HR
Source qualified candidates faster, handle routine HR tickets automatically, and free up hours every week.
Updated last week · 9 sources
Proof over hype
Rate any example below — it feeds which ones stay on this page.
- Enterprise
Chipotle
Chipotle's restaurant turnover ran close to 200% a year, and hiring was, in COO Jason Kidd's words, 'one of the most painful processes' restaurant managers dealt with — every step of scheduling an interview competed with actually running a shift.
With Ava Cado handling the back-and-forth of scheduling and answering applicant questions around the clock, time from application to first day fell from 12 days to four.
The takeaway The time saved wasn't recruiter time, it was manager time. The person actually running the shift got the scheduling taken off their plate — a different, and often more binding, constraint than HR headcount.
How they did it
Ava Cado, Chipotle's AI hiring assistant, has been live since 2024. It talks with applicants, answers their questions about the job, collects their information, and schedules the interview itself, so no manager has to find a gap in their shift to make a phone call. Because it runs around the clock, about 30% of that scheduling now happens outside business hours, when applicants are actually free to deal with it and no general manager is on the clock to do it by hand. Once a hiring manager approves a candidate, Ava Cado sends the offer letter too. Kidd was careful to frame this as reinvestment rather than headcount reduction: everything freed up from AI went back into restaurant operations, and 'at the end of the day, that applicant ends up working for a human.'
- EnterpriseWorkday HiredScore
Formula 1
Formula 1 received 78,000 job applications a year, about 520 for every open role, with exactly three full-time recruiters to work through them across an 800-person organization — plus separate HR systems for hiring, payroll, time tracking and performance that didn't talk to each other.
Workday HiredScore reads each CV against the job description and ranks candidates by fit. With it, F1 cut screening time 16%, took 10 days off time to hire, automated 97% of screening, and gave every applicant an actual CV review instead of some going unread.
The takeaway The tool needed better inputs before it gave better outputs — F1 had to tighten its own job descriptions before the ranking got good. That prep work is usually the real bottleneck, not the software. One thing this source doesn't say: whether the ranking is free of bias by race, age or disability — F1 doesn't publish an audit and the article never raises it. Before taking that on faith with any vendor, read the Workday card further down this page: same underlying product line, currently being sued in federal court over exactly that question. Ask a screening vendor for their own bias-audit numbers before you sign, not after a rejected candidate asks first.
How they did it
F1's HR systems lead, Alastair Goss, picked HiredScore specifically because Workday had just acquired it and it slotted straight into the Workday system F1 already ran everything else on: 'I knew I needed something that fit well with our existing tech stack. I knew I needed a solution that was already proven.' HiredScore's job is narrow: it reads each applicant's CV against the job description and ranks candidates by fit, screening out weak matches so recruiters only look at the strong ones — it doesn't source candidates, schedule interviews or make the hiring decision itself. The tool didn't work well out of the box — F1's team found it needed genuinely precise job descriptions to rank candidates accurately, which forced recruiters to sit down with hiring managers and actually pin down what a role required before HiredScore could screen for it well. That upfront work paid off: with 97% of screening now automated, recruiters stopped doing manual CV review and shifted toward a hiring-strategy role, and for the first time, every applicant's CV gets an actual review instead of some going unread in the pile.
- Mid-market
Asymbl
Asymbl is a workforce company that runs on a mix of human staff and AI 'digital workers' side by side. When it hired 100 people in 100 days, its small HR team got buried under a flood of routine, repeat questions — mostly about benefits and time off — that it couldn't answer fast enough.
Polly now answers the routine benefits and time-off questions that used to land on a person, which freed up about 10 hours of HR work a week.
The takeaway The AI here is deliberately narrow: a Slack chatbot that only answers what it's been explicitly trained on, and hands off anything requiring judgment to a human instead of guessing. That limit is what made it safe to let employees talk to it directly — nobody gets a wrong benefits answer with no person accountable for it. If your HR team's time is going to answering the same handful of policy questions over and over, that's the specific, boring problem this kind of tool solves — not the harder judgment calls.
How they did it
Asymbl built Polly People Ops, an AI chatbot that lives in the company's Slack and answers employee questions on benefits, retirement plan changes and time-off requests, using the company's own written policies as its only source. It only answers what it has been trained on; anything needing judgment or falling outside documented policy gets routed straight to a person, instead of Polly guessing. CEO Brandon Metcalf said the framing mattered as much as the function — calling it 'onboarding' and treating Polly as a colleague rather than a tool shaped whether employees actually trusted it enough to use it. COO Greg Symons said the payoff wasn't just time back — it let HR staff spend that time on the parts of the job that actually needed a person.
- Learn from thisEnterprise
Workday
Derek Mobley is Black, over 40, and has documented anxiety and depression. He applied to more than 100 jobs through Workday's AI hiring platform and was rejected from every one — several rejection emails arrived in the middle of the night, a pattern suggesting no human had reviewed his application before it was rejected.
The age-discrimination claim was certified as a nationwide collective action in May 2025, and the case is proceeding against Workday itself as a legally responsible 'employer' — not just the companies that used its software.
So: Ask any screening vendor exactly which signals feed the ranking, get it in writing, and require a person to review and approve every rejection, not just the ones flagged as borderline.
The takeaway The detail that should worry you isn't the discrimination claim, it's the automated-speed one: rejections arriving in the middle of the night were the evidence that no human was in the loop at all. And because the vendor can be held responsible too, not just the employer, that exposure isn't fully outsourced just because you bought the screening tool instead of building it.
How they did it
Mobley sued, arguing Workday's screening software itself was making the rejection decisions, not the employers using it. A federal court let the case proceed on a novel theory: the complaint sufficiently alleged that Workday acted as an 'agent' of the employers who used its platform, which brings a software vendor within the legal definition of 'employer' under discrimination law — meaning Workday itself, not just the companies that hired it, can be held directly liable for how its own screening tool behaves. In May 2025, the same court granted preliminary certification of the age-discrimination claim as a collective action, letting other rejected applicants over 40 join the case.
- Learn from thisEnterprise
SiriusXM
Arshon Harper applied to about 150 roles at SiriusXM, all screened by the company's automated resume-screening software. He was rejected from all but one, despite, he says, meeting or exceeding the qualifications on paper every time.
So: Strip school name and zip code from what the screener sees, then check your rejection rates by race and age every quarter, before a rejected candidate does the math for you.
The takeaway Zip code and school name feel neutral. They aren't — they correlate closely enough with race that using them at all is now the basis of a lawsuit. Audit exactly what fields your screener looks at, not just what it's supposed to look at.
How they did it
Harper sued in August 2025, alleging SiriusXM's automated resume screening used his educational institutions, employment history and zip code as stand-ins for race, and that the system removed his resume from consideration for reasons that had nothing to do with whether he could do the job. The case is a class action, still in its early stages, and follows the same pattern as the Workday litigation: a rejection rate high enough, across enough applications, that the plaintiff argues it can only be explained by what the software was measuring rather than what the job actually required.
Try this today
Rewrite a policy page into plain English
20 minFewer repeat questions to your team, and a policy people will actually read.
Copy the prompt
You are an HR communications editor. Below is a company policy written for lawyers. Rewrite it for employees at [COMPANY] reading on a phone. Rules: keep every rule, deadline and dollar figure exactly as written; change no entitlement; use short sentences and the word 'you'; cut words that carry no meaning. Format it as a one-line summary, then four headed sections: What you get, What you have to do, Dates that matter, Who to ask. At the end, list under 'Check with legal' any sentence you were unsure how to simplify. Policy: [PASTE POLICY]
One check first. Read it against the original, line by line. Legal or benefits signs off before it replaces the real policy.
Find the themes in your exit interview comments
30 minOne page of the real reasons people leave, ranked, ready for your next leadership meeting.
Copy the prompt
You are a researcher who reads employee feedback for a living. Below are the written comments from [NUMBER] exit interviews at [COMPANY]. Group them into no more than seven themes. For each theme give: a plain-English name, how many comments sit in it, two short quotes as evidence, and whether it points at pay, the manager, workload, career path or something else. Rules: use only what is in the comments; invent nothing; say so when a theme rests on fewer than three comments. End with the three themes a leader could act on this quarter. Comments: [PASTE COMMENTS]
One check first. Strip names and job titles before you paste. Check every quote against the source; the counts and the grouping are guesses.
Rehearse a hard conversation before you have it
15 minYou walk in with the three hardest questions already answered, not hearing them for the first time.
Copy the prompt
You are coaching me before a hard conversation at work. I am [YOUR ROLE]. I need to tell [WHO THIS IS] that [WHAT YOU HAVE TO SAY]. Background: [TWO OR THREE LINES OF CONTEXT]. Give me four things: an opening of no more than four sentences; the five hardest questions this person is likely to ask, each with a straight answer; three phrases to avoid because they sound rehearsed or evasive; and one sentence to close on. Rules: no jargon, no filler sympathy lines. Flag anything in my background notes that a lawyer should read before I say it out loud.
One check first. Anything touching pay, dismissal or a live complaint goes to employment counsel first. The model does not know your contracts.
Specialized tools, and what to ask vendors
| Tool | What it does | Setup | Best fit |
|---|---|---|---|
| Eightfold AIenterprise | Reads your open roles and your applicant pool and ranks who is actually a fit, including people already inside your company. | MonthsIT sign-off | You hire at volume and your team is drowning in inbound applications.Skip it ifUnder roughly 50 hires a year. Setup cost outweighs the saving, and thin data makes the ranking guesswork. |
| Paradoxmid-market to enterprise | Handles the back-and-forth of scheduling interviews and answering candidate questions over text. | WeeksIT sign-off | High-volume hourly or retail hiring where speed decides who you get.Skip it ifSmall executive search, or roles where candidates will not text. A bot cannot sell a senior person on the job. |
| Gloatenterprise | Matches your existing employees to internal projects and roles based on skills rather than job titles. | MonthsIT sign-off | Large workforce, real retention pressure, and a leadership team that will back internal moves.Skip it ifUnder a few thousand employees. The marketplace needs density, or people open it once and never return. |
Questions to ask before you buy
Eightfold AI — 7 questions to ask them
- Which of your customers is closest to my headcount and industry, and can I talk to them directly?
- Can we require a person to approve every rejection, and what record shows who decided?
- How much of my HRIS (the system holding employee records) and ATS (the system that stores job applications) data has to be clean before this produces anything useful?
- How do you handle bias auditing, and can you produce the documentation my legal team will ask for?
- Where does my employee data live, who can access it, and what happens to it if we leave?
- What is the total first-year cost including implementation, and what does my team have to staff?
- What is the most common reason a rollout like mine disappoints?
Paradox — 7 questions to ask them
- What happens when a candidate asks something the system cannot answer?
- Where does candidate chat data go, how long do you keep it, and does it train your models?
- Which of your customers hires at my volume in my industry?
- How do you handle accessibility and candidates who cannot or will not use text?
- What compliance documentation do you provide for automated screening — specifically for NYC Local Law 144 and Illinois HB 3773?
- What is the pricing model, and how does it change as our hiring volume moves?
- Show me a deployment that underperformed and tell me why.
Gloat — 7 questions to ask them
- How do you get skills data if our HRIS records are incomplete, which they are?
- What adoption rate do your comparable customers actually reach in year one?
- What do managers have to do differently for this to work, and how do you get them to do it?
- How does this connect to our existing HRIS and learning systems?
- What is the realistic timeline to the first measurable retention effect?
- Where does our employee data live, does it train models for other customers, and what happens on exit?
- Which customers stopped using you, and what did they say?
What everyone is asking
- NewJul 2026
Best AI for hiring hourly and frontline staff right now?
Text-first assistants that screen and book interviews. Speed decides who you get. Chipotle cut application-to-start from 12 days to four. Formula 1 automated 97% of screening.
What to watch for
Nobody argues about the speed gain. People disagree about letting software rank candidates, which is where the lawsuits start.
Also worth a look. Paradox, Workday HiredScore, Phenom
- NewJul 2026
Will AI resume screening get my company sued?
It can. A judge let discrimination claims against Workday move forward in 2026, and SiriusXM faces a race claim over automated screening. Buying the tool does not move the risk.
What to watch for
Lawyers agree the exposure is real. They split on how far an audit protects you — in the Workday case, its own testing was shielded as legal advice.
Also worth a look. Keep a person reviewing every rejection, Run an independent bias audit — an outside check for skewed results — every year, Turn off automatic ranking and use the tool only for scheduling
- NewJul 2026
Best AI for answering employee HR questions?
A small agent trained on your own policy, sitting where people already talk. Asymbl put Polly in Slack and took about 10 hours a week off its HR team.
What to watch for
The split is over judgment calls. These agents answer policy questions well and hand anything needing interpretation to a person.
Also worth a look. An agent trained on your own policy documents, ServiceNow HR agents, The assistant already built into your HRIS (your core employee record system)
- NewJul 2026
Best AI for internal mobility and skills mapping?
It depends on your size. Gloat suits enterprises above roughly 5,000 staff losing people to outside roles. Eightfold suits you when external hiring and internal moves share one system.
What to watch for
Both comparisons come from vendors in the same market. Read them for the tradeoffs, not the verdict.
Also worth a look. Gloat, Eightfold AI, Engagedly, Beamery
Worth following
Josh Bersin
Industry analyst covering HR technology and the future of work.
weekly · newsletter
Why them
He evaluates HR tech vendors independently and says when something does not work.
Hung Lee — Recruiting Brainfood
Curator of the most-read weekly newsletter in recruiting.
weekly · newsletter
Why them
One email replaces a week of scanning. Strong on what practitioners are actually adopting.
HR Brew
Daily HR news briefing for people teams.
daily · newsletter
Why them
Short, readable, and it names the companies running the experiments.
HR Dive
Daily trade publication covering HR news, compliance, and technology.
daily · publication
Why them
Best early warning on new state hiring laws before your lawyers call you.
UNLEASH
HR technology media and conference organisation.
daily · publication
Why them
Publishes named case studies with real numbers, which almost nobody else does.
SHRM technology coverage
The professional body's technology and compliance reporting.
weekly · publication
Why them
Where the legal side gets explained in plain terms. Much of it sits behind membership.