June 2026
The AI Proficiency Report
Letter from the CEO
I live in San Francisco, where every bus is plastered with an ad for some agentic technology startup. But it’s not just the buses. In every board room and c-suite, there is widespread delusion about enterprise AI, especially agents. If you listened to earnings calls, you’d think that every company in the world was running on agents and the AI value problem was solved.
Our most recent AI Proficiency Report paints a much more challenging picture. Most workers (69%) report some type of agentic action from their organization – giving access to agent-capable tools, encouraging employees to build AI agents, etc.
But only 16% of workers say they use the agentic tools they have access to. Just 13% are building AI automations. And when asked to define an AI agent in their own words, less than 10% know what one is.
Unsurprisingly, the C-Suite is still delusional. 57% say AI is widely integrated in workflows, compared to 18% of ICs. Execs are 2x more likely to think the organization feels positive about AI.
When we ran the first AI Proficiency Report in 2024, the headline was: Few organizations are ready to deploy AI, because most employees are untrained and unprepared. Today, the AIs are smarter but the workforce is still not ready. Enterprises continue to invest in AI licenses and tools while drastically under-investing in what we call “the transformation layer.” CEOs brag at the All Hands and the earnings call, but don’t back it up with resources for their workforces.
It’s even more obvious now that AI-powered organizations will have a meaningful advantage. We’re already seeing it with the first “supercompanies” - organizations like Anthropic, which run on AI with a fraction of the number of employees of SaaS companies. But transformation starts with empowering the workforce, and there, enterprise organizations still have a lot of work to do.
Our most recent AI Proficiency Report paints a much more challenging picture. Most workers (69%) report some type of agentic action from their organization – giving access to agent-capable tools, encouraging employees to build AI agents, etc.
But only 16% of workers say they use the agentic tools they have access to. Just 13% are building AI automations. And when asked to define an AI agent in their own words, less than 10% know what one is.
Unsurprisingly, the C-Suite is still delusional. 57% say AI is widely integrated in workflows, compared to 18% of ICs. Execs are 2x more likely to think the organization feels positive about AI.
When we ran the first AI Proficiency Report in 2024, the headline was: Few organizations are ready to deploy AI, because most employees are untrained and unprepared. Today, the AIs are smarter but the workforce is still not ready. Enterprises continue to invest in AI licenses and tools while drastically under-investing in what we call “the transformation layer.” CEOs brag at the All Hands and the earnings call, but don’t back it up with resources for their workforces.
It’s even more obvious now that AI-powered organizations will have a meaningful advantage. We’re already seeing it with the first “supercompanies” - organizations like Anthropic, which run on AI with a fraction of the number of employees of SaaS companies. But transformation starts with empowering the workforce, and there, enterprise organizations still have a lot of work to do.
Greg Shove
Section CEO
Our methodology
We surveyed 5,026 U.S. knowledge workers across industries and functions and analyzed the following characteristics and behaviors.
1
AI knowledge
Understanding of how AI works, its limitations, and how to use AI tools safely to protect data and mitigate bias
2
AI usage
Frequency, depth, and sophistication of use, including their most valuable use cases and typical behaviors when using AI
3
AI skill
Ability to prompt and build automations effectively and identify high-value applications of AI, measured by hands-on tests rather than self-reporting
4
AI attitudes
Feelings about AI and its impact on their work
5
Organizational AI readiness
Company actions to encourage or discourage the use of AI, including manager support, company strategy, AI policies, chat and agentic deployment, and training
Based on the individuals’ usage, knowledge, and skill scores, we placed them into four levels of AI proficiency.
Level 1
Novice
Does not have the skills and knowledge needed to get value out of AI.
Level 2
Experimenter
Uses AI frequently, but for one-off tasks – no repeatable workflows yet.
Proficient
Level 3
Practitioner
AI is embedded into how they work – into repeatable workflows.
Advanced
Level 4
Expert
Builds AI workflows for themselves and others – automation, agents, new use cases.
Want to know vour
Al proficiency?
What we found
Key Finding 1
Agents are here.
Agentic readiness is not.
69%
of workers report their org has taken some action on AI agents
16%
actually use an agentic tool at work
<10%
can correctly define an AI agent in their own words
The enterprise has moved. Most workers (69%) report that their organizations have done something related to agents — given them access to agent-capable tools, encouraged them to build or use AI agents, deployed agents for a business process, or provided some training. Only 22% say their organization has done nothing on agents at all.
But what's happening below the surface tells a different story.
Only 16% of workers say they actually use an agentic tool – platforms like Claude Code, Copilot Studio, Glean, or similar. Only 13% have set up any kind of automated workflow as their most advanced AI use.
But what's happening below the surface tells a different story.
Only 16% of workers say they actually use an agentic tool – platforms like Claude Code, Copilot Studio, Glean, or similar. Only 13% have set up any kind of automated workflow as their most advanced AI use.
When we ask workers to define an AI agent in their own words, fewer than 10% give a directionally correct answer. On our structured knowledge test – multiple choice – 82% of the workforce cannot correctly identify what an agent is.
Even among organizations that have deployed agents, the access-to-enablement gap is large. Only ⅓ of employees in “agentic” organizations have received training on using agents.
Agents are deployed unevenly across levels. C-suite executives are more than twice as likely as ICs to report their organization has given them access to agent-capable tools and five times as likely to have received agentic training.
The companies deploying agents alongside real enablement are already separating. Workers at organizations that have deployed agents for business processes and trained their people show significantly higher proficiency scores than those at organizations that haven't.
Key Finding 2
More people are using AI every day. They’re still not that good at it.
67% of knowledge workers now use AI at least once a week, up from 55% in October 2025. Daily use has risen from 22.5% to 36.7%.
67%
of knowledge workers use
AI at least once a week
AI at least once a week
Up from 55% (October 2025)
More people have value-adding use cases than they did last year – defined as use cases likely to drive ROI for the business through significant time savings, revenue impact, or other KPI drivers. The % of employees with no single AI use case dropped 72% in the last six months.
But only 5.5% of the workforce meets the bar for AI proficiency today. The vast majority (73.5%) are Experimenters: people who use AI for basic, one-off tasks. Another 20.9% are Novices who barely engage with AI.
Most use cases are still quite basic. The most commonly reported use case (61% of workers) is essentially “looking things up” - using AI as a replacement for Google. Half of workers say they use AI to revise what they’ve written, but only a third use it to create first drafts.
What people are struggling with most: building. Prompting is improving – 21% of workers can write an effective prompt today, up from 7% in October 2025.
But almost no one can write good instructions for an AI assistant (a Custom GPT, Gemini Gem, Copilot Agent, or similar) – a foundational skill for agent-building. Only 3.8% of respondents provided workable assistant instructions in our hands-on testing.
3.8%
of people can write workable
AI assistant instructions
(scored as 3 out of 5 or higher)
AI experts are a small group - but they do exist. This is what they look like.
They think in workflows, not tasks
“Research the topic → draft an outline → refine messaging → simulate stakeholder Q&A”
They’ve automated something real
“I built a deck creation GPT that turns training PPTs from raw spec sheets into decks for new hires”
They’ve set up AI to monitor and act on their behalf
6/10 experts have set an AI to monitor data and trigger an action
Their job has actually changed
9/10 experts said:
“I’ve taken on new types of tasks I didn’t do before”
Download the report for free and get the full status of AI proficiency in 2026.
Key Finding 3
Executives are excited. Everyone else, not so much.
82% of C-suite workers describe themselves as excited about AI. 30% of individual contributors say they’re excited. And the inverse is true: 8% of C-suite workers feel anxious or overwhelmed. 22% of ICs do.
This data reflects different experiences of AI at work – driven by dramatically different levels of access, support, and organizational investment.
Key Finding 4
AI transformation is dying with managers
Managers are the connective tissue between an organization's AI strategy and the workforce that needs to execute it. When managers set expectations, demonstrate AI use, and hold their teams accountable, workers are significantly more proficient. When managers don't – which is most of the time – the strategy stays at the top of the org chart.
Manager support is one of the most important things an organization can do to drive AI proficiency and excitement. Employees whose managers expect and require AI use score 1.5x higher in AI proficiency and are 2.7x more likely to be excited about AI, than those who have no expectations.
Most managers aren’t using AI well themselves. Only 33% of managers use AI daily, and less than half are excited about AI. Their average proficiency is barely higher than the ICs they manage. This creates a foundational problem - managers aren’t excited enough to motivate their teams, and they aren’t proficient enough to guide direct reports in using AI effectively.
Managers are mostly silent or weak on AI. 65% of managers either don’t set any expectations about AI, or encourage it but don’t hold employees accountable. Only 7% of managers require AI use and tie it to performance evaluations.
Most managers haven’t done anything to motivate their team’s AI use in the last month. Only 21.3% of managers have demonstrated their own AI use to their team in the past 30 days. 37.5% did none of the following: demonstrate AI use, ask a team member to use AI for a task, or discuss AI goals or expectations.
Want to know your
Al proficiency?
Key Finding 5
Companies are making moves - but there are still gaps
Companies are moving (slowly) in the right direction. 62% of workers say they’ve received training, up 40% from six months ago. Slightly more workers have access to tools - up to 53% from 50% in October 2025.
But the data also reveals significant gaps.
Training covers compliance but not building. 37.8% of workers have received no AI training at all. Of those who have, 15% only received training on safety and compliance. Only 17% were trained on agents or automations – skills that are now the bar for AI proficiency. Workers at organizations that have provided agent-specific training score an average of 44 out of 100, compared to 25 at organizations that haven't – 1.8x higher.
Leadership calls AI a priority but hasn't appointed anyone to own it. 53.2% of workers say their organization either has no Head of AI or they're not sure if one exists. Workers at organizations with a full-time Head of AI score an average of 42 in AI proficiency, compared to 33 at organizations without one.
Tool access is still a problem. Only 53.2% of workers have clear, sanctioned access to AI tools. Workers with clear tool access score an average of 43 out of 100, compared to 25 for workers at organizations with no sanctioned tools – 1.7x higher.
Formal strategy matters more than most organizations realize. Workers at organizations with a formal AI strategy score an average of 43 out of 100, compared to 26 at organizations with no strategy – 1.6x higher.
Training covers compliance but not building. 37.8% of workers have received no AI training at all. Of those who have, 15% only received training on safety and compliance. Only 17% were trained on agents or automations – skills that are now the bar for AI proficiency. Workers at organizations that have provided agent-specific training score an average of 44 out of 100, compared to 25 at organizations that haven't – 1.8x higher.
Leadership calls AI a priority but hasn't appointed anyone to own it. 53.2% of workers say their organization either has no Head of AI or they're not sure if one exists. Workers at organizations with a full-time Head of AI score an average of 42 in AI proficiency, compared to 33 at organizations without one.
Tool access is still a problem. Only 53.2% of workers have clear, sanctioned access to AI tools. Workers with clear tool access score an average of 43 out of 100, compared to 25 for workers at organizations with no sanctioned tools – 1.7x higher.
Formal strategy matters more than most organizations realize. Workers at organizations with a formal AI strategy score an average of 43 out of 100, compared to 26 at organizations with no strategy – 1.6x higher.
Download the report for free and get the full status of AI proficiency in 2026.
What leaders need to do
5 tactics proven to drive AI value
Don't deploy agents without enablement.
82% of your workforce can't correctly describe what an agent is. Before your next agent rollout, answer three questions: Do workers understand what agents do? Do they know how to use them safely and effectively? Is there a governance structure for building and maintaining AI agents?
Audit tool access before you do anything else.
47% of your workforce either lacks clear access to sanctioned AI tools or doesn't know how to get access to them. Workers with clear tool access score an average of 43 out of 100, compared to 25 for workers at organizations with no sanctioned tools. Get this sorted before investing in anything else.
Make manager AI accountability real.
Workers whose managers expect AI use are significantly more proficient than workers whose managers don't, but only 7.7% of managers tie AI use to performance evaluations. Don’t let change die at the manager level – make team AI fluency an expectation of managers and build it into their performance metrics.
Train people on building with AI.
Safety and compliance is the most common AI training topic in the workforce today. That made sense in 2023. In 2026, with agents being deployed, the training priorities need to shift: agent literacy, function-specific use cases, workflow integration. If your training program is still primarily about what not to do with AI, it's at least a year behind where it needs to be.
Prioritize the IC population before
agents make the gap permanent.
agents make the gap permanent.
Individual contributors are 3.3% proficient, least supported by managers, and least likely to have received training. They also have the most repetitive, automatable work, which means the upside of getting them proficient is significant. This is where AI transformation needs to focus, and it's the population most organizations are neglecting.
