Across ten real employers, from Chipotle and McDonald's to IBM, Bank of America, and Delta, HR runs on AI the same way. The software takes the highest-volume, most repetitive parts of the employee lifecycle, screening and scheduling candidates, answering routine HR and benefits questions, and matching employees to open internal roles, while an HR professional keeps the judgment and signs off on the decision. The teams reporting the biggest gains did not buy one grand HR AI platform. They put one reviewed tool on one high-volume task and kept a person in charge of the outcome.
If you run HR, recruit at volume, or own the people function, you have heard two years of noise about AI in HR. Some of it is real. A lot of it is software companies describing a future that has not arrived. This piece does something narrower and more useful. It looks at ten named employers, what they actually use, and what actually happened, with a source for each.
They are not all giants doing giant things. They are a restaurant chain hiring hourly crew, a bank answering its own staff, an airline moving people between roles, and a biotech company where HR built its own tools. The industries differ, the tools differ, the numbers differ. The pattern does not. Read them as a working menu, not a sales pitch, and pay as much attention to the limits, especially the legal ones, as to the wins.
Key takeaways
- Ten real employers, from restaurant chains to a global bank and an airline, across the United States and Europe.
- The pattern is identical everywhere: AI runs the highest-volume parts of the employee lifecycle, and an HR professional keeps the judgment and the decision.
- Front-door hiring: Chipotle, General Motors, 7-Eleven, and McDonald's use Paradox assistants to screen, schedule, and make offers. GM cut scheduling from over 5 days to 29 minutes; 7-Eleven saves about 40,000 store-manager hours a week.
- Internal service: IBM's AskHR resolves about 94 percent of common HR questions without a human; Bank of America's Erica for Employees is used by more than 90 percent of staff and halved IT service desk calls.
- Skills and mobility: Vodafone recruiters recovered about 16 hours a week with Eightfold; Delta now fills close to half its manager roles internally with SAP SuccessFactors.
- The honest limits: AI-hiring law, like Connecticut's CART Act and NYC Local Law 144, makes a chatbot no defense to a bias claim, and the reported numbers come from the firms and their vendors, not an independent audit.
| Use case | What AI does | Example employers | Caveat |
|---|---|---|---|
| High-volume hiring | Screens, schedules, and sends offers by chat | Chipotle (Ava Cado), McDonald's (Olivia) | Bias and data-security rules still apply; a human should own final selection. |
| Interview scheduling | Books interviews automatically across teams | General Motors (Ev-e), 7-Eleven (Rita) | Strong for volume; low value for senior or bespoke roles. |
| HR service desk | Answers benefits, policy, and IT questions | IBM (AskHR), Bank of America (Erica) | Needs a clean, current knowledge base or it answers wrong. |
| Employee-built assistants | Custom GPTs for HR tasks like interview prep | Moderna (Hiring Advisor GPT) | Governance and confidentiality controls are on you. |
| Skills and internal mobility | Matches employees to open roles by skill | Vodafone (Eightfold), Delta (SuccessFactors) | Only as good as your skills data; setup is real work. |
| Workforce AI literacy | Trains staff, gates tool access behind training | Colgate-Palmolive (AI Hub) | Training is the start, not the finish; measure real use. |
The front door: AI that screens, schedules, and makes the offer
The clearest wins in HR are at the top of the funnel, in high-volume hourly and early-career hiring. Four employers here run the same play with the same vendor, Paradox, and a named conversational assistant that talks to candidates so a recruiter does not have to.
Chipotle: a virtual recruiter across 3,500 restaurants
Chipotle deployed an AI assistant called Ava Cado across more than 3,500 restaurants in North America and Europe. It chats with applicants, answers their questions about the job, collects basic information, schedules interviews for hiring managers, and sends offers, in English, Spanish, French, and German. Chipotle expects the platform to cut the time it takes to hire an in-restaurant worker by as much as 75 percent.
Source: Chipotle newsroom, "Chipotle Introduces New AI Hiring Platform," October 22, 2024. newsroom.chipotle.com
McDonald's: the application that went from 10 minutes to 2
McDonald's uses Olivia, a Paradox assistant, to pre-screen candidates and book interviews for hourly roles through its McHire system. The application dropped from about 10 minutes to two, and McDonald's reported time-to-hire down roughly 60 percent, with about five hours a week returned to store managers. It is also a cautionary tale worth reading to the end, and the caveats section below explains why.
Source: Paradox case study, "How McDonald's cut hourly time to hire by 60%." paradox.ai
General Motors: scheduling from five days to 29 minutes
GM runs a Paradox assistant it calls Ev-e for interview scheduling. Ev-e has automatically scheduled more than 74,000 candidate interviews, cut the average scheduling timeline from over five days to 29 minutes, and saved the company about $2 million a year in hiring automation. This is the single cleanest before-and-after number on the page.
Source: Paradox, General Motors case study. paradox.ai
7-Eleven: 40,000 manager hours back every week
7-Eleven's assistant, Rita, saves roughly 40,000 store-manager hours a week across the chain and schedules 85 percent of applicants within an hour of applying. In high-turnover retail, that is time store leaders spend running the store instead of chasing candidates by phone.
Source: Paradox, 7-Eleven case study. paradox.ai
The internal help desk: AI that answers the workforce
The second pattern points inward. Once you have hired people, they generate a steady stream of routine questions, about benefits, policy, PTO, and IT. Three employers put an assistant on that stream so HR staff stop repeating themselves.
IBM: AskHR resolves 94 percent of common questions
IBM built AskHR on its own watsonx Orchestrate platform. It automates more than 80 HR tasks, resolves about 94 percent of common employee inquiries without a human, and helps managers complete work like promotions an estimated 75 percent faster. It is one of the most-cited internal HR deployments precisely because IBM runs it on its own staff at scale.
Source: IBM, "IBM AskHR" case study, and IBM Think, "Enterprise transformation and extreme productivity with AI." ibm.com/case-studies/ibm-askhr
Bank of America: an assistant more than 90 percent of staff use
Bank of America extended its consumer virtual assistant into Erica for Employees, an internal assistant for HR and IT questions. More than 90 percent of employees now use it, and it has cut calls into the IT service desk by about 50 percent. When nearly everyone in the building uses the tool, the help desk load drops in a way you can measure.
Source: Bank of America newsroom, "A decade of AI innovation: BofA's virtual assistant Erica," August 2025. newsroom.bankofamerica.com
Moderna: HR that builds its own tools
Moderna went further than buying a tool. In late 2024 it merged HR and IT into a single People and Digital Technology function, and on ChatGPT Enterprise its employees have built more than 3,000 custom GPTs. HR built one of them, a Hiring Advisor GPT that drafts role-specific interview questions. It is the clearest proof on this page that a small HR team can build useful AI without an engineering department.
Source: OpenAI, "Moderna" customer story, and Constellation Research and Harvard Business School coverage. openai.com/index/moderna
Skills and internal mobility: AI that moves people up
The third pattern is quieter and, over a career, may matter most. Instead of hiring from outside, these employers use AI to see the skills they already have and route people to open roles internally.
Vodafone: 16 hours a week back to each recruiter
Vodafone runs Eightfold's talent-intelligence platform to keep its skills and job architecture current in real time. Recruiters recovered about 16 hours a week that used to go into sourcing, and Vodafone reduced global time-to-hire by half. The time did not vanish, it moved to talking with candidates and hiring managers.
Source: Eightfold and UNLEASH, "Vodafone on how Eightfold AI drove candidate satisfaction into positive figures." eightfold.ai
Delta Air Lines: filling half its manager roles from inside
Delta built a skills-based talent strategy on SAP SuccessFactors, with an Opportunity Marketplace and a jobs taxonomy that maps 482 skills to 100 percent of its roles across more than 100,000 employees. Using that infrastructure, Delta now fills close to half of its managerial roles with internal hires. For a workforce that size, that is a different way to run a people function.
Source: Delta ESG report, "Talent Management," and HR Brew, "How Delta got its skills-based talent strategy off the ground," November 2024. hr-brew.com
Upskilling the function itself
Colgate-Palmolive: training as the gate, not an afterthought
The last example is not a hiring bot. Colgate-Palmolive requires every employee to complete responsible-AI training before they can access its internal AI Hub, an HR and learning-led program that treats AI literacy as a condition of use rather than an optional lunch-and-learn. Thousands of employees report better quality and creativity in their work as a result. It is the least flashy case here and the most transferable: whatever you deploy, someone has to teach people to use it well.
Source: MIT Sloan Management Review, "The GenAI Focus Shifts to Innovation at Colgate-Palmolive." sloanreview.mit.edu
What still does not work, and what every team kept human
Ten wins can read like a clean line, so here is the other half, drawn from the same employers and the laws they operate under.
The last mile is judgment. AskHR clears about 94 percent of common questions, which means the rest, the sensitive, the ambiguous, and the legally fraught, still lands on a person. Screening bots move volume. They do not decide who is right for the team, and they should not.
AI hiring is regulated, and a vendor is no shield. Connecticut's CART Act and New York City Local Law 144 treat automated employment decision tools as your responsibility. Using one is not a defense to a bias claim. You still owe candidate notice, bias auditing, and a human decision-maker on the outcome. For the current state of that rule, see our desk note on the Connecticut CART Act and AI employment chatbots. Decide your compliance posture before you turn a screening tool on.
Confidentiality and security are real, not theoretical. Applicant and employee data is sensitive and often regulated. In 2025, security researchers showed that applicant records on McDonald's McHire hiring system could be reached through weak access controls, a blunt reminder that these systems hold real personal data and need a real security review before launch. The convenience is not free of duty.
The numbers come from the firms and their vendors. The outcomes here are reported, not independently audited. They are credible and attributed, but they are also the best cases the companies and their software partners chose to publish. Treat them as what is possible, not what is guaranteed on your headcount.
It rewards a method. Moderna's HR built a useful GPT because someone was willing to experiment and check the output. Teams without that person usually need a defined workflow and some training to get the same result, which is the gap a structured program is meant to close. If you want a low-risk place to build that habit, the Leverage Starter is where most people begin.
How to think about your first step
If these employers point at one move, it is this: do not start with a platform, start with a task. Pick the single highest-volume job in your HR week. For most functions that is first-round screening, interview scheduling, or the same benefits and policy questions you answer fifty times a month. Point one reviewed tool at that one task, keep an HR person on every decision, and measure the hours for a month.
Then write down two rules before anyone goes further: what candidate and employee data may enter a tool, and where a human has to sign off. Get compliance and confidentiality down first, especially if a tool touches hiring, train the team on the one workflow, and only add a second task once the first is boring. That is the unglamorous order every team on this page followed, whether they said so or not.