AI-native versus AI added on
| AI added on | AI-native | |
|---|---|---|
| Unit of work | A feature you click: summarize, rewrite, suggest | A task that gets done: screen these 140 applicants, book the panel |
| Screening | Keyword match or a score with no explanation | Each requirement of a written brief checked, with the evidence quoted |
| Automations | Rule builder configured by an admin | A sentence in plain English that becomes a readable workflow |
| Agents | Limited API, perhaps a read-only connector | Every action is a tool; agents are actors with their own permissions |
| Control | On or off | Approvals for sensitive actions, dry runs, an audit log per actor |
| Pricing | AI credits or an add-on | Part of the plan |
How AI resume screening works
Older applicant tracking systems matched keywords: a resume without the exact phrase ranked low. Language models read a resume the way a recruiter does. In Evander you write a brief for the role with must-haves, either/or profiles and bonuses. Every application is checked against each requirement, and each check links to the evidence in the resume, so you can see why someone scored 91 or 34 and correct the brief when it is wrong.
Which products score applicants with AI, how each charges for it and how to test whether a score explains itself: AI resume screening: which ATS shows why.
Two rules make this safe to use. The model judges only job-relevant evidence and is instructed never to consider age, gender, ethnicity, nationality, religion or similar characteristics. And the score only ranks: a person decides who moves forward and who is rejected.
What the law expects
- EU AI Act. AI systems used to recruit or select people, including those that filter applications and evaluate candidates, are classed as high-risk (Annex III). Human oversight, transparency and record keeping are central duties.
- GDPR, Article 22. People have the right not to be subject to a decision based solely on automated processing that significantly affects them. A person making the decision, with the AI's reasoning in front of them, is the standard answer.
- New York City Local Law 144. Employers using automated employment decision tools for candidates in the city need an annual bias audit and must notify candidates.
This is general information, not legal advice. In practice: keep a person in every decision, keep the reasoning for each score, tell candidates that AI assists the review, and delete candidate data when your retention period ends.
Seven questions that reveal an AI-native ATS
- Can I create a complete job, with stages and scorecards, from one sentence?
- Is every applicant screened, or only the ones I click on?
- Does each score show the evidence per requirement?
- Can anyone on the team write an automation in plain language?
- Is there an official MCP server, and does it cover every action or a subset?
- Which actions can an agent take without a person's approval?
- Is AI part of the plan, or billed by credits?
Who offers what
Evander and Kula describe themselves as AI-native. Ashby, Gem, Dover, Greenhouse and Workable have added AI review, agents and MCP servers to their platforms. How each one charges for AI:
| Product | AI | Starting price |
|---|---|---|
| Evander | Every applicant scored against a written brief, with the evidence for each requirement. 500 scorings a month on Startup | Free, then from $79/month |
| Kula | AI scoring, notetaker, writing assistant and scheduling included with unlimited usage | From $5,120/year |
| Ashby | AI-assisted application review, metered by credits (1,500 a month on Foundations); AI Notetaker is a paid add-on | From $300/month |
| Gem | Sourcing agents, ranking and AI application review on the AI-powered seat | Custom; startup program |
| Dover | AI job descriptions and interview questions on Free; AI resume review (applicant scoring) and AI notetaker on Premium | Free, Premium $199/month |
| Greenhouse | AI notetaker and suggestions for scorecards, interview questions and job posts; sourcing automation from Plus | Quote only |
| Workable | Screening, sourcing and engagement agents metered by credits (3,000 included; 5,000 more for $600) | From $299/month |
| Breezy HR | Breezy Intelligence is an optional add-on billed in credits, from $30 per 100,000 | From $157/month |
| Manatal | AI recommendations with scoring, candidate enrichment and AI Copilot on all plans; AI Interviewer is an add-on | From $15/user/month |
Questions
What is an AI-native ATS?
An applicant tracking system designed from the start so that AI does the routine work (writing job posts, screening every applicant, drafting replies, scheduling, chasing feedback) while people make the decisions. Every action in the product is also available to AI agents with permissions, approvals and an audit log.
How is AI resume screening different from keyword matching?
Keyword matching ranks a resume by whether exact words appear. AI screening reads the resume against each requirement of a written brief, the way a recruiter would, and shows the evidence for each check, so a strong candidate who uses different words is not missed.
Does an ATS automatically reject resumes?
Most do not, and under GDPR a rejection based solely on automated processing is restricted. In Evander AI ranks and suggests. A rejection proposed by an assistant or an automation waits for a person's approval; only a screening rule a person wrote, such as a missing must-have, rejects on its own, and each such rejection is logged and can be undone.
Is AI screening of job applicants legal?
Yes, with conditions. The EU AI Act classes recruiting AI as high-risk and requires human oversight and transparency; GDPR Article 22 restricts decisions based solely on automated processing; New York City requires bias audits for automated employment decision tools. Keep a person in every decision and keep the reasoning.
Which ATS products are AI-native?
Evander and Kula describe themselves as AI-native. Ashby, Gem, Dover, Greenhouse and Workable have added AI review, agents and MCP servers to their platforms.
What is an AI-native ATS, and which products are?
An AI-native ATS is built so that AI does the routine hiring work (job posts, screening every applicant, replies, scheduling) and every action is also open to AI agents with permissions and approvals, while people decide. Evander and Kula describe themselves as AI-native; Ashby, Gem, Dover, Greenhouse and Workable added AI features and MCP servers to established products.