Most companies have more hiring tooling than they have ever had, and the hiring outcomes are not noticeably better. Time-to-fill is climbing. Quality of hire is roughly the same. Recruiters are exhausted. Candidates are angry. Founders are confused, because they bought the software the demos said they needed, and the system still does not produce reliable results. Hiring is broken, and not for the reason most people think.
The problem is not effort. It is that almost every applicant tracking system is optimizing the wrong layer of the funnel.
What is actually broken about hiring
Hiring is broken at the screening layer, not the workflow layer. Screening is where you decide who gets read carefully and who does not. Everything downstream, the interviews, the debriefs, the offer, sits on top of that one decision. According to the SHRM 2025 Recruiting Benchmarking Report, the average US time-to-fill is around 44 days, and the long pole is the screening-to-shortlist phase. The Harvard Business School study on hidden workers found 27 million qualified Americans get rejected by automated screening systems for reasons unrelated to whether they could do the job. That is not a workflow problem. That is a screening model problem. The ATS gets candidates into stages and emails them updates competently. It just rank-orders the wrong people at the top, and no amount of email automation downstream fixes that.
So when a founder says “my hiring is broken,” what they almost always mean is “my shortlist is bad.” The shortlist is the output of screening. Fix the shortlist and the rest of the funnel calms down.
Why most ATS optimize the wrong layer
Almost every applicant tracking system on the market was designed for a buyer that is not you. The original buyer was a 5,000-person enterprise HR team whose hardest problem was workflow visibility. Twenty recruiters running ten reqs each. Compliance reporting for EEO and OFCCP. Audit trails for every status change. Those teams already had screening capacity (humans reading resumes), so the ATS focused on giving them dashboards. Screening got bolted on as a Boolean keyword filter.
That architecture is what almost every modern ATS still ships, with cosmetic upgrades. Workable starts at $149/mo for a small team and scales with employee headcount, not hire count. Greenhouse contracts run roughly $12,000+ per year for under-100-employee teams, per buyer-reported data on PriceLevel. You are paying for workflow, and the screening layer is whatever the keyword filter happens to do that week. That filter is what is wrong.
Three specific things go wrong on the keyword layer. It rewards keyword stuffing, which is documented widely, including by Jobscan. It rejects candidates for word choice (“led” instead of “managed”) even when their actual experience is stronger. And it has no concept of impact, so a senior engineer who shipped a real-time fraud detection system loses to a junior engineer whose resume mentions Python four times.
The startup founder is the worst-served buyer
If the original ATS buyer was the enterprise HR team, the worst-served buyer in this market is the startup founder. The reason is structural, not personal.
A founder hiring without a recruiter does not need EEO dashboards or twenty-stage workflow customization. They need to read 150 resumes for a role and find the eight worth talking to. That is a screening-quality problem, not a workflow problem. But the products on the shelf are workflow products with screening as a feature, and they cost as much as a junior engineer’s salary. Workable’s $149/mo Starter scales by employee count, which means a 30-person team that fills three roles a year pays the same as a 30-person team that fills thirty. Greenhouse buyers report contracts in the $10,000+ range for sub-100 employee teams.
So the founder buys the product, gets a shortlist that includes the keyword stuffers and excludes the strong candidates, and assumes hiring is just hard. It is hard. It is also being made harder by the wrong layer of automation.
What we learned at Amazon about ranking systems
Before CurriculoATS, our founder Dev spent years at Amazon working on search and recommendation systems. The single biggest lesson from those systems applies almost word-for-word to hiring: when the input is messy and the user has limited attention, the only thing that matters is whether the top results are right. Position 1 through position 10 decides the outcome. Position 11 onward might as well not exist.
Hiring works the same way. The recruiter or founder will read the top ten or fifteen ranked resumes for a role. Whether the model thinks candidate 73 is slightly better than candidate 84 is irrelevant; nobody will ever look at either. The only question is whether the right people are in the top ten. That is the metric the screening layer needs to be good at, and almost no legacy ATS measures itself this way.
CurriculoATS is built around that single observation. The model reads resumes for measurable outcomes, revenue generated, teams led, systems shipped, problems solved, and produces a 0–100 composite score with a written reasoning paragraph for every candidate. The paragraph is the thing that makes the system trustable. You can read it, agree with it, override it. That is what an outcome-based ATS looks like in practice, and it is structurally different from a Boolean keyword filter sitting under a polished dashboard. See the impact scoring page for the full mechanism.
The compliance turn nobody is talking about yet
Hiring tools are quietly entering the same regulatory phase financial-services models entered around 2018. Two pieces of law set the bar: NYC Local Law 144 (effective July 2023) requires a published bias audit and candidate notice for any automated employment decision tool used on a NYC-resident candidate, and the EU AI Act classifies hiring AI as high-risk under Annex III, with the relevant obligations enforceable from August 2026. Both regimes share one technical assumption: the model’s reasoning has to be documentable. A 73% match score with no explanation cannot be audited; a paragraph that names the candidate’s specific outcomes and how they map to the role can. Most legacy ATS have not retooled for this. Their AI features were bolted onto keyword pipelines specifically because that was cheaper than rebuilding the ranker, and the documentation burden lands on whichever customer is unlucky enough to get audited first. Founders shopping ATS vendors in 2026 should ask three procurement questions: where is the bias audit summary published, what does the candidate notice template say, and can the vendor produce the reasoning paragraph for any specific score on demand. Vendors who answer all three crisply have done the engineering work. Vendors who deflect have not, and that gap is going to widen as Annex III enforcement starts in late 2026. The screening-layer fix doubles as a compliance fix, which is rare in regulatory transitions and worth treating as a feature rather than a footnote.
What a founder changes this week
If you are a founder and your hiring feels broken, the highest-leverage moves are not what you might guess. They are not posting on more boards or running more outreach.
- Stop ranking candidates by keyword density. If your current ATS sorts the inbox by token match, every candidate who has applied to more than a handful of jobs has gamed it.
- Rewrite the JD as outcomes. “In the first six months, this person will ship X” beats “responsible for X.” The screening model has more to work with.
- Cap the shortlist at twelve. Past twelve, attention falls off and the choice gets random.
- Look at the candidates the system rejected. If your last good hire would have been rejected by your current ATS, the ATS is the problem.
- Switch to flat pricing for hire volume, not headcount. Per-seat or per-employee pricing punishes you for being the kind of company you want to be.
Founder questions
What is the single biggest reason hiring is broken?
The screening layer in legacy ATS uses keyword matching, which rewards candidates who learned the rules and rejects strong candidates who wrote honest resumes. Every other hiring problem (long time-to-fill, weak shortlists, expensive interview loops) sits downstream of this one decision. Fix screening and the funnel calms down. Leave it broken and no other tool helps.
Is the problem the ATS or the people using it?
Both, but the ATS is the leverage point. People using a bad screening model can compensate by reading every resume manually, which is what most founders end up doing. That is not a process problem; it is a tooling problem the founder is patching with their time. Replace the model and the founder gets the time back.
Why don’t legacy ATS just upgrade to better AI?
Three reasons. The buyer (enterprise HR) does not push for it. The architecture (workflow-first) makes it expensive engineering. And the marketing (“we have AI now”) works fine without doing the underlying work. The incentives line up against real change.
How does CurriculoATS handle compliance for AI screening?
CurriculoATS produces a written reasoning paragraph for every candidate, which is exactly what regulators are converging on. NYC Local Law 144 requires bias audits and candidate notice. The EU AI Act lists hiring AI as high-risk. A model with documented reasoning makes both regimes far easier to comply with than a black-box keyword filter does.
How long does it take to see screening quality improve after switching ATS?
Faster than most founders expect. Within the first week, the top-10 ranking on an open role changes meaningfully because the ranker is reading for outcomes instead of token density. Within the second week, hiring managers report fewer interview slots burned on candidates who looked good on paper but were keyword-stuffers. Within a month, time-to-fill on the role compresses by 20-40% in the data we see, because the shortlist is correctly ordered and the founder is not re-reading 200 resumes manually.
What does CurriculoATS cost?
Free Starter (1 active job, unlimited team members), Pro at $100/mo flat (currently $50/mo early bird, indefinite), Enterprise custom. There are no per-seat fees. Add the entire team without an upcharge. See pricing for details.
What to do next
If your hiring funnel is the bottleneck for company growth right now, the fastest thing you can do this week is move the screening layer off keyword matching. Read our what is an ATS page for the canonical breakdown of how modern systems differ, then look at compare for how CurriculoATS stacks up against Greenhouse, Lever, Workable, and Ashby on the screening layer specifically. Free Starter is enough to test it on one open role.