in a future issue.

If you want to go to the source:

- The Extra Time Problem — Issue 001, 2026-04-25 (the deliberate-subtract setup)
- Feeling like it's Groundhog Day? AI can help. — Issue 002, 2026-05-09 (the deliberate-codify middle move)
- What only a human can do — even when AI could — Issue 003, 2026-05-16 (the deliberate-protect move)
- Tech:NYC × Accenture — A Path Forward for Apprenticeships & Early Careers in the Age of AI, April 30, 2026. Report: nyctech.org/accenture
- Center for Urban Future — NYC labor-market data
- Pursuit — AI-native training program model
- NYC Apprenticeship Accelerator — 220 new programs statewide

If someone you know is navigating the same question, forward this their way. New here? Subscribe at thehrbplab.ai to get the next one. Want to catch up on past issues? Browse the archive.