Original AISA research · 89 people · September 2026
Most Important AI Skills for Employees [Survey]
Which AI skills for employees matter most? AISA asked 89 people which skill they'd improve across a team. See the results and what they mean.
When asked which single AI skill they would improve across their entire team, over half chose building AI into workflows. AISA asked 89 people starting the AI fluency assessment the question "If you could mass-improve one AI skill across your team, what would it be?" and 51.7% picked workflow integration, more than all other options combined. The result highlights a clear gap between the most important AI skills for employees as people see them and the skills that quietly keep AI output reliable.
- Writing better prompts30.3%27
- Knowing which tool to pick7.9%7
- Verifying AI output4.5%4
- Building AI into workflows51.7%46
- Understanding AI risks5.6%5
89 people answered, 31 August to 7 September 2026. Source: AISA research, aisa.to
What Is the Most Important AI Skill to Build Across a Team?
The skill people most want to spread across a team is the ability to weave AI into day-to-day work, not any single technical capability. That is the headline from AISA's research: 51.7% of the 89 people who answered chose "Building AI into workflows" when given five options.
The Full Picture
Writing better prompts came second at 30.3%. The remaining three options (knowing which tool to pick (7.9%), understanding AI risks (5.6%), and verifying AI output (4.5%)) together account for just 18.0% of votes. The concentration is striking: eight out of ten votes went to either workflows or prompting (82.0% combined).
What This Tells a Manager
If you manage a team and you are deciding where to invest in AI upskilling, this data gives you a starting point. People overwhelmingly feel that the bottleneck is not knowing how AI works in theory, it is knowing how to plug AI into the processes they already run. That instinct aligns with what BCG found in its AI at Work 2025 study: only 36% of employees feel adequately trained in AI use (BCG, June 2025). The gap is practical, not conceptual.
Why Do So Many People Pick Building AI Into Workflows?
Workflow integration won because it sits at the intersection of every other AI skill, you cannot build AI into a process without some prompting ability, some tool awareness, and some sense of where things can go wrong. It is the skill that turns isolated experiments into repeatable value.
From Experiment to Habit
Many teams have individuals who use AI well in isolation. The challenge is making that usage systematic. Task decomposition (breaking a job into steps and deciding which steps benefit from AI) is the core competence inside workflow integration. Without it, AI stays a novelty rather than a default.
The Multiplier Effect
A single person who prompts well produces better individual output. A team that has AI built into its workflows produces better collective output. The 51.7% vote for workflows over the 30.3% vote for prompting suggests people already sense this difference. They want the multiplier, not just the tool.
What Gets Left Behind
The dominance of workflow integration means the three remaining skills (tool selection, output verification, and risk awareness) received minimal attention. Combined, they drew 18.0% of votes. That imbalance matters, and the next sections explain why.
Where Does Prompting Rank?
Prompting placed second at 30.3%, drawing 27 of the 89 votes. It is clearly valued, but it is no longer the skill people see as the single most important lever for a team.
Prompting as a Prerequisite
Writing better prompts is foundational. You need it before you can build reliable workflow automation. But the data suggests people view it more as a prerequisite than a differentiator, important, but not the thing that unlocks team-wide improvement on its own. If your team already has basic prompting ability, doubling down on prompting alone may yield diminishing returns compared to investing in workflow design.
When Prompting Still Deserves Priority
If your team is early in its AI journey (if most people have not moved past copy-pasting questions into a chat window) prompting is the right first investment. AISA's free prompting skills guide covers the fundamentals: iterative refinement, structured output, and context-setting. Once those basics are in place, the conversation naturally shifts to workflows.
Why Do So Few Pick Verifying AI Output?
Only 4.5% of the 89 people chose verifying AI output as the skill they would mass-improve. That is 4 people. It was the least-selected option.
A Dangerous Blind Spot
This matters because verification is the skill that prevents AI from introducing errors at scale. KPMG and the University of Melbourne found in a 2025 global study that 66% of respondents rely on AI output without evaluating its accuracy (KPMG, 2025). When a team builds AI into workflows without building in verification steps, every process that touches AI output becomes a vector for undetected mistakes.
Verification Is Invisible Until It Fails
Part of the reason verification ranks so low may be that it is a defensive skill. It does not produce visible output. It does not speed anything up. It catches problems that, by definition, have not happened yet. Managers rarely get credit for errors that never shipped. But the cost of a hallucination that reaches a customer, a board deck, or a codebase is real. AISA's critical thinking guide walks through practical verification methods (source triangulation, claim decomposition, and stakes-based review) that can be embedded into any workflow.
What This Means for Upskilling Plans
If you are building an AI upskilling programme and you let your team vote on priorities, verification will probably lose. That does not mean it should be deprioritised. It means it needs to be built into the workflow training itself, not offered as a standalone module people can skip.

Curious about your AI Fluency?
AISA helps you measure, prove and improve your AI skills — free report in a 20–40 minute chat.
What About Understanding AI Risks?
Understanding AI risks drew 5.6% of votes, 5 people out of 89. Combined with verification (4.5%), the two safety-adjacent skills account for just 10.1% of responses.
Risk Awareness Is a Team Responsibility
AI risks (data leakage, bias amplification, regulatory exposure) are not hypothetical. The EU AI Act is already in force, and organisations that handle high-risk AI systems face compliance obligations. Yet when people think about what their team needs, risk awareness barely registers. This is not a knowledge gap that fixes itself with time; it requires deliberate programme design.
Pairing Risk With Workflow
The practical move is to embed risk checkpoints inside the workflow training that 51.7% of people are asking for. When you teach someone to build an AI-assisted process, you also teach them where to check for PII exposure, where to add a human review step, and when to avoid AI entirely. AISA's responsible AI use guide provides a framework for exactly this.
How Should You Prioritise AI Upskilling?
Start with what people want (workflow integration) but layer in what they are overlooking: verification and risk awareness. The data from AISA's research gives you a clear picture of demand. Your job as a manager is to balance demand with need.
A Practical Sequencing Framework
The table below maps each skill to its vote share, its risk if neglected, and a suggested priority tier.
| Skill | Vote share | Risk if neglected | Suggested tier |
|---|---|---|---|
| Building AI into workflows | 51.7% | AI stays experimental, no team-wide value | Tier 1, lead with this |
| Writing better prompts | 30.3% | Poor input quality limits every downstream step | Tier 1, teach alongside workflows |
| Verifying AI output | 4.5% | Errors scale with adoption; trust erodes | Tier 1, embed inside workflow training |
| Knowing which tool to pick | 7.9% | Teams default to one tool regardless of fit | Tier 2, address once workflows are running |
| Understanding AI risks | 5.6% | Compliance gaps, data exposure, reputational harm | Tier 2, integrate into policy and workflow checkpoints |
Measure Before You Train
Before you invest in any upskilling programme, you need a baseline. What does your team actually know today? Where are the gaps? A team AI assessment gives you scored, evidence-based data on each person's strengths and weaknesses across prompting, critical thinking, technical understanding, workflow application, and safety. That baseline turns a generic training plan into a targeted one.
Do Not Let Popularity Drive the Curriculum
The 51.7% vote for workflows is a signal, not a syllabus. If you build a programme that is 100% workflow design and 0% verification, you are building speed without brakes. The strongest upskilling programmes treat verification and risk as integral parts of workflow competence, not optional extras.
Track Progress, Not Just Completion
Course completion rates tell you who showed up. They do not tell you who improved. Reassessing your team after training (using the same measurement framework you used for the baseline) shows you whether the investment moved the needle. AISA's assessment is designed for exactly this: 98% of people who retake it land in the same or an adjacent tier, which means score changes reflect genuine skill movement, not measurement noise.
Connect Upskilling to Barriers
If your team is not adopting AI despite training, the problem may not be skill, it may be barriers to adoption like unclear policies, tool access, or lack of management support. Skill training works best when those structural blockers are addressed in parallel.
About this research
This finding comes from AISA's own research. AISA asks a short research question of people as they start the AISA AI fluency assessment, one question at a time. The question in the chart above was asked from 31 August to 7 September 2026, worded exactly as shown and with the answers in the order shown, and 89 people answered it. Each person counts once: when someone answered more than once, their latest answer is the one counted.
More findings from this research are collected in AISA research and data.
How to cite this research: AISA (2026). "Most Important AI Skills for Employees [Survey]". AISA research. https://aisa.to/blog/most-important-ai-skills-for-employees
Related reading: Barriers to AI Adoption at Work [2026 Data], what actually stops people from using AI at work.
Related reading: Critical Thinking With AI: How Good Are You?, why checking AI output is harder than people think.
Related reading: What AI Skill to Learn First? [2026 Data], how individuals rank the same question for themselves.
Frequently Asked Questions
What AI Skills Do Employees Need?
Employees need a combination of prompting, workflow integration, output verification, tool selection, and risk awareness. AISA's research shows that people overwhelmingly prioritise workflow integration (51.7%) and prompting (30.3%), but verification and risk awareness are equally important for safe, reliable AI use at scale.
Which AI Skill Should a Team Learn First?
If the team has basic prompting ability, workflow integration is the skill that unlocks the most collective value, 51.7% of the people AISA asked chose it as the single skill they would mass-improve. If the team is new to AI, start with prompting fundamentals and layer workflow design on top.
How Do You Check AI Output at Work?
Checking AI output involves verifying claims against original sources, decomposing complex outputs into individually checkable statements, and applying a stakes-based review where higher-consequence outputs get more scrutiny. AISA's critical thinking guide covers these methods in detail. Despite its importance, only 4.5% of people in AISA's research chose verification as the top skill to improve across a team.

Curious about your AI Fluency?
AISA helps you measure, prove and improve your AI skills — free report in a 20–40 minute chat.

