On October 5, 2026, Workday, which makes HR and finance software, released its “Global Workforce Report.” Workday pooled the skill requirements that 554 large companies using its recruiting system listed when hiring. Requirements for knowing how to use AI tools peaked in January 2026 and then declined. In the same data, requirements for knowing how to build AI and plug it into work rose for all 11 months. The press release subtitle summed up the shift: companies are looking for people who can build with AI, not just people who prompt it.
Eleven months of hiring requests from 554 large employers show AI skill demand splitting in two
The data comes from skills listed on job requisitions in Workday Recruiting, which Workday collected with customer names removed. A job requisition is the document a company files in its hiring system to fill one open role, spelling out the skills it needs. The period covers 11 months, from September 2025 to July 2026. The report sorted the AI skill requirements on those requisitions into two groups.
| Report's category | Report's definition | What this article calls it |
|---|---|---|
| Basic AI skills | Everyday use of AI tools, from prompting to using copilots | Skills for using AI |
| Applied AI skills | Building, integrating, and running AI through machine learning, workflow automation, LLMs, and AI engineering | Skills for building AI into work |
The group for using AI covers not only writing prompts but also using copilots on the job. The group for building AI into work covers building AI as well as connecting it to workflows and keeping it running.
There are limits to keep in mind. The full report does not say whether the changes are measured in number of requisitions, share of requisitions, or number of skill mentions. The public materials also do not explain how individual skills were assigned to each group. The sample is Workday Recruiting customers with 250 or more employees, so it does not represent the whole labor market. Workday, the publisher, also sells recruiting software, skills management products, and Sana, a tool for building agents.
Skills for using AI tools, like prompting and copilots, fell 25% after a January peak. Skills for building AI into work, like machine learning and workflow automation, rose 51% from September 2025
Companies asked most often for skills in using AI in January 2026. In the report's words, demand then fell “steadily,” dropping 25% overall. Demand for skills in building AI into work “rose every period” and ended 51% above September 2025. The public materials do not say whether a “period” is a month or a quarter.
The two numbers use different baselines. The 25% is the drop from the January 2026 peak. The 51% is the rise from the September 2025 starting point. So their sizes can't be compared on the same scale. What the data does show clearly is the direction each group moved. Over the six months from January to July 2026, companies listed skills in using AI less often on requisitions and skills in building AI into work more often.
The data can't tell whether using AI matters less or is now simply assumed
Requisition data shows what companies wrote down. It does not show why they stopped writing something. The report's methodology says all trends should be read as descriptive, not causal. Learning News, an education industry outlet, made the same point after reporting the figures.
“Advertised requirements do not establish whether basic AI skills are becoming less useful or more widely assumed.”
If the first reading is right, the ability to use AI tools has become something companies weigh less in hiring. If the second is right, it has become so expected that nobody bothers to write it down. The report offers no evidence to settle which is true, and it does not argue for either.
It is also not disclosed whether the 25% refers to requisition counts or shares. If it is based on counts, part of the drop may reflect less hiring overall. Using different data (Workday HR system usage), the same report found that a typical company's monthly job openings fell 4%, from about 83 in March 2025 to about 79 in July 2026. Because the periods and data differ, you can't subtract that figure from the 25%. Over the same 11 months, demand for skills in building AI into work went up, not down.
Only the tech sector asked for more of both: using AI and building with it
Across all 554 companies, demand for skills in using AI fell after January, but the tech sector broke from that trend. The report names tech as the exception, where demand grew for talent who can use AI well and also build with it. The original line reads: “Technology is the exception, with rising demand for talent who can both use AI fluently and build with it.”
That sentence is all the report gives. It has no numbers for tech or for any other sector. So we can't tell how much demand for skills in using AI grew in tech, or how much steeper the decline would be without tech. What the sentence does support is this: in tech, skills for building AI into work were added on top of skills for using AI, not in place of them.
Whatever the reason, large employers are writing “can use AI” less and “can build with AI” more
The 554 companies are Workday Recruiting customers with 250 or more employees, and the report presents the trend as descriptive only. Within that scope, the AI requirement that kept growing on requisitions after January 2026 was the skill of building AI into work.
What building AI into work looks like on the job shows up in a blog post Workday published four days after the report to introduce Sana, its own agent tool. According to the post, an HR business partner (HRBP) at Workday built a prototype calibration agent with Sana without going to the engineering team. Calibration is the process of aligning the performance ratings managers give across teams, and the agent was trained on questions posted in the company's internal Slack. Of Workday's HR business partners, 83% said the agent was very useful, and half of its users cut calibration prep time by up to 75%. The post does not say how many people responded.
The same post describes an interview question agent built by one recruiting team manager. Workday has more than 1,600 roles and a list of over 800 skills. For each role, the agent writes two tailored interview questions for every skill that role requires.
“an agent built by one person does it for all of them.”
Both agents were built with Sana by employees who are not developers. The public materials do not say whether this kind of building counts toward the report's applied AI skills. The post is also a product piece from Workday, which sells Sana, so read it as an example of what building AI into work looks like, not as evidence behind the report's numbers.
Through January 2026, large companies asked most often whether candidates could use AI tools. That requirement then shrank on requisitions. The AI requirement still growing on large companies' requisitions today is whether candidates can build AI into their work.
Further reading
Global Workforce Report: The Adaptability Advantage The full report PDF, with the original definitions of basic and applied AI skills (Chapter 3, pp. 11–13) and the methodology (p. 22). forms.workday.com
Are Jobs Report Headlines Missing the Plot? Workday guest economist Aaron Terrazas walks through the same data alongside broader hiring trends. workday.com
Jobs Mentioning AI Are Growing Amid Broader Hiring Weakness Indeed Hiring Lab's January 2026 labor market update, which found that U.S. Indeed postings mentioning AI or AI-related terms grew more than 130% through late 2025. Unlike Workday, it counted every mention of AI. hiringlab.indeed.com