Key findings

1
Between the week of 3 Aug 2026 and the week of 24 Aug 2026, open roles in Agentic Tasks & RL Environments on the platforms tracked since 3 Aug 2026 fell from about 17 to a low of 11 in the week of 17 Aug 2026, then rose to 15 (-12.7% over the 4 weeks, against +9.0% for all AI training work on the same platforms).
Figure 1
2
4 platforms listed the 17 open roles in Agentic Tasks & RL Environments; Mindrift held 7 of them and the three largest 16 (HHI 3,149, highly concentrated, against 1,993 for all AI training work).
Figure 3
3
17 of the 17 open roles in Agentic Tasks & RL Environments (100.0%) stated an hourly rate, with a median of $50/hr (middle half $40 to $80), 23% below the $65/hr of all AI training work.
Figure 4
4
5 of the 17 open roles in Agentic Tasks & RL Environments accepted applicants from any country, against 51.3% of all AI training roles; of the 20 countries compared, residents of the United States (13), then the United Kingdom (11), then Argentina, Brazil and Colombia (9 each) could apply to the most, and residents of 8 of the countries (among them Bangladesh and Egypt) to the fewest (5 each).
Figure 9
5
The words used most in the titles of open roles in Agentic Tasks & RL Environments were Engineer or developer (12), Evaluator or reviewer (10) and Expert or SME (5), of 17 open roles.
Figure 10
6
Over the 31 days from 1 Aug 2026 to 31 Aug 2026, open roles in Agentic Tasks & RL Environments on the platforms tracked at its start went up 13.3% (15 to 17): platforms listed 36 new roles and removed 34, and a removed role had been listed for a median of 7 days (22 across all AI training work).
Figure 2

Figure 1

Open roles, week by week

Is the number of open roles in Agentic Tasks & RL Environments rising or falling?

Week	Open roles (fixed panel)	Index	All AI training work (index)	All tracked platforms
3 to 9 Aug	17	100.0	100.0	17
10 to 16 Aug	13	75.4	102.2	13
17 to 23 Aug	11	64.4	107.7	11
24 to 30 Aug	15	87.3	109.0	15
31 AugPartial week	18	106.8	110.5	18

Sample. Average open roles in Agentic Tasks & RL Environments over each week's days, on the 15 platforms tracked since 3 Aug 2026 (the fixed panel; 4 of them list this work), 3 Aug 2026 to 31 Aug 2026.

The index is the panel's open roles against the week of 3 Aug 2026 (100), beside the same index for all AI training work on the same platforms.

Every platform listing this work in the span was tracked from the start of the series.

Weeks with fewer than 10 open roles on the panel, or cut by the span, are marked.

The series starts in the week of 3 Aug 2026 at the earliest: before it, this dataset's feeds were still being added and tuned.

Source. AITraining.jobs Research, The Market for Agentic Tasks & RL Environments Work, August 2026 edition.

Figure 2

Listings added and removed

How fast do roles in Agentic Tasks & RL Environments turn over, and is the stock growing on the same platforms?

Work	Open on 1 Aug 2026	Open on 31 Aug 2026, same platforms	Change	New listings	Removed	Median days listed
Agentic Tasks & RL Environments	15	17	+13.3%	36	34	7.3 days
All AI training work	1,520	1,591	+4.7%	938	867	21.6 days

Sample. Roles listed from 1 Aug 2026 to 31 Aug 2026: 51 in Agentic Tasks & RL Environments, 2,458 across all AI training work.

Change compares the roles open on the first and the last day of the window on the platforms tracked when it opened, so a platform added during the window does not read as growth.

A new listing first appeared in the window on a platform already tracked before that week; a platform's first week (its back catalogue) is not counted.

Median days listed runs from a removed role's listing date to its last sighting, for removed roles with a known listing date (at least 10; 34 here).

Source. AITraining.jobs Research, The Market for Agentic Tasks & RL Environments Work, August 2026 edition.

Figure 3

Who lists it

Which platforms list roles in Agentic Tasks & RL Environments, and what does each advertise?

Platform	Open roles	Share of open roles	Stating an hourly rate	Median per hour
Mindrift	7	41.2%	100.0%	n/a
Mercor	5	29.4%	100.0%	n/a
Braintrust	4	23.5%	100.0%	n/a
Terac	1	5.9%	100.0%	n/a

Sample. 17 roles in Agentic Tasks & RL Environments open at the end of 31 Aug 2026, by the platform that lists them: HHI 3,149 (highly concentrated), against 1,993 for all AI training work.

HHI: the sum of every platform's squared percentage share of the open roles, 0 to 10,000; under 1,500 is unconcentrated, 1,500 to 2,500 moderately and above 2,500 highly concentrated (US merger-guideline thresholds).

Shares are of listings, not of people hired or money paid: a platform posting one role per language or country weighs more here than its hiring may.

A median needs at least 8 roles stating an hourly rate.

Source. AITraining.jobs Research, The Market for Agentic Tasks & RL Environments Work, August 2026 edition.

Figure 4

What it pays

What do roles in Agentic Tasks & RL Environments advertise per hour, against AI training work as a whole?

Work	Median per hour	25th percentile	75th percentile	Stating an hourly rate	Roles stating pay
Agentic Tasks & RL Environments	$50	$40	$80	100.0%	17
All AI training work	$65	$40	$85	89.1%	1,418

Sample. Roles open on 31 Aug 2026: 17 in Agentic Tasks & RL Environments, 17 of them stating an hourly rate; 1,591 across all AI training work.

Pay is the hourly figure the posting advertises (the middle of a stated range). The middle half runs from the 25th to the 75th percentile.

Roles paid per task or per project, or that state no figure, do not state an hourly rate. A median needs at least 8 roles stating one.

Source. AITraining.jobs Research, The Market for Agentic Tasks & RL Environments Work, August 2026 edition.

Figure 5

How its rates are distributed

What share of roles in Agentic Tasks & RL Environments pay $50 an hour or more?

Advertised rate	Open roles	Share of priced roles	This rate or less	All AI training work
Under $15/hr	0	0.0%	0.0%	2.8%
$15 to $25/hr	0	0.0%	0.0%	3.9%
$25 to $35/hr	2	11.8%	11.8%	5.1%
$35 to $50/hr	5	29.4%	41.2%	21.1%
$50 to $75/hr	3	17.6%	58.8%	32.9%
$75 to $100/hr	5	29.4%	88.2%	15.0%
$100 to $150/hr	2	11.8%	100.0%	14.9%
$150/hr or more	0	0.0%	100.0%	4.4%

Sample. 17 roles in Agentic Tasks & RL Environments open on 31 Aug 2026 that state an hourly rate, against 1,418 across all AI training work.

Each role counts once, at the middle of its advertised range. The bands are those of the pay-transparency report.

Source. AITraining.jobs Research, The Market for Agentic Tasks & RL Environments Work, August 2026 edition.

Figure 6

Applicant demand per open role

Do roles in Agentic Tasks & RL Environments draw more or fewer applicants per opening than the average AI training role?

Roles	Clicks per role a week	Index	Clicks	Avg open roles
Agentic Tasks & RL EnvironmentsSmall sample	0.14	1.81x	9	14
All AI training work	0.08	1.00x	544	1,548
Its roles stating an hourly rateSmall sample	0.12	1.47x	7	14
Its roles stating no hourly rateSmall sample	0.74	9.29x	2	1
Its roles open worldwideSmall sample	0.00	0.00x	0	1
Its roles limited by countrySmall sample	0.15	1.93x	9	13

Sample. Apply clicks on a specific role on AITraining.jobs, from 1 Aug 2026 to 31 Aug 2026: 9 on roles in Agentic Tasks & RL Environments, against the days those roles were open.

Clicks per role a week = apply clicks / (open role-days / 7). Index 1.00 is the rate of every role on the site.

Rows with fewer than 30 clicks or 3 roles open on an average day are marked as a small sample.

Apply clicks measure interest among visitors to AITraining.jobs, who see roles open in their country and recent, priced roles first. Clicks are never split by platform.

Source. AITraining.jobs Research, The Market for Agentic Tasks & RL Environments Work, August 2026 edition.

Figure 7

Applicant demand by listing age

Do newer roles in Agentic Tasks & RL Environments draw more applicants than older ones?

Listing age	Clicks per role a week	Index (this work = 1)	Clicks	Avg open roles
Listed during the windowSmall sample	0.22	1.50x	6	6
Listed up to 4 weeks before itSmall sample	0.04	0.28x	1	6
Listed 4 to 12 weeks before itSmall sample	0.11	0.79x	1	2
Listed more than 12 weeks before itSmall sample	0.72	5.01x	1	0

Sample. Apply clicks on roles in Agentic Tasks & RL Environments, from 1 Aug 2026 to 31 Aug 2026, by when each role was listed relative to the window's first day (1 Aug 2026).

Listing date: the date the job board gives for the posting, or else the first day the role appeared in the feed. A role that arrived with a platform's back catalogue and gives no date has none.

The index divides each row's rate by the rate of all roles in this kind of work, so it reads within the kind.

Rows with fewer than 30 clicks or 3 roles open on an average day are marked as a small sample.

Source. AITraining.jobs Research, The Market for Agentic Tasks & RL Environments Work, August 2026 edition.

Figure 8

Who may apply: location rules

Can people apply to roles in Agentic Tasks & RL Environments from anywhere?

Location rule	Open roles	Share of its open roles	All AI training work
Open worldwide	5	29.4%	51.3%
Worldwide except listed countries	0	0.0%	9.8%
Only listed countries	12	70.6%	38.9%
No rule stated	0	0.0%	0.0%

Sample. 17 roles in Agentic Tasks & RL Environments open at the end of 31 Aug 2026, against 1,591 across all AI training work.

The rule is read from each posting's location text and, where the posting is silent, from what the platform states about where it hires.

Source. AITraining.jobs Research, The Market for Agentic Tasks & RL Environments Work, August 2026 edition.

Figure 9

Who may apply: roles open to each country

How many roles in Agentic Tasks & RL Environments can a resident of each country apply to?

Country	Open roles	Share of its open roles	Median per hour
Any country (worldwide roles only)	5	29.4%	n/a
United States	13	76.5%	$80
United Kingdom	11	64.7%	$80
Argentina	9	52.9%	$80
Brazil	9	52.9%	$80
Colombia	9	52.9%	$80
India	7	41.2%	n/a
Chile	6	35.3%	n/a
Germany	6	35.3%	n/a
Japan	6	35.3%	n/a
South Korea	6	35.3%	n/a
Philippines	6	35.3%	n/a
Saudi Arabia	6	35.3%	n/a
Bangladesh	5	29.4%	n/a
Egypt	5	29.4%	n/a
Ghana	5	29.4%	n/a
Indonesia	5	29.4%	n/a
Cambodia	5	29.4%	n/a
Nigeria	5	29.4%	n/a
Pakistan	5	29.4%	n/a
Palestine	5	29.4%	n/a

Sample. 17 roles in Agentic Tasks & RL Environments open at the end of 31 Aug 2026, for residents of a fixed set of 20 countries.

A country's count is the roles open worldwide, plus roles whose list names it, plus roles whose exclusion list does not. Roles with no rule stated count for no country.

The 20 countries are the same in every report of this series: those named most often by the location rules of open AI training roles on 9 Oct 2026, either as the only countries accepted or as excluded.

Median pay is over the roles open to that country that state an hourly rate (at least 8).

Source. AITraining.jobs Research, The Market for Agentic Tasks & RL Environments Work, August 2026 edition.

Figure 10

What the roles are called

What are roles in Agentic Tasks & RL Environments called: annotator, evaluator, expert or trainer?

Title word	Open roles	Share of its open roles	Median per hour	Platforms
Engineer or developer	12	70.6%	$41	3
Evaluator or reviewer	10	58.8%	$40	2
Expert or SME	5	29.4%	n/a	1

Sample. 17 of the 17 roles in Agentic Tasks & RL Environments open on 31 Aug 2026 name one of these words in their title (100.0%); 3 title words are named.

A title can use several of these words ('Expert AI Trainer') and counts for each. The vocabularies are those of the job-titles, languages and expert-fields reports.

Median pay needs at least 8 roles stating an hourly rate.

Source. AITraining.jobs Research, The Market for Agentic Tasks & RL Environments Work, August 2026 edition.

For labs, data vendors and researchers

Buy the data behind this report

The figures above are a summary. Behind them are records we collect ourselves, every day:

9,572 roles tracked, each with its history
Every role on 29 job boards, read daily since 13 Jul 2026: its pay and place, and the first and last day it was open.
6,200 apply clicks in 28 days
Which roles people actually apply to, by field, pay and kind of work. It happens on this site, so nobody else can count it.
39 lab and vendor hiring boards
The job boards of 25 AI labs and 14 human-data vendors, read every day: who is hiring for human data, and who is hiring to buy it.

What you can buy

The records
The rows behind any figure, with their history, not only the totals.
A regular feed
The same figures every week or month, as files.
A custom cut or benchmark
Your countries, languages or fields, or your own platform against the market.

Method and definitions

The public job boards of the AI training platforms this site tracks, read every day. Apply clicks are made on AITraining.jobs; full descriptions come from the platforms whose boards publish them. Every figure is computed by one database query over the window shown, the same query for the live report and for every edition. Each figure states its sample beneath it.

Kind of work: Each role is classified from its title and description into one of the twelve kinds of AI training work this site tracks; this report counts the roles classified as Agentic Tasks & RL Environments.
Fixed panel: The platforms already tracked when the weekly series opens. The index follows their open roles, so a platform added later (such as OpenTrain AI and Welo Data in September 2026) never reads as growth.
Index: The panel's average open roles in a week, divided by the same in the base week, times 100; shown beside the same index for all AI training work on the same platforms.
Open role: A role listed by a given day and still in its platform's feed that day.
New listing: A role that first appeared during the window on a platform already tracked before that week. A platform's first week in this dataset is its back catalogue and is not counted as new.
Removed listing: A role that left its platform's feed during the window.
Days listed: From a removed role's listing date to the last day it was seen in its feed. A median needs at least 10 removed roles.
HHI: Herfindahl-Hirschman Index: the sum of every platform's squared percentage share of open roles, 0 to 10,000. Under 1,500 unconcentrated, 1,500 to 2,500 moderately, above 2,500 highly concentrated.
Advertised pay: The hourly figure a posting states (the middle of a stated range). Medians need at least 8 roles stating a rate; the middle half runs from the 25th to the 75th percentile.
Clicks per open role a week: Apply clicks on AITraining.jobs on a specific role, divided by the days the roles were open, times 7.
Demand index: Clicks per open role a week divided by the same rate across every role on the site. Above 1, the roles draw more applicants than the average role.
Listing age: When a role was listed relative to the window's first day: during the window, up to 4 weeks before it, 4 to 12 weeks before or longer. The listing date is the job board's posting date, or else the first day the role appeared in the feed.
Location rule: Read from each posting's location text and, where that names no country, from what the posting's own description states: open worldwide, worldwide except listed countries, only listed countries, or no rule stated.
Open to a country: A role is open to residents of a country if it is open worldwide, its list names the country, or its exclusion list does not. The 20 countries compared are the same in every report of the series.
Window: Live: the last 28 complete days, in UTC, and the last twelve complete ISO weeks (from the week of 3 Aug 2026 at the earliest) for the weekly series. Edition: one calendar month, in UTC, and its weeks.

What this data cannot tell you

Platforms that recruit without a public job board are not counted. Several large ones invite workers to projects directly.
A listing is not a hire. Some platforms keep evergreen roles open for months, and some post one listing per language or country.
Apply clicks measure interest among the people who use AITraining.jobs, and what is shown first gets clicked more. Clicks are never published by platform.
Full descriptions are published by a few platforms only, so the requirements figure describes their roles and appears only when at least 10 roles of this kind are described.
A kind of work with few open roles gives figures that move with a handful of postings: rows on too little data are marked as a small sample.

How the job feeds are read and roles are classified: site methodology. How fresh, complete and consistent the data is, checked every day: data quality and coverage.

Editions

The live report covers the last 28 complete days and changes every day. Each complete month is also published as an edition, frozen two days after the month ends, so a figure you cite from an edition never moves.

Live dataLast 28 complete days, recomputed daily
September 2026 editionFrozen 10 Oct 2026
August 2026 editionFrozen 10 Oct 2026

Cite and reuse

Free to quote, chart and republish under CC BY 4.0 with a link to this page. This edition is frozen, so its figures will read the same later.

Citation

AITraining.jobs (2026). The Market for Agentic Tasks & RL Environments Work, August 2026 edition. https://www.aitraining.team/research/agentic-rl-environment-market/2026-08

BibTeX

@misc{aitrainingjobs_agentic_rl_environment_market_2026_08,
  author = {{AITraining.jobs}},
  title = {The Market for Agentic Tasks & RL Environments Work, August 2026 edition},
  year = {2026},
  howpublished = {\url{https://www.aitraining.team/research/agentic-rl-environment-market/2026-08}},
}

Data: every figure as CSV (one line per value) and JSON, sent to your work email.Pay and open-roles data

More research

The same figures for every other kind of work:

All 78 research reports