The short answer
You can add AI to your ATS without ripping it out. There are four routes: switch on the AI features your ATS vendor already offers, add point tools from its integration marketplace, connect an AI agent that works across your ATS, calendar and messaging tools, or build your own on the ATS's API. Most teams combine the first and third. Whatever you choose, the ATS should stay the system of record: the AI reads from it, does the work, and writes notes, scores and stage changes back so recruiters keep one place to look. Start with one role family, set approval rules, and measure against a baseline.
Why add AI to your ATS instead of switching
An applicant tracking system holds years of candidate history, requisition approvals, reporting, compliance data and integrations with HR and payroll. Moving all of that to a new platform is slow and risky, and recruiters lose habits that work. Adding AI for the applicant tracking system you already use keeps all of that in place while taking repetitive work off the team.
It also avoids a trap: buying a new all-in-one tool for its AI and ending up with two ATSs, because some teams never fully move. If the AI works with the system you have, adoption is easier and reporting stays in one place.
The four ways to add AI to your ATS
A simple way to choose: if your ATS vendor's AI covers your slowest step, start there. If your slowest step is the hand-off between steps (sourcing to outreach to interview to scheduling), an agent across the stack usually helps more than adding another point tool.
What to connect, and why
AI in recruiting is only as useful as what it can reach. These are the connections that matter most, roughly in order:
- The ATS. Jobs, candidates, stages, notes and past outcomes. Reading past candidates powers rediscovery; writing back keeps one record.
- Calendars. Google Workspace or Microsoft 365, plus video tools like Zoom, so interviews can be booked without email ping-pong.
- Messaging. Email for candidates, and Slack or Microsoft Teams for the hiring team, so updates arrive where people already work.
- Job boards and sourcing tools. Inbound applicants and outside profiles, scored the same way as everyone else.
- The HRIS. Org structure, hiring managers, and internal candidates. Useful for routing approvals and for spotting internal mobility.
- Single sign-on. So access follows your identity rules and leavers lose access automatically.
You don't need all six on day one. The ATS and a calendar are enough for a first pilot.
What to ask about any ATS integration
Integration claims are easy to make and hard to compare. Ask every vendor the same questions, and ask to see the answer in a live account rather than a slide.
- What does it read? Jobs only, or candidates, notes, interview feedback and dispositions too?
- What does it write back? Notes, scores, attachments, tags, stage changes? Where do they appear for a recruiter?
- Who is the author? Do actions show up as the AI, a service user, or a named recruiter? You want to tell them apart.
- What can it change? Can it move stages or reject candidates, and can you switch that off per action?
- How are duplicates handled? If a sourced candidate already exists in the ATS, does it merge, link or create a second record?
- What happens when it breaks? Who gets alerted if the connection fails, and does work queue up or get lost?
- How long is setup? Ask for a typical range and what makes it longer, such as custom fields or approval chains.
A 30-day pilot plan
Keep the pilot small enough that one recruiter can review everything. A pilot that's too big to watch closely teaches you less.
Compliance checks before you switch it on
Adding AI to screening or ranking can bring legal duties, wherever the software lives. Three examples from primary sources (this is not legal advice):
- In New York City, an automated employment decision tool needs a bias audit within one year of use, a public summary of the results, and candidate notice 10 business days before use, according to the city's DCWP. Our NYC Local Law 144 guide explains when a tool counts.
- In Illinois, since January 1, 2026, employers must give notice when they use AI in recruitment and hiring decisions, and may not use AI that has a discriminatory effect or uses zip codes as a proxy for protected classes.
- In the EU, AI used "to analyse and filter job applications, and to evaluate candidates" is high-risk under the AI Act, with those duties applying from December 2, 2027 according to the consolidated Article 113.
Ask each vendor what has been independently audited and which features the audit covers. An audit of one feature doesn't extend to the rest of the product. For permission design, see our AI recruiting guardrails guide.
How to tell it's working
After the pilot, compare against the baseline. Good signs: time to first interview drops, recruiters spend fewer hours per hire on logistics, hiring managers accept a higher share of the shortlist, and pass-through rates by group stay level or improve. Warning signs: recruiters re-doing the AI's work, duplicate records in the ATS, or candidates confused about who they're talking to. Fix the cause before you expand to more roles.
Adding NTRVSTA to your ATS
We build NTRVSTA, which takes option 3. Our agent Carolina works inside the systems you already use: ATSs such as Greenhouse, Lever and Workday, HRIS and payroll, Google Workspace or Microsoft 365 calendars, and Slack or Teams, among 70 connected systems, plus a REST API and SFTP. It sources, runs outreach, scores and ranks resumes with its reasoning, runs live adaptive AI interviews and schedules, and each action is set to suggest, ask first or auto. Setup usually takes a few days, depending on customization. Starter is $0 for one open role with 10 AI interviews; Team is $490 a month billed annually. Full details are on our pricing page.
When it isn't the right choice: if you want one tool to become your ATS and system of record, an ATS with built-in AI fits better, and for high-volume hourly hiring built around text-to-apply, Workstream is designed as that kind of all-in-one platform. NTRVSTA is also newer and smaller than the incumbents, with no published case studies yet. To compare more options, see the best recruiting automation software and the best AI recruiting software for enterprise.
Questions
Can I add AI to my ATS without replacing it?
Yes. Most AI recruiting tools are built to work alongside an ATS: they read jobs and candidates, do the work, and write notes, scores and stage changes back. Your ATS stays the system of record for reporting and compliance.
Is my ATS's built-in AI enough?
It can be, if it covers your slowest step. Built-in features are often strongest inside the ATS itself, such as summarizing or matching. If your bottleneck is sourcing, outreach, first-round interviews or scheduling across calendars, a connected AI tool or agent may cover more of it.
What data does an AI tool need from my ATS?
At minimum, open jobs and their descriptions, plus candidates and their current stage. For better results, add past candidates with outcomes, interview feedback and scorecards. Grant only the access the tool needs, and decide separately what it may write back.
How long does it take to connect AI to an ATS?
Connecting the ATS and calendar usually takes days, not months, depending on custom fields and approval rules. Plan another two to three weeks to calibrate scoring and approval settings on a pilot role family before rolling out more widely.
Bring your ATS and one open role, and we'll show you what Carolina would do with it: book a demo. Or try a sample AI interview on your own.
