If your skills line says "ChatGPT" or "Prompt Engineering," it lists the one AI skill employers have stopped asking for. Workday's requisition data shows demand for basic AI skills peaked in January 2026 and has fallen 25% since, while demand for people who build AI tools and automate workflows rose 51%.
What Workday's Global Workforce Report found
Workday found that demand for basic AI literacy is falling, demand for applied AI skills is rising, and each hire now draws more applicants than before. The data comes from requisitions at more than 550 employers that use Workday Recruiting.
Here are the numbers side by side:
| Signal | Earlier | Latest |
|---|---|---|
| Basic AI skills (simple prompting) | Peak, January 2026 | Down 25% |
| Applied AI skills (building tools, automating workflows) | September 2025 baseline | Up 51% by July 2026 |
| Median applicants per filled job | 58 | 69 |
| Time-to-fill | About 60 days | About 60 days |
The two AI figures cover different windows. The 25% drop is measured from the January peak, and the 51% rise runs from September 2025. The direction is still clear. Employers now take basic prompting for granted and pay for the ability to build something with it.
Phil Willburn, Workday's VP of people systems, intelligence and support, summed it up: "Employees may not be changing jobs, but their jobs are changing around them." Workers feel this too. In Workday's survey, 62% of people who use AI for nearly all their work said they think it will make their current skills less valuable.
Non-technical resumes are listing the wrong skill
Most marketers, analysts, designers and ops managers list the AI skill that is losing demand, not the one that is gaining it. In Refolk's index of US professional profiles, across five non-technical titles (Marketing Manager, Marketing Specialist, Data Analyst, Operations Manager, Product Designer), "Prompt Engineering" shows up 3.4 times as often as "Workflow Automation."
| Skill tag | US profiles | Ratio vs Workflow Automation | Top title in sample |
|---|---|---|---|
| Prompt Engineering | 463 | 3.4x | Marketing Manager (7 of 25) |
| ChatGPT | 385 | 2.8x | Marketing Manager (9 of 25) |
| Workflow Automation | 138 | 1.0x | Data Analyst (8 of 25), Ops Manager (6 of 25) |
Profiles can carry more than one tag, so the rows don't add up to a total. The title breakdowns come from 25-profile samples, so read them as a rough signal. Of the profiles that carry either the prompting tag or the automation tag, about 23% (138 of 601) carry the automation one.
From Refolk's index, across Marketing Manager, Marketing Specialist, Data Analyst, Operations Manager and Product Designer.
Why supply runs opposite to demand
The explanation is cost. Adding "ChatGPT" to a skills line takes a few seconds and needs no proof, so almost everyone adds it. Building an automation that runs every week takes real work, so fewer people have one, and fewer still write it down.
That leaves the generic tag crowded and the in-demand one fairly empty. With 69 applicants per hire, the people who list "Prompt Engineering" all look alike. The people who can describe a workflow they automated face less competition for the skill employers say is growing.
Why "ChatGPT" in your skills section no longer earns credit
Hiring managers now expect basic AI use the way they expect Excel. They treat it as a baseline skill, meaning something every candidate is assumed to have, so listing it doesn't set you apart. Proof that you used it to get a result does.
The Resume Genius 2026 Hiring Insights Report, a survey of 1,000 US hiring managers, shows this:
- 78% said AI tool proficiency is at least somewhat important.
- 19% called it very important.
- 60% said they want to test, discuss or see proof of a candidate's AI abilities.
- AI and machine learning ranked sixth among the skills hiring managers prioritize.
The gap between 78% and 19% matters most. "Somewhat important" means managers expect you to have the skill. They won't reject you over it, and they won't pick you for it either. The 60% figure tells you what does get credit: something they can test. A bare "ChatGPT" gives them nothing to test. A bullet with a number does.
A tool name tells a recruiter you have a login. A workflow tells them you shipped something.
More applicants in the same 60 days means faster skimming
Recruiters are reading more resumes per hire in the same amount of time, so each resume gets less attention. Median applicants per filled job went from 58 to 69, roughly 19% more, while time-to-fill stayed near 60 days.
Two more findings make the skim harsher:
- 58% of hiring managers in Resume Genius's 2026 survey of 1,500 managers said they use AI to screen resumes.
- 86% of US hiring managers in the Express Employment Professionals and Harris Poll survey from February 2026 said AI makes it too easy to exaggerate skills on resumes.
So two readers check every resume. Screening software looks for tool names. Then a skeptical person decides whether to believe the claim. A skills line full of AI logos can get past the software and still fail with the person, who has seen the same list on plenty of other resumes. With 69 applicants on the same posting, an identical skills line just puts you deeper in the pile.
The workflow automation resume bullet: input, tool, output, saved
A workflow automation resume bullet describes one repeating task you handed to software. It names what went in, which tool did the work, what came out, and how much time or error it saved. That format works for both readers. The screener finds the tool name, and the person gets a claim they can check in an interview.
The four parts:
- Input. What triggers the work or feeds it. Inbound leads, support tickets, a weekly data export, design requests.
- Tool. What did the work. Name it: Zapier, ChatGPT, Claude, or whatever you used. This is where the keyword goes.
- Output. What came out the other end. A routed lead, a drafted reply, a finished report, a tagged asset.
- Saved. The measurable change. Minutes per task, hours per week, errors avoided, turnaround time.
CareerAddict gives a good real example: "Built a ChatGPT prompt library for the support team, and it cut our average first-response drafting from 12 minutes down to 4 minutes." The input is support tickets. The tool is ChatGPT. The output is a drafted first response. The saving is 8 minutes per reply. You can defend every part of that sentence in an interview.
Before and after, by role
Here is the shape for each role. Fill the brackets with your own numbers, and only numbers you can explain out loud:
- Marketing. Before: "Skills: ChatGPT, Prompt Engineering." After: "Built a Zapier and ChatGPT flow that scores and routes inbound demo requests to reps, cutting lead response time from [X] hours to [Y] minutes."
- Data analysis. Before: "Skills: AI tools." After: "Automated the weekly pipeline report from a CRM export into a written summary with Claude, saving [X] hours a week and removing manual copy errors."
- Operations. Before: "Skills: Generative AI." After: "Set up an intake workflow that classifies vendor requests and drafts first replies, reducing ticket backlog by [X]% in [Y] weeks."
- Product design. Before: "Skills: ChatGPT, AI design tools." After: "Automated design-request intake so briefs arrive with [required fields] filled in, cutting back-and-forth before kickoff by [X] rounds."
If you can't put a real number in a bracket, don't make one up. With 86% of managers already worried about exaggeration, a made-up figure that falls apart in the interview does more damage than leaving the number out. Write the change in plain words instead, like "replaced a manual daily copy-paste step," and be ready to explain it.
You also have to match the bullet to each posting. An ops role wants the backlog bullet up top, and a growth marketing role wants the lead-routing one. Doing that reordering for every application is the work Refolk takes off you: it writes your resume from your own history, and when you paste a posting it rewrites the resume to put the automation work that posting asks for first.
Marketers: you're probably mislabeling, not missing the skill
Many marketers already automate plenty of work. They just label it "ChatGPT" instead of describing the workflow. In the samples from Refolk's index, Marketing Manager is the top title among "Prompt Engineering" profiles (7 of 25) and "ChatGPT" profiles (9 of 25), but only 2 of 25 among "Workflow Automation" profiles. Data Analysts (8) and Operations Managers (6) lead that tag, and Operations Manager doesn't appear among the top titles for prompting at all.
Think about the recurring work in a marketing manager's week:
- Lead routing between forms, CRM and sales.
- Recurring campaign and channel reporting.
- Content pipelines from brief to draft to publish.
Each of these is a workflow, and many marketers have already automated parts of them. For a marketer, the quickest win is usually renaming work you already did rather than taking a course. Find the task you stopped doing by hand and write it up with the four-part formula.
Analysts and ops managers have the opposite problem. They tag the right skill but often bury it as a single word in a skills list. Employers among the "Workflow Automation" profiles in Refolk's index include FedEx, Morgan Stanley and CareFirst BlueCross BlueShield (one profile each, so take them as examples, not a ranking). Automation work shows up in logistics, finance and healthcare. On most resumes it's still a skills-list keyword with no bullet behind it.
How to list AI skills on a resume, section by section
Keep the tool names, but move the proof. Put the name in the skills line so the screening software finds it, and put the result in a bullet so the person reading believes it.
- Skills line. List the specific tools you used, such as Zapier, ChatGPT or Claude. Write "Workflow Automation" as a phrase if you have a bullet that backs it up. Drop "Prompt Engineering" unless you built something reusable, like a prompt library other people use.
- Experience bullets. Each role gets one or two input, tool, output, saved bullets. Put the strongest one first under the role it happened in, because that's where a skimming recruiter looks.
- Summary. Use one clause, not a list: "Operations manager who automates intake and reporting workflows." That reads as a skill, not a tool.
- Interview prep. For each bullet, be ready to explain what broke, how you checked the output, and what it looked like before. That's what the 60% who want proof will ask.
Before you send anything, check whether your automation bullets actually match what the posting asks for. Refolk scores how well your history fits a specific job, which tells you whether to lead with the automation bullet or whether that posting really wants something else.
FAQ
Should I remove ChatGPT from my resume skills section entirely?
No. Keep it as one tool among several, because 58% of hiring managers screen resumes with AI and the tool name helps you match. What you should drop is the idea that the word alone earns credit. After Workday's 25% drop in demand for basic AI skills, "ChatGPT" in the skills line only counts when a bullet below it shows what you did with it.
Is prompt engineering on a resume still worth listing?
Only if you built something reusable, like a shared prompt library, a template set or an internal assistant other people use. In Refolk's index, "Prompt Engineering" already appears on 463 US non-technical profiles, 3.4 times as many as "Workflow Automation," so as a standalone tag it just blends in. Turn it into a bullet with a measured result, like CareerAddict's example of cutting reply drafting from 12 minutes to 4.
What are the best AI skills for a non-technical resume?
The ones Workday saw rise 51%: building AI tools and automating workflows. You don't need to write code. A Zapier flow that routes leads, a ChatGPT step that drafts first replies, or a Claude summary of a weekly export all count, as long as you can name what went in, what came out and what it saved.
What if I've never automated anything?
Pick one task you repeat every week and automate it this month. Then write it up using the four parts. One real workflow you can explain in an interview counts for more than five tool names, because 60% of hiring managers want to see or test your AI ability, and a tool you only experimented with gives them nothing to test.