99.8% Bought AI Agents. 90% Missed Goal. The 1:77 Ratio Explains Why.

GoodTime's 2026 Hiring Insights Report landed with a number that should have embarrassed every AI vendor at HR Tech: 99.8% of TA teams use, pilot, or plan to use AI agents, and 90% of U.S. companies still missed their hiring goals. The industry spent a week arguing about whether AI recruiting is broken. It is not broken. It was pointed at the wrong part of the funnel.

The headline number everyone is quoting

99.8% AI adoption produced a 10% goal-attainment rate because AI agents were bought to solve a sourcing problem most teams do not actually have, while the coordination problem that eats 38% of a recruiter's week went largely untouched. That is the through-line of the GoodTime 2026 report, and it is the reason AI recruiting agents ROI keeps disappointing the CFOs who signed the POs.

The GoodTime Hiring Insights Report surveyed more than 500 U.S. TA leaders at companies with over 1,000 employees in November 2025. The findings that matter:

• 90% of U.S. companies missed hiring goals, with 1 in 3 missing by a wide margin.
• 60% of organizations saw time-to-hire increase in 2025. Only 1 in 9 reduced it.
• Recruiters spend 38% of their week scheduling interviews.
• Tech-hiring goal attainment fell from 58% in 2023 to 50% in 2024 and held at 50% in 2025.
• Fraudulent or AI-generated candidates are now ranked the #1 threat for 2026, above "lack of qualified talent."

99.8%

Adoption at 99.8% is not a competitive edge. It is table stakes. Anything that universal is, by definition, a lagging indicator of buying behavior, not a leading indicator of performance. Which raises the real question: what were the 10% who hit their goals actually doing differently?

Why AI sourcing agents did not move goal attainment

AI sourcing tools are not working at the aggregate level because most funnels never had the sourcing discipline for an agent to amplify, and generating more candidates into a coordination-starved pipeline just increases the scheduling tax that was already the biggest drain on recruiter time.

Three mechanisms explain the gap between spend and results:

1. You cannot automate a function that does not exist. An AI agent is a force multiplier. Multiply zero sourcers by any coefficient and you get zero pipeline. The generalist recruiter running req intake, screening, scheduling, and offers does not suddenly become a sourcer because a Chrome extension shipped.
2. More candidates make the scheduling problem worse. If 38% of the week is already scheduling, dumping more top-of-funnel volume into that same coordination layer does not shrink time-to-hire. It stretches it. That is exactly what 60% of respondents reported.
3. Inbound-oriented AI amplifies fraud. With AI-generated candidates now the #1 threat, resume-parsing and inbound-scoring agents are ingesting a poisoned corpus. The signal-to-noise ratio at the top of the funnel is getting worse, not better.

The AI recruitment market was valued at $704.54 million in 2025 and is projected to hit $1.12 billion by 2032, with 54% of firms planning 40%+ increases in AI recruiting spend. Goal attainment, over the same window, fell. Dollars are chasing tools that measurably have not moved the KPI.

The 1:77 ratio nobody is talking about

The U.S. has roughly one dedicated sourcer for every 77 generalist recruiters, and that ratio, not AI adoption, is the constraint on hiring goal attainment in 2026. In Refolk's index of professional profiles, the numbers are stark.

Read the last two rows together. There are more Heads of TA in the U.S. than there are people whose full-time job is finding candidates. The average enterprise TA org has bought an AI agent for a function it never staffed. That is not an AI failure. That is an org design failure with an AI label on it.

1 : 77

Where the sourcers actually are

Dedicated sourcing talent concentrates in the San Francisco Bay Area, followed by Austin and San Diego. The top employers of dedicated sourcers in Refolk's index skew AI-native and platform: Pinterest, Rippling, MongoDB, Verkada, Anthropic, Zoox, EvolutionIQ. These companies are not winning hiring because they have better agents. They are winning because they staffed the top of the funnel with humans who know how to run one, then pointed automation at the parts a human should not be doing.

For a non-coastal enterprise, the implication is grim: you literally cannot hire the sourcer you would need to feed the agent you just bought. The realistic move is to buy the sourcing capability as a service. That is the exact gap Refolk closes: you describe the person in plain English and get a ranked shortlist back across GitHub, LinkedIn, and the open web, without needing to headcount a sourcing team you cannot recruit for.

What the top 10% actually do differently

The 10% of TA teams hitting their hiring goals in 2025 restructured the work around AI rather than layering AI onto the existing work, and their edge shows up in three measurable operational habits from the GoodTime data.

Top performers were:

• 74% more likely to keep headcount flat while reorganizing roles, moving generalists into specialized lanes (sourcing, coordination, closing) instead of everyone doing everything.
• 58% more likely to use a centralized platform for texting candidates, collapsing the channel sprawl that fragments recruiter attention.
• 20% more likely to use AI agents specifically for interview scheduling, the exact task consuming 38% of recruiter time.

Notably absent from that list: "used AI for sourcing." The winners are not out-sourcing the competition with better agents. They are freeing up recruiter hours by automating coordination, then spending those recovered hours on judgment work: outbound to specific named humans, calibration with hiring managers, and closing.

Ahryun Moon, CEO of GoodTime, framed it this way in the report:

The teams outperforming everyone else have restructured their organizations around an AI-enabled future, where automation handles coordination so humans keep their focus on judgment.

Translated into recruiter time allocation: the goal is not to replace the recruiter with an agent. It is to reclaim the 38% of the week they spend on Doodle polls and reschedule threads, then reinvest that time into the one activity no agent has cracked, which is convincing a specific senior engineer at a specific competitor to take a first call.

The market that is actually hiring for this

The demand signal for people who can run this redesigned funnel - sourcers, TA ops, coordinators - is real, and it clusters in the same cities where the top performers already sit.

If you are a TA leader trying to move goal attainment in 2026, the hiring plan implied by the GoodTime data is not "add three generalist recruiters." It is "add one sourcer, one coordinator, and a scheduling agent, and let the recruiters you already have run reqs."

Rebuild the top of the funnel before adding another agent

The correct sequence is to fix sourcing intent and coordination first, then layer AI on the specific tasks the top-10% data says actually move the needle. Doing it in the other order is how you end up in the 90%.

A concrete rebuild, in order:

1. Audit where recruiter hours actually go. If scheduling is above 30%, buy a coordination agent before you buy anything else. This is where the top 10% got their 20% edge.
2. Name your ICP in plain English, per req. Not a Boolean string. A sentence a smart human could act on. "Backend engineer, 6+ years, has run Postgres at scale outside FAANG, in or willing to move to Austin." Vague ICPs produce vague pipelines regardless of tool.
3. Decide outbound vs. inbound per req. With fraudulent candidates ranked the #1 threat for 2026, inbound-heavy funnels for senior roles are a losing bet. Outbound to verified humans - verified via GitHub commits, real employment history, and open-web signals - is the defensible mode. This is where Refolk fits: you describe the person, Refolk returns real, ranked candidates across GitHub, LinkedIn, and the open web, and your recruiters spend their reclaimed hours writing first messages instead of parsing fraud.
4. Only then, add sourcing automation. An agent on top of a well-defined ICP and a coordinated funnel is a multiplier. An agent on top of neither is a burn rate.
5. Measure per-req time-to-first-conversation, not per-req applications. Applications are the metric AI vendors love. First-conversation-with-a-qualified-human is the metric that predicts goal attainment.

The uncomfortable truth in the GoodTime numbers is that most TA orgs bought AI to avoid the harder conversation about how their function is structured. AI agents are cheaper than reorgs, faster to procure than a new sourcing hire, and easier to defend in a board deck. They also, on the current evidence, do not move goal attainment on their own.

The 10% figured that out. Everyone else has a renewal coming up.

FAQ

Why did 90% of companies miss hiring goals despite near-universal AI adoption?

Because most teams bought AI sourcing agents to solve a problem that was not their actual bottleneck. GoodTime's 2026 report shows recruiters still spend 38% of their week scheduling interviews, and 60% of orgs saw time-to-hire increase in 2025. Adding candidate-generation agents on top of a coordination-starved funnel produced more volume, not more hires. The 10% who hit goals invested in coordination automation and role specialization first, then layered sourcing tools on a funnel that could actually absorb them.

Is AI recruiting agents ROI actually negative, or just misattributed?

Mostly misattributed. When 99.8% of teams have adopted a category, that category cannot be the source of competitive advantage; it is table stakes. The ROI question worth asking is per-agent and per-task: coordination and scheduling agents show up in the top-10% differentiators from GoodTime's data. Sourcing agents deployed without a defined ICP or a sourcing function do not. The negative signal is really "we bought the wrong AI for our funnel," not "AI does not work in recruiting."

What does the GoodTime hiring insights report say the top 10% do differently?

Three things measurably separate them: they are 74% more likely to keep headcount flat and reorganize roles into specialized lanes, 58% more likely to run a centralized candidate-texting platform, and 20% more likely to use AI agents for interview scheduling specifically. Notably absent from the list is "used AI for sourcing." The winners restructured the work, then automated the coordination layer, then reinvested recruiter hours into outbound and closing.

How should I fix recruiter time allocation if I cannot hire a sourcer?

Buy the sourcing capability as a service and reclaim the scheduling hours in parallel. Dedicated sourcers concentrate in the Bay Area, Austin, and San Diego per Refolk's index, so most enterprises cannot realistically hire one locally. A tool that lets you describe the person you want in plain English and returns a ranked shortlist across GitHub, LinkedIn, and the open web removes the need to staff a full sourcing bench. Pair that with a coordination agent for scheduling and you have replicated the top-10% operating model without a headcount fight.