Indeed Sourcing Assistant: 370M Profiles, Wrong Ones for Engineers

If you hire senior engineers, you now have to pick an AI sourcing agent, and the shortlist most vendors are pushing is wrong. The candidates you want are not applying to jobs, which means the size of an agent's index matters less than what is actually in it.

On June 15, 2026, Indeed shipped Sourcing Assistant on its Professional and Enterprise Smart Sourcing tiers: a natural-language, autonomous agent that "sources while you sleep" across 370 million profiles, drafts outreach, integrates with ATSs, and posts a headline 2.9x hire-rate lift. It landed the same quarter LinkedIn Talent Solutions' agentic products crossed a $450M ARR run rate in Microsoft's FY26 Q3. Both are real products. Neither is built for the role you are probably trying to fill.

The headline numbers, in one place

Indeed's Sourcing Assistant monitors 370M "Sourceable Profiles," claims applicants it surfaces are 2.9x more likely to be hired than applicants from other sources, and says employers saved 7 hours per week during beta. Those numbers describe an apply-driven graph, not a passive-engineer graph.

Here is the comparable data every engineering leader evaluating AI recruiting agents in 2026 should have on one page.

The uncomfortable ratio: the entire enumerable US population of senior software engineers, ~246K in Refolk's index, is 0.066% of Indeed's 370M sourceable pool. For technical roles, Sourcing Assistant is scanning a haystack roughly 1,500 times larger than the actual signal.

Why 2.9x of Indeed's baseline is still a low number

Indeed's 2.9x hire-rate lift is measured against Indeed's own applicant baseline, not against LinkedIn, GitHub, or a human sourcer, and that baseline is unusually low. By Indeed's own reporting, fewer than 5% of Indeed applications lead to a callback, and 97 to 98% of applicants do not meet basic qualifications.

Multiply 2.9 by a very small number and you still get a small number. That is the mechanism, not a vibe. The agent is not smarter than the pool it reads.

The sample behind the headline is also thin:

• 225 jobs in the hire-rate study
• 17 self-reporting beta employers behind the "7 hours saved per week" claim
• No disclosed role mix, which means engineering subgroups are almost certainly not represented in either sample

0.066%

The 370M vs 665M gap is the real story

Indeed has 665 million Job Seeker Profiles globally but only 370 million are "Sourceable." Roughly 44% of Indeed's own graph is invisible to Sourcing Assistant, likely opted-out or dormant, and no amount of agent autonomy fixes a data-consent ceiling.

For high-volume roles (warehouse, retail, entry customer support) the remaining 370M is more than enough. For a Staff Rust engineer at Cloudflare who last updated a resume in 2019 because she has never needed one, none of Indeed's tiers matter. She was never in the 665M. She is definitely not in the 370M.

This is what recruiters have been saying for years, and it survives the agent era intact:

A senior staff engineer who's happy at her current job will never appear in your Indeed funnel. She isn't applying to anything.

Where senior engineers actually work in 2026

Senior US engineers concentrate at a small set of employers that do not rely on Indeed as a top-of-funnel channel, which is why sourcing them requires a graph agent or an open-web agent, not a job-board agent.

From Refolk's index, the top employers of Senior/Staff/Principal software engineers include:

• Google
• NVIDIA
• Datadog
• Shopify
• Omada Health

For the ~55,236 US senior engineers with Rust, Kubernetes, or Go on-profile, the concentration shifts toward infrastructure and security:

• Cloudflare
• Palo Alto Networks
• Cisco
• Datadog
• Modal
• ServiceNow
• NVIDIA

None of these companies fill senior roles primarily by ingesting Indeed applications. Their recruiters live in LinkedIn Recruiter, GitHub, ex-employer alumni graphs, and referral chains. An autonomous agent scanning Indeed's 370M pool is optimizing for a top-of-funnel that these employers do not share.

Indeed vs LinkedIn Hiring Assistant vs open-web agents

The right 2026 comparison is not Indeed vs LinkedIn. It is job-board agent vs graph agent vs open-web agent, and each reads a fundamentally different data source.

• Job-board agent (Indeed Sourcing Assistant). Reads a resume graph of ~370M sourceable profiles. Best fit: apply-driven, high-volume roles where the candidate is actively searching.
• Graph agent (LinkedIn Hiring Assistant). Reads a ~1B-member professional graph. Charter customers include AMD, Canva, Siemens, and Zurich Insurance. Best fit: mid-to-senior professional roles where the candidate maintains a public profile.
• Open-web agent (Refolk, and to a lesser extent hireEZ, SeekOut, AmazingHiring). Reads LinkedIn plus GitHub plus the open web. Best fit: senior technical roles where the strongest signal is code, conference talks, and employment history, not a resume upload.

For a Senior Rust role in Berlin (n = 335 in Refolk's index, ~85% concentrated in Berlin and Munich, top employers Siemens, DeepL, IONOS, Matter Labs, IOHK), only the third category actually enumerates the market. Sourcing Assistant will happily draft outreach at 3am to whoever in its 370M pool self-identifies as "Rust developer" in Germany. That set overlaps the real 335 by a rounding error.

Note the overlap in the LinkedIn charter list and the Germany Rust list: Siemens appears in both. The same employer is being sourced from through a graph agent and being sourced into through an open-web agent. The two tools are not substitutes.

What "sources while you sleep" does to a senior-engineer pipeline

Autonomous agents amplify the base rate of the pool they read from, so on a senior engineering role, "sources while you sleep" burns outreach on wrong-fit candidates 24 hours a day instead of 8. That is the failure mode nobody is naming.

The mechanism is simple:

1. Sourcing Assistant sees a job description for "Staff Backend Engineer, Distributed Systems."
2. It ranks the 370M pool for token overlap and recency of application activity.
3. It drafts and sends outreach to the top N, then repeats every night as new applicants enter.
4. The top N is dominated by people who apply to a lot of jobs, which correlates negatively with the profile of a Staff engineer at Cloudflare.

The output feels productive (dashboards fill up, response rates look fine against Indeed's own baseline) while the actual pool of ~55,236 US Senior Rust/K8s/Go engineers goes untouched. This is where an open-web tool like Refolk changes the shape of the funnel: describe the person in plain English ("Senior backend engineer with production Rust and Kubernetes, currently at a Series C infra company, based in the US") and get a ranked shortlist against the enumerable pool, not against Indeed's applicant queue.

An AI recruiter comparison for engineering leaders in 2026

For engineering-heavy hiring in 2026, the right stack depends on role seniority and how public the candidate signal is, not on which vendor shipped an agent most recently. Here is a practical split.

High-volume, apply-driven roles

• Buy: Indeed Smart Sourcing Professional plus Sourcing Assistant.
• Why: The 370M pool is the point. Apply-intent is a feature, not a bug. The 2.9x lift applies to exactly this segment.

Mid-level engineers with public LinkedIn profiles

• Buy: LinkedIn Recruiter plus Hiring Assistant.
• Why: ~1B-member graph, agentic ARR already at $450M+ in Microsoft's FY26 Q3, and enterprise-tech charter customers (AMD, Canva, Siemens, Zurich Insurance) that share your candidate pool.

Senior/Staff engineers, specialized stacks, or small country pools

• Buy: An open-web sourcing tool, and treat LinkedIn Recruiter as a verification layer.
• Why: The pool is small enough to enumerate (245,899 US Senior+ SWEs; 335 Senior Rust engineers in Germany), which means precision beats volume. Refolk is built for exactly this: ask in plain English, get the right people across GitHub, LinkedIn, and the open web, and never see the wrong 369M profiles Sourcing Assistant would have queued for you.

Mixed role types, 5 to 20 recruiters

• Buy: One job-board agent (Indeed Smart Sourcing Professional) plus one AI-native platform for specialized searches (hireEZ, SeekOut, or Refolk for engineering).
• Why: The job-board tier handles the fat middle of the funnel. The AI-native layer handles the top 10% of roles that produce 80% of the hiring pain.

The uncomfortable takeaway

Indeed Sourcing Assistant is a good product for the job it was designed for: closing high-volume roles faster on an apply-driven graph. It is a weak tool for sourcing engineers because the graph it reads does not contain them, and no agent, however autonomous, can source people who are not in the index.

The 2026 AI recruiter comparison worth running is not Indeed vs LinkedIn Hiring Assistant. It is: which agent's underlying graph actually contains the 246K US senior engineers, the 55K Rust/K8s/Go specialists, or the 335 Berlin Rust engineers you are trying to reach? Answer that first. The agent layer is downstream.

FAQ

Is Indeed Sourcing Assistant worth it for technical roles?

For junior and mid-level roles where candidates actively apply, yes. The 2.9x hire-rate lift and 7-hour weekly time savings are credible for that segment, with the caveat that both come from small samples (225 jobs and 17 self-reporting employers). For senior, staff, or principal engineers, the answer is no: those candidates are largely absent from Indeed's 370M sourceable pool, and automating outreach against the wrong pool produces more wrong outreach, not better hires. A useful heuristic: if your target hire has updated a resume in the last 18 months, Indeed can help. If they have not, buy something that reads GitHub and the open web.

How does Indeed compare to LinkedIn Hiring Assistant for engineering hiring?

LinkedIn Hiring Assistant wins on graph size (~1B members vs Indeed's 370M sourceable), on enterprise-tech distribution (charter customers AMD, Canva, Siemens, Zurich Insurance), and on ARR momentum ($450M+ agentic-product run rate in Microsoft's FY26 Q3). Indeed's tools win on apply-driven, high-volume roles. Neither reads GitHub, which is why teams hiring senior infrastructure or ML engineers usually pair one of them with an open-web tool.

Why does Refolk's index show only 245,899 US senior software engineers?

Because that is roughly the enumerable size of the market at Senior/Staff/Principal seniority in the US. It feels small next to Indeed's 370M number because it is measuring a different thing: real, currently employed senior engineers at named companies (Google, NVIDIA, Datadog, Shopify) versus every profile Indeed can legally show a recruiter. Precision at that pool size is what makes plain-English sourcing viable. You can review the whole ranked shortlist.

What should a founder buying an AI recruiting agent in 2026 actually do?

Split the decision by role type. For sales, support, and operations roles, put Indeed Sourcing Assistant on the Professional or Enterprise tier and let it run overnight. For mid-level engineers, keep LinkedIn Recruiter with Hiring Assistant on. For any senior engineering role, staff-plus role, or specialized stack (Rust, Kubernetes, kernels, ML infra), add an open-web sourcing tool and ask for candidates in plain English. That third layer is where the 246K real senior engineers actually live, and it is the layer job-board agents structurally cannot reach.