Zulfugar Graphics // Shutterstock
Companies spend six figures on AI. A third of employees don’t know what it costs
Every year, someone in finance asks whether the company’s AI spend is worth it. Every year, leadership says yes. And every year, the employees actually using the tools have no idea the question was even asked. Now, corporate America is starting to sweat over its AI bill.
Microsoft canceled most of its Claude Code licenses, in part over costs. Uber’s COO said AI expenses are getting harder to justify. One AI consultant reported that a client spent half a billion dollars in a single month after failing to put usage limits on employee licenses.
Most companies haven’t gotten that memo. Or if they did, they summarized it with AI and never read the output. According to a Zapier survey of 715 U.S. professionals conducted in April and May 2026, 91% of managers and above say their AI tools are clearly worth the cost, and 86% plan to spend even more over the next 12 months.
The confidence is real, but it’s concentrated at the top. According to the Zapier survey, nearly 37% of individual contributors (IC) either don’t think about the cost of AI usage (25.2%) or don’t know their usage has a price tag at all (11.7%). Leadership is doubling down on a bet that most of their workforce can’t see, let alone evaluate. That disconnect between boardroom conviction and ground-level awareness is exactly where the AI ROI story gets complicated.
Key findings:
- 52% of managers say their organizations spend more than $100K per month on AI-related tools.
- Only 34% of individual contributors know where the AI budget actually goes.
- Security concerns are the biggest barrier to getting more from their AI spend, cited by 39% of organizations.
- Companies growing their AI budgets plan to invest in training, integration, and new licenses equally.
52% of managers say their organizations spend more than $100K per month on AI-related tools
The financial anxiety making headlines hasn’t slowed most organizations down. If anything, budgets are moving in the opposite direction.
Apparently, nobody forwarded those headlines to the finance team. Nine out of 10 (91%) managers and above—a group that includes managers, senior managers, VPs, and C-suite executives—say AI tools are clearly worth what they’re paying. And they’re putting real money behind that conviction.
More than half (52%) of survey respondents say their organizations now spend over $100,000 per month on tools purchased specifically for their AI capabilities. And 40% have routed more than half of their total software budget to AI, which, if you’ve ever seen what “the software budget” has to cover at a 500-person company, is not a casual reallocation.
Zapier
That spending trajectory is only going up—86% of all respondents plan to increase AI investment over the next 12 months, and half of those expect growth of 50% or more.
Only 34% of individual contributors know where the AI budget actually goes
Budget visibility drops sharply the further you get from the executive level. C-suite respondents and company owners are 2.5 times more likely than individual contributors to have direct visibility into how AI dollars are being spent.
Eighty-five percent of executives say they have a clear view of where AI dollars go. But only 34% of ICs say the same, and 23% say, with admirable candor, that it’s simply not their job to know.
Zapier
When the people using AI every day don’t know what it costs or why it was purchased in the first place, they can’t reasonably be expected to use it in ways that actually justify the investment. Recent conversations around tokenmaxxing, a trend where companies treat the sheer volume of AI activity as a measure of success, illustrate the exact problem. Treating AI usage as a volume game rather than a value game leaves leadership with no real way to tell whether the tools are being used well or just used a lot.
Nearly 37% of individual contributors either don’t think about what their AI usage costs (25.2%) or don’t know it has a cost at all (11.7%). By contrast, 75% of C-suite respondents actively think about cost efficiency when using AI tools, meaning leadership is far more mindful of spend on a daily basis than the people actually driving that spend through their usage.
The result is a spending structure where budgets keep growing, confidence stays high at the top, and nobody in the boardroom actually knows if their employees are doing meaningful work with AI or just burning tokens to summarize their own emails.
Security concerns are the biggest barrier to getting more from their AI spend, cited by 39% of organizations
Leadership conviction does not, as it turns out, solve the hard problems. Most organizations are still running into real obstacles that limit the returns on what they’re spending.
Security, compliance, and governance concerns top the list, cited by 39% of respondents as the single biggest challenge to getting more value from their paid AI tools. Data readiness issues (36%) and lack of tool integration (30%) are right behind. These are organizations spending six figures a month and planning to spend even more, and they’re still hitting the same walls. Budget size alone isn’t solving it.
Only 17% say they have no challenges at all. The other 83% have some notes.
The security and governance numbers stand out for a reason. As more employees gain access to AI tools, open questions about their impact stack up:
- Who has access to what?
- What data are agents touching?
- Are those actions auditable?
Without controls in place across employees, AI agents, and company systems, most organizations have no reliable way to answer any of those questions.
That same accountability problem shows up in the integration data. More than half (52%) of organizations already pay for AI automation and agents, yet almost 1 in 3 still cite a lack of tool integration as a top barrier. Paying for automation and actually governing how that automation runs across your systems are two different problems, and right now, most companies are only solving one of them.
Companies growing their AI budgets plan to invest in training, integration, and new licenses equally
The next round of AI budget increases isn’t going entirely toward more tools and more tokens. Among organizations increasing their AI budgets over the next 12 months, employee training (54%), AI automation and agents (53%), and software licenses (53%) are essentially tied as the top priorities. The next wave of AI investment looks less like a shopping spree and more like an infrastructure build.
That training focus makes sense given what happens without it. Zapier’s 2026 AI workslop survey found that untrained workers are six times more likely to say AI makes them less productive. Handing someone a powerful tool with no context for how to use it doesn’t just fail to deliver returns—it can actively drag productivity in the wrong direction.
And despite cost pressures, governance challenges, and the disconnect in visibility between leadership and the people actually using the tools, most organizations aren’t second-guessing their direction. Seventy-seven percent said they would make the same AI investments again, and nearly half said they would have moved even faster from the start.
Leadership is confident, and budgets are growing, but that confidence doesn’t automatically translate into accountability or visibility across the rest of the organization. The bigger the AI budget gets, the more that blind spot costs.
Methodology
The survey was conducted by Centiment on behalf of Zapier. The survey was fielded between April 28, 2026, and May 18, 2026. The results are based on 715 completed surveys. To qualify, respondents were screened to be U.S. professionals who work for a company with 500-plus employees and have direct knowledge of their company’s AI initiatives. The company must use at least three paid AI tools (e.g., ChatGPT Enterprise, Claude for Business, OpenAI/Anthropic/Google Cloud APIs). The respondents were screened to create a 50/50 split (within 5%) between individual contributors and those at the manager level or above. Data is unweighted, and the margin of error is approximately +/-3.6% for the overall sample with a 95% confidence level.
This story was produced by Zapier and reviewed and distributed by Stacker.
