Andrew Ng Just Priced the "AI-Forward Recruiter" at $90K. The Title Is a Trap.

DeepLearning.AI reposted a Mountain View "Recruiter (AI Forward)" role about nine days ago on Built In, banded at $90,000 to $130,000 and explicitly framed as "not a traditional recruiter role." Andrew Ng's company is the first frontier-AI-adjacent org to write the new recruiter archetype into a public posting, complete with the line "Recruiting like it's 2022 (before modern Generative AI) no longer makes sense." If you run talent at a startup or you are a recruiter trying to read the next 12 months, this JD is doing more work than the salary band suggests.

What the JD actually says

The role is classified Junior, in-office, and reports to a Lead Recruiter. The opening paragraph reads: "DeepLearning.AI is seeking an AI-forward Recruiter to support our Lead Recruiter and help build a world-class talent engine. In this role, you'll focus on the tactical execution of recruiting: sourcing candidates, screening, managing pipelines, coordinating interviews, and supporting offer processes, while using advanced AI tools to reimagine how recruiting should work."

So far, standard tactical recruiter. Then the JD pivots:

"You will partner closely with our engineering team as a domain expert on recruiting, and provide feedback on what tools would be most helpful. Thus, you will help shape our portfolio of advanced AI-enabled recruiting tools."

That second paragraph is the real job. The recruiter is being hired as an internal product user embedded with engineering, not just as a pipeline operator. They will source AI engineers, data scientists, and full-stack engineers across the broader AI Fund ecosystem of companies, while simultaneously acting as the design partner for the AI recruiting tools Ng's team is building. This is the recruiter-shaped version of the "AI Forward Deployed Engineer" pattern Ng popularized in The Batch, where an engineer embeds with a client to tune agentic workflows. Same template, applied inward, on the talent function.

The pay band is the story, not the title

Read the salary against any honest benchmark and the framing gets uncomfortable.

$90K to $130K

Built In SF reports the average technical recruiter base in San Francisco at $132,248, with $12,795 in additional cash compensation and a total of $145,043. Indeed pegs the California average at $125,695 across 325 job postings. Levels.fyi has the Bay Area recruiter range running from $138,000 to $215,000 in total compensation. DeepLearning.AI's top of band ($130K) sits roughly at the bottom of Levels.fyi's range and just under the Built In SF base average. The midpoint ($110K) is meaningfully below market for a Bay Area technical recruiter who can credibly source AI engineers.

The implicit thesis is straightforward: with the right AI tooling, one recruiter does the work of two or three, so paying a single junior headcount still beats a traditional team on output. That math may even be right for DeepLearning.AI. But if you are a talent leader benchmarking against this post in Q1, you should ask the harder question: is "AI-forward" quietly becoming the euphemism that justifies paying recruiters less while expecting them to absorb the work of the role beneath them?

AI-forward risks becoming the euphemism that justifies paying recruiters less while expecting them to absorb the role beneath them. </pull> The productivity claims supporting that math are vendor-sourced, not audited. The figure making the rounds, "AI sourcing has expanded candidate pools by an average of 340% while reducing sourcing time by 67%," comes from a Second Talent 2025 study amplified through recruiting-tech blogs. SHRM's own number is more defensible: AI adoption in HR doubled in a single year, from 26% to 43%. That is a step-change in adoption, not a step-change in proven ROI. Treat the productivity stats accordingly. ## The candidate pool barely exists yet Here is the part that should make every talent leader move faster. The recruiters who can credibly claim "AI-forward" today are a rounding error in the market.

stat number: ~36 label: U.S. "Technical Recruiter" profiles listing Generative AI as a skill note: Early employers include Apple, Figma, AWS, Aledade, and Inductive Automation.

A query across the open professional-network index for U.S. "Technical Recruiter" titles with "Generative AI" listed as an explicit skill returns roughly 36 matching profiles. Apple, Figma, AWS, Aledade, and Inductive Automation show up among the early employers. That is the entire visible pool DeepLearning.AI is fishing in at $90 to $130K. There are obviously more recruiters who use ChatGPT every morning and have not bothered to put it on their profile, but the explicitly-tagged pool is tiny, and that is what most boolean searches will surface.

If you are trying to source against this archetype, the bottleneck is not your ATS, it is your query. A boolean string for "AI recruiter" returns thousands of marketers. A boolean for "Generative AI" AND "recruiter" returns mostly vendor employees. This is exactly the friction we built [Refolk](/) to remove: you describe the person in plain English ("US technical recruiter who has actually shipped AI sourcing workflows at a startup, not a vendor") and get a ranked shortlist instead of a thousand false positives.

What changes for talent leaders in the next 12 months

Four predictions, none of them safe.

1. The JD language will spread fast

Ng coined the "AI Forward Deployed Engineer" framing in The Batch. His org publishing the recruiter version essentially names the archetype for everyone else. Expect "AI Forward Recruiter" or close cousins in JDs from AI Fund portfolio companies, Anthropic, OpenAI, Mistral, Scale, and Glean inside 90 days. If you want to be early, the time to write your version is now, before the candidate pool is even more contested than the 36-profile floor suggests.

2. "Domain expert plus tool shaper" replaces "pipeline manager"

The JD's second paragraph is the template. Recruiters will be hired partly to provide structured feedback to the engineers building internal AI hiring tools. This is a feedback-loop role. If your recruiter cannot describe, in writing, why a particular sourcing query failed and what the tool should do differently, they will not pass the new bar. Recruiters who can ship a Loom about a broken eval will out-earn recruiters who can ship a calendar invite.

3. The skill profile shifts faster than the title

One of the better lines from the broader 2026 AI-in-recruiting research: "Recruiters who will thrive are those who can interpret AI insights, identify when AI recommendations need to be overridden, and apply emotional intelligence to candidate interactions that AI cannot replicate. The recruiter's title may stay the same; their skill profile is evolving rapidly." Practically, this means the screen on your next recruiter hire should test for taste, not tool literacy. Anyone can run a prompt. Almost nobody can look at a 200-person AI shortlist and tell you which 12 to actually contact.

4. The first movers set comp norms

If DeepLearning.AI lands an "AI Forward Recruiter" at $110K, the band sticks. If they get 200 applications and only three pass the screen, they will quietly move the band to $140 to $170K by Q2, and the rest of the market will follow. The first 18 months of any new archetype are when comp gets locked in. This is the same dynamic that played out for ML engineers in 2017 and prompt engineers in 2023. The companies who paid early paid less.

How to actually source against this archetype today

The honest answer: stop searching for "AI recruiter." It is a polluted query. Search instead for the artifacts of the work.

Recruiters worth hiring against this JD have one or more of the following on a public surface: a write-up of how they built an outbound eval, a comparison of Findem versus Fetcher versus a hand-rolled stack, a GitHub repo (yes, recruiters with repos exist now) containing scrapers or scoring scripts, or a public Loom of their sourcing workflow. The signal is in the artifact, not the headline.

This is the second place Refolk earns its keep for talent leaders: the in-prose query "find me recruiters who have publicly written about building their own sourcing workflows in the last year, prioritize those at AI-native startups under 200 people" returns a list you can actually call, not a list you have to re-filter. It also catches the recruiters who are already doing AI-forward work but have not labeled themselves that way, which is the larger and more interesting pool.

For the rare candidate who applies cold to the DeepLearning.AI role itself, your first screen should be one question: "Show me a recruiting workflow you built or broke in the last six months, and tell me what the eval looked like." A traditional recruiter cannot answer that question. An AI-forward recruiter has three answers ready.

What this means for your own JD

If you are about to write your version of this posting, three edits worth making before you publish.

First, name the tools you actually use internally. Vague JDs ("leverage AI tools") attract vague applicants. Specific JDs ("we use Findem for attribute search, a custom GPT for outbound drafts, and an internal eval harness for sourcing quality") attract operators.

Second, price the band against the work, not the title. If you want a recruiter who can also act as a product user for your engineering team, that is a $140K to $180K job in the Bay Area, not a $90K job. DeepLearning.AI's band works because the brand subsidizes the comp. Yours probably does not.

Third, write the screen into the JD. Tell applicants exactly what artifact you want them to send. The recruiters who can produce it will self-select. The ones who cannot will save you a phone screen. And for the candidates you do source proactively (using Refolk or otherwise), give them a 48-hour artifact prompt before the first call. The hit rate on that filter is brutal in a good way.

The DeepLearning.AI posting is not just a job. It is a category being named in public, at a price point most of the market has not metabolized yet. The companies that read it carefully will hire well in Q1. The ones that copy-paste the title without copying the substance will spend Q2 explaining why their AI-forward recruiter quit for a $50K raise.

FAQ

Is $90K to $130K actually underpaying for an "AI Forward Recruiter" in Mountain View?

By the public benchmarks, yes, at least at the midpoint. Built In SF's average technical recruiter base is $132,248 and Indeed's California average is $125,695. Levels.fyi shows Bay Area recruiter total comp running from $138K to $215K. DeepLearning.AI's top of band lands at the bottom of those ranges. The Andrew Ng brand premium likely closes some of the gap, but talent leaders benchmarking against this JD should not use it as an excuse to underpay. The candidate pool that can credibly do the job is too small for that math to hold.

How is this different from a regular technical recruiter role?

The second paragraph of the JD is the difference. The recruiter is being asked to partner with engineering as a domain expert and shape DeepLearning.AI's portfolio of internal AI recruiting tools. That makes the role a hybrid of tactical recruiter and internal product user, modeled on the "AI Forward Deployed Engineer" pattern Ng has been amplifying in The Batch. A traditional technical recruiter manages a pipeline. An AI-forward recruiter manages a pipeline and a feedback loop into the tooling.

Where do I actually find recruiters who fit this profile?

Not by searching "AI recruiter," which returns mostly vendor staff and marketers. Search for artifacts: public writeups on building sourcing workflows, comparisons of Findem, Fetcher, Eightfold, and HireVue, GitHub repos with sourcing scripts, or Loom walkthroughs of evals. The current explicitly-tagged pool of US technical recruiters listing Generative AI as a skill is around 36 profiles, with employers like Apple, Figma, AWS, and Aledade. The undeclared pool is larger but requires a plain-English query rather than boolean to surface cleanly.

Will every recruiter JD look like this in 12 months?

The language will spread quickly at AI-native companies and frontier labs, probably within 90 days at AI Fund portfolio companies, Anthropic, OpenAI, Mistral, Scale, and Glean. It will spread more slowly at traditional enterprises where the recruiter is still scoped as a pipeline operator. The bigger shift is the skill profile underneath the title. Even recruiters whose job titles do not change will be expected to interpret AI insights, override bad recommendations, and apply human judgment AI cannot replicate. The title may lag. The work already shifted.