What 2026 job data says about AI skills, and how to show yours
Short answer: Basic AI use is becoming normal, so “I can write prompts” says less than it did. Two new reports point the same way. Employers look for people who can build a repeatable AI workflow and check the result. Neither report studies virtual assistants, so treat the findings as a signal and not as a rule.
What the two reports found
Workday published its Global Workforce Report on October 5, 2026. IVADO, a Quebec research institute, published a worker study on October 15, 2026.
| Source | What it measured | Main findings |
|---|---|---|
| Workday report | Job posts from about 550 employers that use Workday Recruiting (September 2025 to July 2026). Surveys of about 6,000 employees. | Demand for basic AI skills, such as simple prompting, peaked in January 2026 and then fell 25%. Mentions of hands-on skills, such as building AI tools and automating workflows, rose 51%. |
| IVADO study | A survey of 1,015 Quebec workers (June 26 to July 2, 2026) and 94 Reddit posts. | 43% of workers use AI at work. 64% of those users use it for less than a quarter of their tasks. 38% of all respondents have no plan to adapt. That figure is 18% for AI users and 54% for non-users. |
What the data does not show
- No virtual assistant data. Workday covers its own customers. IVADO covers Quebec. Neither report gives Philippine data.
- Job posts are not hires. A fall in “prompting” mentions does not prove that people with prompting skills lose work.
- Mentions are not pay. The reports do not say that a skill leads to higher rates.
- Sponsor and survey limits. Workday sells HR software. Both reports use self-reported surveys. They show what people think, not what happened.
- The “51%” group is wide. It includes AI engineering. Only part of it applies to assistant work, such as automating a routine task.
Why this matters for you
If most people can write a prompt, a client cannot tell you apart by that skill. A client can tell you apart by proof that your work is accurate, on time, and easy to hand over. That proof comes from a process you can describe. A tool name alone does not give it.
Example (hypothetical)
This example is not a real client story. An assistant supports an online store owner. The owner sends 30 supplier emails each week. The assistant uses an AI tool to draft replies. Before the change, the owner found two wrong order numbers each month. The assistant now follows this routine:
- The AI tool drafts a reply from a saved template.
- The assistant checks each order number against the order sheet.
- The assistant marks unclear items for the owner and does not guess.
- The assistant logs each week’s errors in a short note.
This routine is easy to explain in an interview. It shows a workflow, a check, and a record.
How to show a skill, not a tool name
Start here
- Pick one task you repeat each week.
- Write the steps in five lines or fewer. Say where the AI helps and where you check.
- Add one measure you can track, such as errors found before sending.
- Put this on your profile in two sentences.
Profile example: “I draft client emails with AI from saved templates. I check every name, date, and number before I send. I track corrections each week.”
Advanced extension
Turn the routine into a reusable checklist or a simple automation. Document the limits: what data it can touch, who approves changes, and what you do when it fails. Do not claim time savings you did not measure.
Limits and trade-offs
A routine takes time to build. A bad routine can spread errors faster than manual work. Start with low-risk tasks. Ask the client before you use their data in any AI tool. Do not share confidential files in a tool that the client has not approved.
Common questions
Should I remove “prompting” from my profile?
No. Keep it, but do not lead with it. Basic prompting is still the base for other skills. Put the result of your process first, then the tool.
Do I need to learn to code?
Not for most assistant roles. The Workday figure includes engineering work. For assistant work, the useful step is often a simple automation, such as a saved template or a form that sends data to a sheet. Learn what your clients need first.
What if my clients do not allow AI tools?
Follow the client’s rule. You can still show a documented process, a checklist, and an error log for manual work. Judgment and reliability matter with or without AI.
How do I know my routine works?
Count one thing for four weeks. Examples: errors caught before sending, or requests that needed a second round of edits. If the count does not fall, change the routine.
Do this week
Write the five-line routine for one repeat task. Send it to a client or peer and ask: “What would you check first?” Update your profile with the result.
Sources
- Workday Global Workforce Report: AI Is Rewriting Jobs More Than It’s Cutting Them. Workday, October 5, 2026. Vendor report. Release read; full report not read.
- AI at Work: Mass Job Displacement Is Not Imminent. IVADO and CRIMT, October 15, 2026. News release; the full report is in French and was not read.
