Five job titles that don't exist yet, but people are already doing the work

I spent twenty minutes last week scrolling my LinkedIn connections. Not looking for a job. Not recruiting. Counting.

I was counting how many people in my first-degree network are doing a job that does not officially exist yet.

The count surprised me. Eleven.

Eleven people whose formal job title is “Staff Engineer” or “SRE” or “DevOps Lead” or “Senior Security Engineer” — but whose public self-description reads like this: “I manage a fleet of 30+ AI agents across four product teams.” “I audit and correct agent-authored pull requests full-time.” “I design orchestration patterns for multi-agent workflows in a regulated environment.” Different jobs than they had in 2023. Same titles. The work has moved. The paperwork has not.

That is the whole of PRED-002 in one observation.

The prediction

PRED-002 — By December 2028, at least five new agent-native job categories will exist on LinkedIn, Indeed, or Glassdoor with 500+ open postings each. These roles will have no human-task predecessor. They will exist because AI agents need human oversight, coordination, or infrastructure.

Confidence: 5 out of 5.

Why I believe it

Chapter 9 puts this prediction at 5 out of 5 — one of the two highest-confidence claims in the book. The reason is not that I see into the future. It is that the work is already here and the titles are lagging.

The prompt engineer precedent is the headline. “Prompt engineer” went from a non-existent title to thousands of job postings between 2023 and 2024 — roughly eighteen months — with salary ranges reaching $100,000 to $300,000 in the US. That is not a prediction. That is the 2023–2024 timeline, fully documented, and it established that AI creates entirely new job categories rapidly, not just rebrands existing ones.

Chapter 9 names five agent-native roles that will follow the same arc: agent fleet manager, agent output auditor, multi-agent orchestration architect, agent trust officer, agent economics analyst. What the chapter does not do — because it is a chapter and not a newsletter — is point at the 2026 shadow titles for these roles. So here are three patterns I can see right now, scrubbed of identifying details:

Pattern 1 — the staff engineer who is really an agent fleet manager. Spends more than 60% of their week on agent lifecycle management: spinning up, scaling down, debugging drift, monitoring cost, handling incidents where an agent did something unexpected at 2 AM. Formal title unchanged since 2022. When asked what they actually do, they describe a job that did not exist in their 2022 JD.

Pattern 2 — the senior QA who is really an agent output auditor. Reviews agent-authored pull requests and agent-drafted customer responses as primary work, not human-authored output. Develops internal rubrics for what “acceptable” agent output looks like in their domain. Teaches junior reviewers how to spot the categories of mistake agents make (which are different from the categories of mistake humans make). No internal title for this discipline yet. The person invents one in their bio.

Pattern 3 — the solution architect who is really a multi-agent orchestration architect. Designs agent-to-agent handoff patterns, not human workflow diagrams. Thinks in terms of agent capability contracts, message formats, failure handoffs. Their day-to-day output is closer to protocol design than to system design. The title their employer assigned them was “Solution Architect” because that was the nearest box in the HR database.

These are not predictions. These are observations of March 2026. The 2028 formalisation is HR catching up with 2026 reality. Confidence 5 out of 5 is not confidence in the arrival of the work — the work is here — it is confidence in the arrival of the paperwork. Paperwork is one of the most predictable things in the world.

What would make me wrong

The published falsification trigger:

If by December 2028, fewer than three genuinely new job categories — not rebranded existing roles — appear with 200+ postings each, this prediction is wrong.

The word that carries the stake is new. Rebranding does not count. If “Staff Engineer, Agent Systems” becomes the dominant title instead of “Agent Fleet Manager,” the work has formalised but the category has not, and PRED-002 fails on the technicality that is also the substance.

The realistic failure mode is consolidation, not absence. HR departments hate new comp bands. The path of least resistance is to bolt “agent” onto existing titles rather than create new ones. If that is what happens, the shadow stays a shadow and the formal agent-native title never arrives. That is the outcome I am actively betting against — and the outcome the reader challenge below is designed to measure.

Reader falsification challenge

Reply to this email with two things: your current formal job title, and one sentence describing what you actually spent the last thirty days doing. Nothing else. No cover letter. One title. One sentence.

I will aggregate the responses into an anonymised shadow-title taxonomy and publish it on atin-agarwal.com/predictions/pred-002-agent-native-job-categories/, refreshed quarterly through December 2028. If fewer than three distinct shadow-title clusters emerge across 100+ responses by August 2026, the “already happening” thesis weakens, and PRED-002 reverts to its weaker “it will happen” form.

If the shadow-title clusters do show up — and I expect they will — the taxonomy itself becomes a quiet leading indicator for when LinkedIn, Indeed, and Glassdoor will add the formal titles. A shadow with 50 practitioners stays invisible. A shadow with 5,000 practitioners forces a form change.

What this means for you

If you are an engineer: rename yourself in your LinkedIn headline. Your headline is free real estate. “Staff Engineer / Agent Fleet Manager — 8 agents, 2.1M tasks closed last quarter” is the 2028 résumé line showing up three years early. When the formal title arrives, the recruiters will search for you by name first. Claim the surface area now.

If you are an HR lead: stop mapping agent work to existing titles. Your engineers are already self-describing in their bios. Your job architecture is lagging the work by at least eighteen months. Start tracking the bios before you update the JDs.

If you are a founder: the first formal agent-native title you post will get roughly 10x the applications of a generic SRE opening. I know this because I watched it happen with “prompt engineer” in 2023–2024. The early mover in titles gets disproportionate candidate funnel, because the title itself is the acquisition channel.

If you are a recruiter: the people you are ranking as “SRE with AI experience” are the people you should be ranking first for agent-native roles that do not exist on your requisition form yet. Your 2027 pipeline is your 2026 “SRE — AI-adjacent” list.

This issue is drawn from Chapter 9 of The AI Agent Economy — 15 falsifiable predictions with dates, numbers, and explicit triggers for being proven wrong. Pre-order on Kindle — $9.99. Release July 1, 2026. atin-agarwal.com/books

Read the full PRED-002 entry on the public tracking page → atin-agarwal.com/predictions/pred-002-agent-native-job-categories/

Previous issue: Issue 01 — Agents will out-work humans in lean software shops by 2029
Next issue: Issue 03 — The $100B market that doesn’t exist yet, and the invisible tax it will charge