Read the titles, not the headlines.
The story we tell ourselves about AI and work usually arrives via headlines or vendor pitches. The story the labour market is telling itself sits in the live job ads. The two don't always match.
What follows is a read on 20,203 live AI roles right now. Borrowed methodology from Anthropic's Economic Index; the data is from our own jobs index. Treat both as a snapshot of this week, not a forecast - see what this dataset permits us to say at the foot of the page.
- physical therapist +5.2×517 active · 0 a year ago
- physical therapist assistant +2.6×256 active · 0 a year ago
- help companies of all sizes embrace industry-changing technologies like rapids to analyze massive amounts of data to make better, faster business decisions. +76%76 active · 0 a year ago
- occupational therapist +46%46 active · 0 a year ago
- physical therapist - prn +44%44 active · 0 a year ago
- travel physical therapist +37%37 active · 0 a year ago
The picture that emerges from the live index right now isn't the headline picture. The headlines say AI is taking jobs. The data says jobs are reshaping, in three specific ways the headlines don't usually pick up. We argue them below - willing to be wrong out loud.
One. Where the titles are being created tells you more than where the layoffs are. "AI is replacing X" is a story that requires before-and-after data we mostly don't have. "AI has created the title Forward Deployed Engineer and there are now hundreds of live ads for it" is a story we do have. The latter is a leading indicator; the former is mostly anecdotal. (Caveat: a job ad existing is not the same as a hire happening, and a new title is not the same as new work.)
Two. The companies hiring hardest this week are not the ones with the biggest brand. Sort by jobs-this-week rather than total volume - see the panel below - and the ranking shifts. A company at 50% recent intensity is typically in an acute hiring round, often around a fundraise, a product launch, or a strategic pivot. That signal moves weeks before press coverage does. (Caveat: ATS plumbing varies - a feed that re-publishes is indistinguishable from a feed of new ads at the API level.)
Three. The senior-to-entry mix is the clearest read on where a discipline is in its life-cycle. A 70% senior split is a mature, saturated function - entry pipelines have thinned and the rewards live downstream. A 40% entry-level split with high volume is a category scaling fast enough to take people without a track record. Marketing, Operations and Sales each tell a different story; engineering tells the story everyone else is reacting to. (Caveat: the seniority classifier reads job-spec language, which compresses a noisy spectrum into three buckets.)
The most popular counter-argument to all of this - best made by Abigail Marks in The Conversation - is that AI hasn't yet caused mass unemployment, and the fear that it has is doing more economic damage than the technology. We think she's right about the fear being premature, and right that the data so far supports reshaping over replacement. The reading we make below assumes that frame.
What the foundation labs hire for first.
Forward Deployed Engineer, AI Deployment Strategist, Agent Reliability Engineer. None of these existed in volume two years ago. The companies running these ads are typically fifty times larger than their eventual buyers - when a title appears in the top tier here, you are looking at the work that will be commodified at scale within the year. Watch which titles rise; that's where to look for skills to pick up.
91 of the most-listed roles right now are titles that did not exist as named categories before the agentic-AI shift. The ones above sit in our current top twelve, ahead of established titles in adjacent disciplines.
Top live titles, ranked by current count- 01 physical therapist 517
- 02 physical therapist assistant 256
- 03 help companies of all sizes embrace industry-changing technologies like rapids to analyze massive amounts of data to make better, faster business decisions. 76
- 04 software engineer 72
- 05 senior software engineer 67
- 06 occupational therapist 46
- 07 physical therapist - prn 44
- 08 travel physical therapist 37
- 09 deployment strategist AI-native 36
- 10 data scientist 36
Who's moving, not who's big.
Total volume tells you who is big. Last-seven-day count tells you who is moving. Sort by recent intensity and the ranking shifts: companies with 40-60% of their current openings posted in the past week are typically in a discrete hiring round - a Series funded, a product launch, a strategic pivot. That signal moves weeks before press coverage does, and the titles they hire for are usually the cleanest read on what they're actually building next.
NVIDIA
1,638 this week · 1,880 total
ATI Physical Therapy
732 this week · 732 total
Sanford Health
530 this week · 530 totalFord Motor Company
126 this week · 529 total
Anduril Industries
92 this week · 499 total
Electronic Arts
86 this week · 396 total
Synopsys
68 this week · 704 total
Databricks
55 this week · 467 totalThe pattern that holds: NVIDIA and the foundation-model labs are the steady-state ceiling - they hire continuously, the index never empties out. Movers come and go above them on the seven-day cut, and that's the more interesting list. If you're reading this to figure out where to apply, ignore the steady names. Look at who's at the top of the panel above today and was not last week.
Where to apply effort, by life-cycle stage.
Engineering dominates by orders of magnitude - that's the easy read. The harder read is the senior / mid / entry split. A category that's mostly senior is mature and saturated at the top; pipelines into it are thin and rewards live downstream. A category with high entry-level volume is scaling fast and accepting people without a track record. If you are choosing a discipline to invest in, look at the entry-level columns below - that's where the doors are still open.
Management Consulting
100% senior. Mature function - the work demands experience, and entry-level pipelines are thin.
Healthcare
14% entry-level. A category scaling fast and hiring early-career people in volume.
| Category | Total | Entry | Mid | Senior |
|---|---|---|---|---|
| Engineering | 11,350 | 928 | 2,433 | 6,647 |
| Sales | 1,799 | 230 | 452 | 774 |
| Healthcare | 1,224 | 173 | 554 | 114 |
| Finance | 1,064 | 102 | 280 | 601 |
| Operations | 942 | 303 | 247 | 297 |
| Marketing | 676 | 75 | 223 | 322 |
Show 108 more categories
Three takeaways for three audiences.
Pick a category with high entry-level volume.
"AI engineer" is over-saturated and skill-gated. "Forward Deployed Engineer" and "AI Deployment Strategist" are reading the same skill stack but coming through doors that are still open - these companies are training people in. Look at the entry / mid columns of the categories above. Pick a category where the entry column is greater than 20% of the total.
The titles you don't recognise yet.
Agent Reliability Engineer. Applied AI Engineer. AI Deployment Strategist. These are not synonyms for "developer." They are roles built around the gap between what an LLM can do and what a business can rely on. Twelve months from now, the question your board will ask is which of these your team has. Use the trend chart in the hero to add the words to your vocabulary.
The work that's being commodified.
The titles being hired into companies fifty times your size are the work that arrives at your door as a SaaS feature within a year. Look at the trending list and ask: what would my team's job description look like if half of this were already automated? That is the readiness conversation worth having now, not in twelve months.
What this dataset permits us to say.
The dataset is young. Twenty weeks of live listings, sourced from 306 active feeds across employer ATS systems and aggregator APIs - no third-party panels, no surveys. Read the numbers on this page as a live tap off a growing dataset rather than the final word.
The exposure framework - "what AI could do" versus "what AI does" versus "what AI replaces" - is borrowed from Anthropic's Economic Index. We use their distinction throughout. Exposure is not the same thing as substitution. A title appearing in volume is exposure data. Whether the humans who used to do that work still have jobs is a separate question this dataset doesn't answer.
"AI-native titles" - Forward Deployed Engineer, AI Deployment Strategist, Agent Reliability Engineer, Applied AI Engineer, MCP Engineer, AI Engineer - are picked by name. The selection is editorial, not algorithmic.
"Aggressive hiring" sorts by jobs posted in the last seven days. The intensity percentage is jobs_last_7d / total_jobs per company. Caveat: ATS plumbing varies, and a feed that re-publishes ads on a cadence is indistinguishable at the API level from a feed of genuinely new ads. Treat the seven-day cut as directional, not surgical.
The entry / mid / senior split is YubHub's own role classifier reading job-spec language. The classifier compresses a noisy spectrum into three buckets - expect 5-10% noise on the mid-senior boundary. Categories with fewer than 200 live roles are excluded from the "where doors are open" callouts because the noise floor is too high.
The arguments on this page are ours. The data is the data's. Where they pull apart, we revise the argument.
Want to query this from inside Claude or ChatGPT?
The same data behind this page is exposed as an MCP server. Install @houtini/yubhub and your agent can query trending titles, hiring companies and category breakdowns directly. Free, MIT, public API.