EP 135: Where to Start With AI in HR - 5 Tasks to Automate First
Series: AI in HR (Ep. 2 of 3) | Host: Kerri Roberts | Run time: ~30 minutes
Episode Summary
Automate the right things and it’s efficiency. Automate the wrong things and it’s risk. And both show up on your P&L.
Last week we asked whether your HR department is AI-ready. This week is the natural next question: where do you actually start? The fastest way to get burned is to automate everything at once. It isn’t sustainable, you put the wrong things in, and the whole thing gets wobbly.
So this episode is a shortlist: the five high-volume, low-judgment HR tasks worth automating right now — resume screening, interview scheduling, onboarding paperwork, an employee FAQ built on your own handbook, and first drafts — plus the three things you should never hand to AI. Kerri also shares how automating admin and marketing work saves her business roughly $65,000 to $70,000 a year.
The rule is simple: if a task is high-volume, repeatable, and low-judgment, it’s a candidate. If a wrong call could hurt a real person or create legal exposure, it stays with you.
In This Episode
The one rule for spotting an automatable task: high-volume, repeatable, and low-judgment.
Why resume screening is the number-one thing to automate, and why it isn’t new (Indeed, LinkedIn, and your payroll system have done it for years).
How to set up a minimum-qualifications screen in Claude, ChatGPT, or Gemini using your job description.
Killing the interview-scheduling back-and-forth with a booking link and a small library of email templates.
Automating onboarding paperwork and new-hire workflows so nothing falls through the cracks.
Building an employee FAQ and policy-lookup assistant on your own handbook, not the open internet.
Using AI for first drafts of job descriptions, offer letters, and policy language, then editing to 100%.
The three things to never automate: final hire and fire decisions, performance and coaching, and sensitive judgment calls.
Why starting with just one task is how it sticks.
Chapter Timestamps
0:00 Where do you start with AI in HR?
1:00 How AI saves me $65K to $70K a year
2:20 Pareto: what to automate vs. what to cut
4:00 The rule: high-volume, low-judgment tasks
4:45 #1 Resume screening and initial filtering
11:50 #2 Interview scheduling and candidate communication
16:30 #3 Onboarding paperwork and new-hire workflows
18:38 #4 Employee FAQ and policy lookups
22:00 #5 First drafts: job descriptions, offers, policies
24:00 3 things you should never automate
26:30 The stats: why this pays off
28:00 Where to start plus resources
Resources Mentioned
The HR Easy Button (book): saltandlightadvisors.com/thehreasybutton — paperback, hardback, and Kindle on Amazon
HR Foundations (course, under $400): saltandlightadvisors.com/hrfoundations
HR Operations (course): saltandlightadvisors.com/resources
Strategic HR coursework: available starting August 2026
Free Mini HR Audit (10 minutes): saltandlightadvisors.com/hraudit
Work with Salt & Light / book a consultation: saltandlightadvisors.com/contact
Tools referenced: Claude (Projects), ChatGPT (custom GPTs), Gemini, Calendly, Microsoft Bookings, Zapier, Slack bots
HR/payroll platforms: Gusto, UKG, Rippling, Paycom, Paycor, Paylocity, ADP, QuickBooks Online
Sourcing platforms: Indeed, LinkedIn
Episodes in the AI in HR Series
Ep. 134: Is Your HR Department AI Ready?
Ep. 135: Where to Start With AI in HR - 5 Tasks to Automate First (you’re here)
Ep. 136: Writing an AI-Use Policy for Your Organization (coming next week)
Your Action Item
Pick just one of the five — resume screening, interview scheduling, onboarding paperwork, an employee FAQ, or first drafts. Start there today. Document how the process works right now so AI has something to learn from, get that one running, and only then move to the next. One at a time is how this sticks. Do it all at once and it just becomes another part-time job.
Get the Monday Email
If this episode resonated, you’ll like the Monday morning email. Every Monday at 5:28am, Kerri sends one practical idea for leaders who want to do this work better — including the conversations behind episodes like this one. 1,000+ leaders. 50%+ open rate. saltandlight.myflodesk.com/saltandlightadvisors
Full Transcript
Where do you start with AI in HR?
Last week we talked about the question, is your HR department AI-ready? And many of you are. So this week we’re jumping into where do you start? That’s exactly what we’re going to focus on, and I’m super excited. We’re talking high-volume, low-judgment tasks that you can automate in your HR team right now. And if you don’t have an HR team, even better — this is exactly where you need to start if you’re the CEO of a small business or an office manager. So let’s get into it.
Welcome back to another episode of Don’t Waste the Chaos. I’m your host, Kerri Roberts. I’ve been doing this HR thing for well over 20 years now, and I use AI a ton in my business.
How AI saves me $65K to $70K a year
Outside of the HR aspect, from the administrative side of my business, I’m saving around $65,000 to $70,000 a year by using AI. How am I doing that math? I used to have two part-to-full-time people who worked for me. One focused on marketing, the other on admin. Those positions have been eliminated and automated, and I’m using AI for that. It’s been fun, challenging, and exciting. But the best part, being a small business owner like so many of you, is that while top-line revenue was growing, bottom-line revenue grew too. Which has been 10 out of 10, thank you very much.
So we’re going to talk about how to implement AI specifically in your HR function. Where I want to start is: not with everything. If you start with everything, you’re going to get burned. It’s not sustainable, you’ll get wobbly, and you’ll put the wrong things in there. Today I’ll cover five things you should automate immediately, then three things you absolutely should not use AI for. While I’m using it for a lot, I’m certainly not using it for everything. My clients don’t talk to AI. When it comes to my client work, it’s me.
Pareto: what to automate vs. what to cut
This is a great time to look at revenue-generating activities and what you actually need to do. I applied the Pareto principle in my business and asked: where’s 80% of my revenue coming from? Those things I focus on, and I take the high-volume, low-judgment pieces of those tasks and automate them. The places not generating much revenue, I just cut. If you’ve got a person doing a task, you’ll make sure they have a fulfilling job assignment. But if you’re deciding what to automate, you’re not going to automate things that don’t matter. This gives you hours back of your week. It didn’t automate my whole job, though — I’m doing more work and I’ve got more clients. June and July were my highest revenue months ever. Automate too much and it’s an expensive mistake; automate too little and that can be one too.
The rule: high-volume, low-judgment tasks
Here’s the rule. If you do it a lot, it’s repeatable, and it’s low-judgment — so it doesn’t really need a person, but it drains your week — that’s when you know it might be a good task to automate. Anything where a wrong call could hurt you, hurt a real person, or cause legal exposure, we don’t want to lean into AI for those things.
#1 Resume screening and initial filtering
I’ll start with one that’s controversial. I’ve seen a ton of people say you absolutely shouldn’t use AI for this, and I’d say they’re wrong — they’re looking at it the wrong way. This is exactly a place I recommend to all of my clients. People say AI is biased, but if you’re feeding the prompts correctly and giving it accurate information, it shouldn’t be any more or less biased than a human screening it.
This is first-pass sorting against must-haves. You spend your time on the short list, not the entire pile. People are still involved — just in the short list. If you’re using Indeed or LinkedIn, or an applicant tracking system inside a payroll platform like UKG or ADP, you’ve already been using AI to screen for minimum qualifications for years. We call it minimum quals, and I’ve had it in place since my career started in 2008.
You can do a simple version using a Claude Project or a custom GPT loaded with your job must-haves first. You load your must-haves, then your qualifiers, and it routes candidates one direction (advancement) or another (declination). Copy and paste the job description, or upload it, then upload the applications or resumes. Give instructions to review against the must-haves. If they don’t have the must-haves, they’re a declination. We don’t screen on things entirely missing like job gaps or years not worked; we’re looking for a specific set of things, and then AI ranks them.
Say I post a role for a social media specialist. My description says a minimum of three years in marketing, experience with Canva and a scheduling tool, maybe a year with podcasts, and a two- or four-year degree. I put the description into AI, add the applications, and say: rank these candidates, four stars down to one, on their fit against the must-have qualifications. Inside an ATS I might auto-route anyone without the must-haves into the declined pile and send a declination email. If you don’t feel comfortable, just sort the pile — don’t reject the pile.
I made a TikTok post a couple months ago about using AI to help screen candidates, and people went bananas: what’s the point of HR if AI does your job? That’s ignorance. Almost every website you apply to already does this. The human checkpoint: read every shortlisted resume yourself, and never auto-reject on anything that may carry bias, like an employment gap, a non-traditional path, or a career changer. Those aren’t minimum qualifications. Minimum qualifications are technology, years of experience, and education. If you’re a candidate, truly capture what’s being looked for in your resume. Without minimum quals, a role gets hundreds of applications and it’s paralysis by analysis. This one is golden.
#2 Interview scheduling and candidate communication
These go hand in hand. The back-and-forth eats days on your calendar, and it shouldn’t. Set up auto-declinations, and a message that advances people to the next step — maybe a questionnaire they complete and email back. That weeds out folks who are mass-applying versus those who really want to work at your company. Some clients in construction, warehousing, distribution, or trucking, I skip the questionnaire so I don’t make it hard to find a candidate. Read the room. If you don’t do a questionnaire, send a scheduling link — Calendly, Microsoft Bookings, whatever you use — so they book directly onto your calendar and attach their resume. Get rid of the back-and-forth. Then build a small library of email templates for each stage: received, moving forward, scheduling, and the decline. Let AI draft and personalize them in seconds. The human checkpoint: the decision is yours, with a human touch on anything sensitive, especially the no. Anyone who went all the way through deserves a personal note or a call. Automate the logistics, not the empathy — people are talking to their friends and family about this potential life change. Always reply. If you aren’t doing declinations, start today.
#3 Onboarding paperwork and new-hire workflows
I love workflows, especially in HR payroll technology. If you don’t have that, use Zapier or scheduled tasks inside a Claude Project. Build checklists: how you handle recruitment, what happens from offer to first day, and what runs from day zero to day 90. If you don’t have a checklist, AI gets you 70 to 80% of the way there, then you refine it. On a platform like Gusto, UKG, Rippling, Paycom, Paycor, Paylocity, or even QuickBooks Online, you can often build the new-hire checklist right in there: offer letter, I-9, W-4, e-signatures, benefits enrollment, IT and account setup, reminders. It pushes to your new hire automatically and keeps you on track. AI can also draft role-specific 30/60/90-day check-ins from your job description. The human checkpoint: the true welcome and the relationships can’t be automated — the manager one-on-ones, the introductions, making someone feel they made the right choice. Everything that would be a paper form gets automated.
#4 Employee FAQ and policy lookups
How much PTO do I have? What’s the holiday schedule? Do we have bereavement leave? Those answers already live in your policies. Build an assistant on your handbook — a Claude Project, a custom GPT trained on your documents, or a Slack bot — so employees get accurate, instant answers from your actual documents, not the open internet. The human checkpoint: anything sensitive or legally attached, like FMLA eligibility, routes to a human. Do the same thing externally with client questions: a website chatbot built from your FAQ answers pricing and cancellation questions and cuts down your email load. AI can even cite the policy — page XYZ, section 3.2.1. Could they scroll and find it themselves? Sure, but this makes it easier, and we’ve got the technology.
#5 First drafts: job descriptions, offers, policies
The first draft of a job description, an offer letter, or policy language. AI drafts it, you edit it — no more blank pages. Give it your templates, your voice, and your brand, and it turns a blank page into a solid draft in minutes. Tell it the compensation, exempt or non-exempt status, and who the role reports to, and it drafts an offer letter. Say you work in a certain state and industry and need a social-media-use policy, and it gives you a starting draft or asks the questions you should consider. It can even review language you already have. The human checkpoint: you edit every single one to fit your culture, and anything with legal weight goes to your attorney. AI gets you to about 80%; you do the last 20%.
3 things you should never automate
Three things I would never automate — not once we have enough data, but for legal protection and for culture. One: the final hire and fire decision. You have context AI doesn’t. It can give you a pros-and-cons list or an outline to consider, but you make the call. Two: performance and coaching conversations. The human relationship is the deal here — use your empathy and context. Three: any sensitive judgment call, like an investigation or an accommodation, or anything touching someone’s individual rights or human dignity. All of these are potential financial decisions. If we don’t automate, we waste time and money. If we automate too much, we create risk. Automate the right things and it’s efficiency; automate the wrong things and it’s risk — and both show up on the P&L.
The stats: why this pays off
Deloitte: HR teams spend up to 57% of their time on administrative tasks. In general, 75% of most HR tasks are administrative. SHRM: 85% of employers using automation or AI report time savings and greater efficiencies, and AI-powered recruitment can cut cost per hire by 30%. Recruiting is the number-one place HR is already applying AI, with 27% of organizations already using it — and I’d say it’s more, because a lot of people don’t realize the minimum-qualifications screen in Indeed and LinkedIn is AI. And 65% of HR professionals still don’t feel ready to work with AI. So if you feel behind, you’re not. It’s just time to step it up, one at a time.
Where to start plus resources
I walk through these exact workflows, especially around hiring and onboarding, in my HR Foundations course. Go to saltandlightadvisors.com/resources for the coursework — Foundations, HR Operations, and Strategic HR, available starting in August. My book, The HR Easy Button, is on Amazon in paperback, hardback, and Kindle. If you’re ready but lean and mean and don’t have the internal capacity, and you want to work with me, go to saltandlightadvisors.com/contact. Or, to see how your HR is standing up right now for free, saltandlightadvisors.com/hraudit gives you a mini HR audit in about 10 minutes and ranks where you stand. Pick one of the five today, and I’d love to hear which one you chose. Next week we’re getting into policy writing with AI — writing a policy about AI use in your organization, which we all need, because our employees are using AI, and without a policy we’re not protected. If you haven’t hit subscribe, do that, and please leave a review — it helps other leaders find the show. Folks, I know this can feel like chaos. But the work is worth it. So don’t waste the chaos. Embrace it. I’ll see you next week.
Resources To Keep Building
Not sure which HR tasks are safe to automate and which ones will burn you? Take the free HR Audit to see whether your people systems are actually ready to hand work to AI. saltandlightadvisors.com/hraudit
Take the free HR Audit — score your HR systems in 5 minutes and see exactly where your gaps are: saltandlightadvisors.com/hraudit
Explore the HR coursework — self-paced courses, tools, and resources: saltandlightadvisors.com/resources
Get the Monday Email — one practical idea for leaders, every Monday at 5:28am: saltandlight.myflodesk.com/saltandlightadvisors
Need fractional HR support or want to talk through a specific challenge? saltandlightadvisors.com/contact
