How should AI agents be used for recruitment in India?

Short answer: Use an agent for the first screen and for the chase: reading a pile, drafting a question, reminding a candidate. Then test the people you keep on the real work. An agent that only makes screening faster has not made hiring better.

I'm Vinay Pasricha. I run GoodSpace AI in Noida, and I wrote Organizational Frequency. Hiring is the work I know most closely. The temptation is to let an agent run the funnel. The funnel is not the hire. The hire is a person who can do this job, in this company, at this pace.

What the agent is for

A recruiter's week in a growing Indian company is full of repetitive motion. Reading resumes that do not fit. Sending the same three questions. Reminding a panel. Logging a status. An agent can do a first pass against criteria a person wrote, draft the questions, and keep the chase moving. That is worth doing. It gives the recruiter the hour the judgement actually needs.

It is not worth doing if the criteria are "culture fit" written as a mood. Write the criteria as work: a skill, a kind of problem, a pace. The agent can match those words. It cannot discover whether you meant them. If two hiring managers would disagree, settle it before the agent reads anyone.

The first screen should be generous enough that a non-standard career is not thrown out because the resume used the wrong noun. Check, on a sample, who the agent dropped. If the dropped pile contains people you would have wanted to talk to, the criteria are wrong. Fix the criteria. Do not blame the model and carry on.

What stays human

• The real work: a problem from last month, done in the room or on a call, judged by the person who owns the job.
• The environment: how this company decides, how fast it moves, how it treats a mistake. A resume cannot show that, and neither can an agent scoring a resume.
• The conversation a candidate deserves: an answer from a person when they ask what the job really is.
• The decision: a named person hires. An agent does not send the offer.

AI alone made screening faster, not better. That sentence is the whole caution. Speed without the real-work test fills the pipeline and empties it at the offer, or worse, fills the team with people who interviewed well against a rubric nobody checked.

There is a longer practical guide on this site, AI in recruitment in India, for where to start in a hiring team. This page is the narrower question: what an agent may touch. Keep the agent on the first screen and the chase. Keep the test of the work, and the decision, with people who will live with the hire.

A boundary that should be written down

The agent may read what the candidate sent. It may not invent experience, guess a community, or score a person on anything you would not say to their face. It may not hold their documents in a place you cannot name. If a vendor cannot tell you where the resumes go, do not connect the inbox.

Do not let the agent email a candidate with a promise about pay, role or timing. A person sends anything that commits the company. The agent can prepare the note. The recruiter can still change it, and should, whenever the case is not routine.

Use the free Risk Ladder before the agent is allowed to act, and Automate, Assist, Escalate to sort the tasks around it. The eight steps are in AI strategy for business. Collapse or Collapse is the longer argument: one person owns one result and runs agents to deliver it.

One thing to do this week

Write five criteria for the role you are hiring now, as work the person must be able to do. Run them past the hiring manager before any tool sees a resume. If the manager rewrites them, you just saved the agent from scaling a muddle.

Where this comes from

This answer draws on What it is not: just AI screening in Organizational Frequency.

Your company will have its own details. To ask a follow-up in your own words, try Ask Vinay's books.

Vinay Pasricha is the founder of GoodSpace AI and the author of six books, including Organizational Frequency.