AI Recruiting Academy › Module 6: AI across the recruiting desk
The short answer
At the end of the process, AI speeds up two jobs. Automated reference checking sends short surveys to references, collects answers within a day or two and flags references that look fake. At the offer stage, AI helps you check pay against market data, draft the offer letter and get it signed quickly.
Two rules keep you safe. Reference checks run by an outside company can fall under background check law, and AI must never set pay using age, sex, race or any other protected trait.
Why does the end of the process matter so much?
Because this is where hires are won or lost after weeks of work. Gem’s 2026 benchmarks found 82% of candidates who reach the offer stage accept. But candidates keep looking. Gartner found only 48% of candidates accepted their most recent job offer in late 2025, and a 2023 Gartner survey reported by SHRM found half of candidates had accepted an offer and then backed out for another employer.
Speed helps. Ashby’s offer data found candidates who accepted did so in about two to three days, while those who declined took about six. Every day of slow references and slow paperwork gives a competitor time.
How does automated reference checking work?
- The recruiter enters the candidate in the tool.
- The candidate adds their references.
- Each reference gets a short survey by email or text, usually a mix of ratings and open questions.
- The tool collects the answers, summarizes them and flags anything unusual.
Harver, which now sells the Checkster product, says its surveys take references about two minutes on a phone and most checks come back within 24 hours. Those are vendor claims, but the time saving over phone tag is easy to believe.
| Phone reference checks | Automated reference checks | |
|---|---|---|
| Recruiter time | Calls, voicemails and notes for each reference | A few minutes to start the check |
| Speed | Often several days | Often within a day or two |
| Consistency | Questions vary by caller | Same questions for every candidate |
| Candour | Some references hold back on the phone | Written answers can be more open, but shorter |
| Follow up | Easy to ask a second question | Harder; call when an answer needs context |
How do these tools spot fake references?
Fake references are part of the wider rise in candidate fraud. Tools check where and how each reference responded. SkillSurvey flags references that answer from the same IP address, and the State of Wisconsin’s recruiter guide tells staff to call the reference to confirm, since two colleagues at one office can share an address innocently. Crosschq added a fraud module in 2026 that flags shared devices and VPN use and can verify identity.
A flag is a reason to check, not proof. Call the reference, or ask for a work email address you can verify.
Which laws apply to reference checks?
| Rule | What it means for you |
|---|---|
| Fair Credit Reporting Act | The FTC has said reference checks run by outside agencies that interview references can count as investigative consumer reports. That brings FTC employer duties: a separate written disclosure, written permission, and notice before you reject someone because of the report |
| Salary history bans | HR Dive counts 46 state and local salary history bans on asking about past pay. Keep pay questions out of reference surveys too |
| Limited reference policies | Many employers confirm only title and dates. SHRM says that is a choice to reduce legal risk, not a law. Ask for references who worked with the candidate directly |
Vendors usually supply the consent wording, but it is your process. Check it with your legal team, and do not treat this lesson as legal advice.
How can AI help you build a competitive offer?
Check the market. Free BLS wage data covers about 830 occupations by state and metro area. Paid tools such as Pave draw pay data from company HR systems. Ask an AI assistant to summarize both for the role and location.
Check your own pay. Compare the offer with what current employees in the same role earn. Pay transparency laws, covered in Lesson 2, mean candidates often know the range already.
Draft the offer letter. Give the AI your approved template and the agreed terms, and ask it to fill in the letter without changing any legal wording.
Prepare the call. Ask the AI to list what the candidate said matters to them in interviews, so the offer call addresses it.
Send for signature quickly. Most ATS and electronic signature tools let the candidate sign on a phone the same day.
Using this approved offer letter template [paste] and these agreed terms [title, start date, base pay, bonus, benefits, manager], fill in the letter. Do not change any wording outside the square brackets. Then list anything in the terms that is missing from the template, and anything in the template I have not given you a value for.
What must AI never do at the offer stage?
Set pay from protected traits. The EEOC says federal law bans pay discrimination based on race, color, religion, sex, national origin, age or disability. A pay tool trained on past offers can repeat old gaps, so check its suggestions against the role, not the person.
Use pay history. Where salary history bans apply, AI must not pull past pay from a resume or a profile and anchor the offer to it.
Hold personal data in the wrong place. OpenAI says consumer ChatGPT conversations can be used for training unless you opt out, while business plans are excluded by default. Offer details belong in a business account or your ATS. See Module 4, Lesson 5.
Make the call. The offer conversation is where a recruiter earns the hire. Use AI to prepare, then talk to the candidate yourself.
Quick answers
Are automated reference checks as good as phone calls?
They are faster and more consistent. For senior roles, many recruiters use the survey first and then call one or two references about anything that needs context.
How many references should I ask for?
Three is common, with at least one or two who managed the candidate. Some employers, such as the State of Wisconsin, require two supervisors among three.
Can I use AI to write a counteroffer response?
Yes, to draft it. Give it only the facts you are allowed to share and read it carefully before sending.
What is a good offer acceptance rate?
Benchmarks from Gem and Ashby sit around 80% for offers made. If yours is much lower, look at pay, speed and what was agreed at intake.
Explore in the directory
Module 6: AI across the recruiting desk
- How can AI help with hiring manager intake meetings?
- How do you write job descriptions and job ads with AI?
- What is a recruiting chatbot, and does it help candidates?
- How does AI help with reference checks and job offers?
- How do agency recruiters use AI to win new clients?