In short
Only 20% of agentic AI roles are remote overall, but the figure ranges from 6% for AI infrastructure to 25% for data science. The gradient tracks proximity to physical compute and to the secure environments that host it, which makes remote share a proxy for what a role actually touches.
Remote share in agentic AI is usually quoted as one number. Across our 1,719 live listings that number is 20%. It is also close to useless, because the spread underneath it is four to one.
The gradient
Remote share by role category, across every category with at least 50 listings:
- Data Scientist: 25% remote, 75 listings
- AI Solutions Engineer: 24%, 88 listings
- General AI: 23%, 624 listings
- ML Engineer: 18%, 120 listings
- LLM Engineer: 15%, 54 listings
- AI Researcher: 14%, 71 listings
- AI Agent Engineer: 13%, 149 listings
- Generative AI: 13%, 55 listings
- AI Infrastructure: 6%, 184 listings
Categories with fewer than 50 listings are excluded, because a percentage from a small slice is noise dressed as a finding.
The pattern is proximity to compute
Order that list and it is not random. The roles at the top work primarily with data and with people. The roles at the bottom work with clusters, networks and the physical facts of where GPUs live.
AI infrastructure at 6% is the clearest case. That work touches capacity planning, interconnect, scheduling and the security boundaries around expensive hardware. A meaningful share of it happens inside environments that are not reachable from a laptop in another country, for reasons that are as much compliance as engineering.
Data science at 25% is the mirror image. The inputs are datasets and stakeholders, both of which travel.
The middle of the list is the interesting part. AI agent engineering sits at 13%, which is lower than most people expect for work that is fundamentally software. The explanation is that agent teams are new, small and still figuring out how they operate. Teams in that phase over-index on being in a room together, and they are the ones writing the job adverts.
Seniority pushes in the same direction
The role gradient is not the only one. Remote share also falls as seniority rises:
- Mid level: 22% remote, 754 listings
- Senior: 19%, 401 listings
- Staff: 16%, 322 listings
- Lead: 15%, 165 listings
Executive roles break the pattern at 23%, though from only 52 listings, and senior executive hiring follows different rules everywhere.
Below that level the trend is consistent: the more senior the role, the more likely it wants you present. Both gradients trace to the same underlying thing, which is that the highest-leverage work sits closest to the systems and to the people making decisions about them. If you are optimising for both remote and seniority simultaneously, you are searching a genuinely small part of the market.
The employer-size effect is the largest of the three
Here is the cut that matters most in practice. Companies with exactly one open role advertise remote at 36%, against 20% market-wide.
That is nearly double, and it is a bigger effect than either role category or seniority. Small employers are remote-friendly because they have to be. A ten-person company hiring its first agent engineer cannot restrict itself to one metro area and expect to fill the role, whereas a company with 200 openings can afford a location policy.
Since 376 of the 486 companies in our corpus have exactly one opening, that is not a niche. It is the majority of employers, and they are the ones most likely to say yes to remote.
What to do with this if you want remote work
If remote is non-negotiable, the practical move is to bias toward the top of the role gradient rather than to filter every listing and feel discouraged. A data science or solutions engineering role in the agentic economy is roughly four times more likely to be remote than an infrastructure role, and both are legitimately agentic AI careers.
Then stack the other two effects. Small employer, mid-level scope, data-adjacent role: that combination is several times more likely to be remote than a staff infrastructure position at a large company, even though both appear under the same search term.
The second move is to accept the trade explicitly where you will not bend. Infrastructure roles pay well and are among the most durable positions in the market, precisely because they are hard to staff. If you want that work, remote is a constraint you are choosing to fight rather than one the market imposes arbitrarily.
What this means for employers
If you are struggling to hire infrastructure engineers and your posting is onsite-only, you are competing inside the 6%. That is a defensible choice when the work genuinely requires physical or network proximity, and an expensive accident when it does not.
The question worth asking is which parts of the role actually need presence. Capacity planning and incident response on physical hardware are one thing. Writing the orchestration layer that runs on it is another, and teams routinely apply one location policy to both.
Why the headline number misleads
A candidate reading 20% concludes remote agentic AI work is scarce. A candidate reading the gradient concludes it is concentrated, which is a completely different search strategy. Scarce means give up or compromise. Concentrated means look in specific places, and there are three separate axes telling you where.
Method
Remote status is derived from the free-text location string on each listing, matching remote, anywhere, distributed, work from home and telecommute. That undercounts hybrid roles advertised without those words, and it undercounts employers who negotiate remote individually without advertising it, so treat every figure here as a floor rather than a ceiling. Categories and cuts below 50 listings are excluded throughout. Figures come from the live corpus and refresh daily. The open dataset carries the underlying counts, and our piece on single-opening employers covers that segment in more depth.
Frequently asked
What share of agentic AI jobs are remote?
20% across all 1,719 live listings, but that headline hides a four-to-one spread. Remote share ranges from 6% for AI infrastructure roles to 25% for data science roles.
Which agentic AI roles are most likely to be remote?
Data science at 25%, AI solutions engineering at 24% and general AI at 23%. These roles work primarily with data and with people, both of which travel, rather than with physical infrastructure.
Why are AI infrastructure roles almost never remote?
At 6% they are the least remote category, because the work touches capacity planning, interconnect, scheduling and the security boundaries around expensive hardware. Much of it happens inside environments not reachable from a laptop elsewhere.
How should I search if I need a remote role?
Bias toward the top of the gradient rather than filtering every listing and getting discouraged. A data science or solutions engineering role in the agentic economy is roughly four times more likely to be remote than an infrastructure one.
Are these remote figures a floor or a ceiling?
A floor. Remote status is derived from the free-text location string, which undercounts hybrid roles advertised without an explicit remote keyword, so actual flexibility is somewhat higher than these percentages show.