At some point in the last two years, every company on earth held a meeting about AI. There were slides. There was enthusiasm. Someone said "paradigm shift." And then, in most cases, employees were handed a Claude or ChatGPT subscription and left to work out the rest themselves.
But lobbing prompts at generative AI doesn't automatically make someone an expert. Just like buying a cookbook doesn't automatically make someone a chef. It takes time and guidance to learn a new skill.
And when it comes to effectively using AI, the pressure is definitely on: 77% of business leaders tell us that leveling up their team's AI literacy is urgent to stay competitive.
To find out how companies are actually making that happen, we surveyed over 500 C-suite executives and decision-makers. The verdict? The urgency is real, the intentions are good, but the training infrastructure is…a work in progress.
Key findings:
63% of executives think AI literacy is valuable to their workforce, but most don't formally train employees
Nearly all executives (94%) say their organization is already using AI in some capacity. But "using AI" covers a lot of ground. Here's what that actually looks like in practice:
23% are piloting tools across select teams.
23% are in the early exploration phase.
28% are scaling their AI use across the company.
19% have fully integrated AI into their operations.
As AI becomes a permanent fixture in the tech stack, 63% of enterprise leaders now view AI literacy as either a mandatory core skill (28%) or a valuable asset that isn't formally required (35%).
But while executives say AI literacy matters, the training infrastructure tells a different story. Across every department, a significant share of employees have AI tools at their fingertips and no structured guidance on how to use them. Even IT and engineering teams only have a coin-flip chance of receiving structured AI training.
Department | Formal, structured training is provided | No training, staff are supported with access to paid tools | No training, staff must self-learn with their own resources | AI skills are not expected to be used in this department |
|---|---|---|---|---|
IT / Engineering | 51% | 22% | 12% | 15% |
Business / Operations | 49% | 28% | 11% | 12% |
Finance | 43% | 30% | 13% | 14% |
Sales / Marketing | 43% | 30% | 14% | 13% |
Product | 42% | 29% | 14% | 15% |
HR | 40% | 31% | 14% | 16% |
Legal | 36% | 26% | 18% | 20% |
The skills leaders are prioritizing vary just as much as the training does, and they break down fairly cleanly by function:
IT and engineering teams: Technical AI skill development is the clear priority here, with leaders prioritizing skills for tasks like coding, API management, and model fine-tuning.
Product, sales, marketing, and finance teams: Leaders from these departments all want to see skill building around immediate utility—using AI for content generation, data analysis, or daily productivity.
Business and operations teams: Leaders here are more likely to prioritize big-picture skills like identifying new use cases and ensuring AI is woven into existing business processes.
HR and legal teams: For these departments, the top priority is building AI skills around security, ethics, and governance.
Knowing what skills to build is only half the battle. The other half is figuring out who's actually responsible for building them.
You don't need to wait for your IT team to build something from scratch to bring AI into your workflow. Zapier connects the apps your team already uses, so you can build your own bespoke AI tools without writing a single line of code.
Only 7% say companies put Learning & Development in charge of critical AI training initiatives
Most executives (64%) say their organizations plan to train current employees to build AI skills. But for many companies, the training plan has a few structural challenges baked in.
The first is ownership. Most commonly, 34% of executives say IT and engineering leadership are responsible for defining and maintaining AI skills at their organization. Learning and Development or HR teams are only owning it in 7% of cases.
IT and engineering may be the most likely to understand how these tools work. But what they might not understand is how to teach them.
HR and L&D exist precisely because subject matter expertise and instructional expertise are different skills. Assuming one implies the other is how you end up with a two-hour technical demo that nobody retains. The most effective AI training programs need both teams in the loop.
The second challenge is speed. AI is evolving fast enough that training programs have a shelf life measured in weeks, not years. Overall, 78% of executives say their organization is facing at least one barrier to building a stronger AI-skilled workforce. The most common one, cited by 18%, is that the rapid pace of AI change makes training obsolete almost as soon as it's developed. Other recurring blockers include:
Difficulty identifying which skills are actually needed: 13%
Unclear strategy or ownership of the upskilling mandate: 12%
Lack of time for employees to train: 10%
Lack of budget or resources: 10%
Despite these barriers, the case for training is hard to argue with. Our previous research on AI workslop found that untrained workers are six times more likely to say AI makes them less productive. When handed a powerful tool without the context to use it well, employees don't just fail to gain ground—they lose it.
Training is important, but so is meeting employees where they are. Zapier Agents connect your data and perform tasks across 9,000+ apps, so employees get the benefits of AI in their everyday workflows without needing to become power users overnight.
48% of executives would pay premium salaries for AI skills, but few formally measure AI proficiency
Enterprise leaders clearly value AI skills. Nearly half (48%) are willing to put money behind that conviction, reporting that AI-focused roles are paid more than comparable non-AI roles at their organization. The message to job seekers is fairly straightforward: it's time to learn this stuff.
But paying more for AI skills only makes sense if you can measure whether employees actually have them. That turns out to be a harder problem than it sounds. A quarter of executives say they gauge AI effectiveness through performance and output. Essentially, if the work is better, the AI is working. It's a reasonable proxy, if a somewhat optimistic one.
Others are taking a more structured approach to measuring AI effectiveness: 21% rely on manager or peer review, and 21% use formal assessments or audits.
What's notable is that the hands-off approach may be intentional. Around three-quarters (76%) of executives say they're confident their organization already has the right talent and skills to achieve their AI goals. That's a striking number given everything we've just covered about training gaps, unclear ownership, and rapidly obsolete curricula.
Measuring AI fluency is hard, but it doesn't have to start from scratch. Zapier's AI Fluency Rubric breaks down what AI proficiency looks like across four components, with role-specific benchmarks for every department.
Build AI skills with repeatable systems your teams will actually use
Developing AI maturity doesn't happen by accident. Employees need structure, clear expectations, and a way to see how these tools actually fit into the work they do every day.
The companies getting this right aren't just teaching AI in the abstract. They're building it into the fabric of how work gets done, so using it well becomes the path of least resistance, not an extra step that requires motivation and a free afternoon.
If your team is ready to move from ad hoc AI experiments to something that scales, Zapier can help build customized and trustworthy AI workflows that work with your existing tech stack. Help your team transform AI from a tab open on their screen to the tool that opens up completely new ways of working.
Methodology
The survey was conducted by Centiment for Zapier and fielded between March 30 and March 31, 2026. To qualify, respondents were screened as residents of the United States, aged 18 or older and identified as C-level executives or decision-makers (including, but not limited to, managers, senior managers, directors, VPs, CEOs and CFOs) at organizations with 1,000 or more employees. The data represents 542 qualified respondents and is unweighted with natural fallout across organization size and industry. The margin of error is approximately +/-4% for the total sample at a 95% confidence level.
Related reading:
