Are your employees stuck in a Paul Hollywood technical challenge?

A new season of The Great British Bake Off is just around the corner, which means it’s time for a new batch of bakers to face the show’s pesky technical challenges. For those unfamiliar, just imagine being asked to complete a technically complex project, but your boss intentionally omits two-thirds of the instructions. You’re a relative-amateur at this work, but you’re expected to know the basics well enough to produce perfection. You’ll have to use your intuition and that solid head on your shoulders to figure it out. Without more clarity on how to produce the desired result, some bakers will fail. It makes for great television, but a rough way to spend a Tuesday. 

According to employees around the world, some leaders are pulling from the Paul Hollywood playbook, which makes using AI at work feel like a massive technical challenge. 

Employees are experimenting with AI, but the vast majority admit they aren’t seeing leaders connect it to the bigger picture. Our first AI effectiveness benchmark, covering about 112,000 employees around the world, spanning company size, industry, and region, gives us our first glimpse into what employees actually think about AI at work. 

We built a survey template for that

Sensing that the work landscape was shifting, our people scientists built the AI Effectiveness at Work template in 2025. The full template can be used on its own as a 20-question pulse survey, but many companies have chosen to include a few of the AI effectiveness questions in their existing engagement surveys to get a lighter-touch read on the topic.

The AI Effectiveness at Work template measures four factors, representing the aspects of the employee experience most impacted by AI:

• Culture. These are the enablers that make AI adoption possible: leadership support, a clear vision for AI use, and an environment that actually encourages people to try things. 

• Confidence. These check for workforce AI readiness: How capable, motivated, and productive do people feel about working alongside AI? 

• Trust. These look at how much people trust the organization to use AI ethically, transparently, and responsibly. 

• Capability. These assess technical foundations, including systems, processes, access, and training. 

Every benchmark result here is a swath of employees telling us how they feel about something, not a measure of output or hours. But employees’ perceptions of themselves and the business undoubtedly impact how well they actually perform. 

Before getting to the survey results, we need to call out that this sample is likely not representative of all companies around the world. 

The companies sending out AI surveys won't surprise you

Companies that choose to run an AI survey are, almost by definition, further along with AI than companies that don't. Read this as a portrait of AI-engaged organizations, not of every employer on earth. North America accounts for 58% of companies using the template, with EMEA and APAC sitting close together at 22% and 20%. By industry, tech leads by a wide margin, with professional services a distant second. 

By size, this is an enterprise story so far. Organizations of 1,000 or more make up about a third of those using the template, with the 500-1,000 band adding another 20%. Running a survey about AI requires having enough of an AI situation to survey, and it’s more necessary when an organization’s layers grow beyond what leaders can see and observe without a measurement tool. That said, let’s jump into the results! 

How employees rate AI at work, overall

The highest-scoring items in the benchmark collectively suggest that companies have been taking a risk-aware, but exploratory approach to AI so far. Highest-scoring items established in the benchmark speak to:

• risk literacy (86% favorable), 

• encouragement to experiment (85%), and 

• trust in the org to use AI responsibly (80%).     

Tied for the third-highest scoring item is a bit of a head scratcher for us, though. Four in five employees at these AI-forward companies believe that they will help shape how AI is used at their company (80%). While this intuitively makes sense in an environment of exploration and experimentation, it breaks with more traditional top-down strategy patterns, particularly across larger organizations. 

If organizational people strategy needed to pivot due to a merger or acquisition, employees generally would not expect a say in how that change happens or is managed. Here, in the case of AI rollouts, are they expecting a say because they aren’t getting clear direction from the top? The survey items scoring the lowest do point in that direction. 

The lowest-scoring items across the benchmark speak to the support and infrastructure around the AI rollouts. They concern: 

• the accompanying systems and processes and how well they support AI tools (64% favorability), 

• how well managers are sharing examples of how to use AI to meet goals (60%), and

• how clearly leaders connect AI use to company goals (58%). 

However capable employees are, they can’t get around a lack of structural and directional support. System and process change takes real investment, managers walking direct reports through AI takes time, and explaining the plan requires the willingness to say something out loud that may later turn out to be incomplete or wrong. 

Calling out the delta between “go try” and “here’s why”

When we take data from the highest-scoring items and put them against the lowest-scoring items, we have to call out the delta. 

There is a 27% point difference between the share of employees saying they’re encouraged to explore and experiment with AI (“go try it”) and those who say their leaders have connected that AI use to company goals (“here’s why”). 

In all fairness to leaders, though, the challenge of connecting the “here’s why” goes beyond internal communications. Getting your message out may be a part of the issue, but the bigger pain point is the lack of certainty about which roles AI reshapes, which skills hold their value against what AI can do well, and how AI ought to be used to keep a product marketable. 

AI is too new to be predictable at the moment, and leaders can’t really be expected to lay out a clear plan of what they can’t foresee. 

We know that going quiet can feel safer than being wrong. But, Amy Lavoie, our VP of People Science, isn’t so sure the silence buys leaders much.

Amy’s advice may be easier said than done, and it may even be a scary proposition for many leaders. But it’s solid, especially considering our next set of findings based on how often people are using AI. 

Because the global sample includes everyone (AI lovers and haterz alike), averaging them together buries most of what's interesting. This data set tells a story of outliers more than one of a unified middle. Using the templated multiple-choice question about how often employees say they use AI at work, we built a frequency-of-use demographic. 

Our research team standardized all answers into four comparable groups: 

• Non-users (no regular AI use at all), 

• Light users (once or twice a week), 

• Daily users (once a day), and 

• Power users (multiple times a day). 

C-suite leaders are most likely to be power users

Consistent with previous findings, our new AI at work benchmark revealed that by level, AI power use is highest amongst the C-suite. 

One in five C-suite leaders are power users, 5% points higher than those at other levels. So, the same leaders that employees are looking to for explicit instructions on connecting AI use to company goals are the ones most likely to be using it multiple times a day.  

By another view, when we look at habitual users (combining daily users and power users), the C-suite’s margin increases. Three in four C-suite leaders fall into that habitual use category (75.5%). From there, the share of habitual users falls steadily (for the most part) through the rungs. 

So executive leaders are saying “go try it” to their workforce, while experimenting themselves. Clearly, they’re not in the dark, as some employees might conclude after a period of silence.  

Seeing that executive leaders are the most likely to be power and habitual AI users, we wanted to dive a level deeper to understand how the experience of work differs by AI-use frequency. So we did, and we found what we expected. 

Frequency of AI use has a profound impact on employees’ experience with it

It turns out the frequency of your AI use has a profound effect on your experience of it. I know, I know, you’re probably thinking “Duh.” That’s an intuitive conclusion, not a revelation. I agree. Buuuuut, the amount of variance based on use level calls for a deeper look. 

When employees respond with agreement about if AI has helped them learn and grow, the global average is 66%, just above those lowest-scoring items. But, non-users are pulling the average down significantly, with only 27% agreeing (curious given they say they aren’t using AI at all, but I digress). 

Because non-users don’t start at zero, we wouldn’t be so quick as to call them resisters. 41% of non-users still believe that the organization's use of AI will deliver more value to customers. These are the employees with no habit of AI use. Whatever they believe about AI, they didn’t come by it from personal experience. They’ll grant the organizational case at roughly twice the rate they’ll grant the personal one. 

Our takeaway from this is that you don't have to sell non-user employees on AI. They've largely accepted that it's happening and that it will probably be good for the business. What they may not understand is what their own part in it is supposed to be.

AI use doesn’t change everything, though

Across nearly all factors, power users are having a better-than-average experience. We’d be troubled if we saw the reverse trend. But, as we’d expect, people power-using AI at work are more positive about AI. Unsurprisingly, they also report having more support from managers and more communication from leadership about the plan. 

AI usage, on this evidence, has minimal impact on workload, challenging the common belief that AI saves employees loads of time. Favorability for the survey item "Generally, my workload feels reasonable for my role" varies slightly across usage levels, ranging from 69% for non-users to 71–72% for those who do use it. While initial AI adoption slightly improves workload perception, increased frequency of use yields no additional relief. 

In its own right, this isn't the kind of stat that should alarm a leader. Employees' plates adjust as they get more efficient, and that's normal. We wouldn't expect a quilter with a quilting machine to make the same number of quilts per year as one working by hand. The whole point of the machine is to increase the volume the plate can hold. 

Where the added capacity is going may be cause for alarm 

What should concern leaders is where the recovered time is actually going. Rebecca Hinds’s research team at Glean surveyed 6,000 digital employees and found that workers report AI saving them about 11 hours a week, but they then spend 6.4 of those saved hours on “bot-sitting.” The extra time goes to feeding the AI tool more content, checking and correcting outputs, debugging mistakes or hallucinations, and generally editing sloppy-looking work. 

Looking ahead, another key concern is what happens down the road. Without a clearly articulated vision connecting AI use to company goals, power users may spend their extra capacity in their own unique and divergent directions. Or, when employees are aware of and can align with broader organizational goals, that added capacity may become a powerful driver for the business. 

If that alignment piece feels too far off because the future with AI is just too unpredictable for your org right now, you might start by encouraging more open discussions around AI and cataloging how employees are using AI for productivity and efficiency. 

Can we move to “show and tell” time?

Fortunately, employees already report that leaders have cultivated a safe environment for sharing experimental AI results. Globally, three in four employees agree: “At our [Company], we feel comfortable sharing how we use AI in our work.” This came as a bit of a surprise, given the widespread anecdotal reports of workers hiding how much they rely on AI to perform their jobs. 

But we were pleased to see that the majority of orgs are moving swiftly to create or sustain cultures where experimentation and innovation are warmly welcomed. Who doesn’t love “show and tell”?  

In our own process of integrating AI into our workflows at Culture Amp, our leaders have gotten creative. After rolling out company-wide access to Claude at the end of July, our people team led a massive initiative throughout the month of August, which they called “Claugust.” Yes, that is a portmanteau of Claude and August. [Chef’s kiss].

How we’re driving AI change at Culture Amp

Within our month-long AI spectacular, leaders hosted “The Claugies," which was a competition around AI workflow experimentation that produced an AI use case library, now open access to every Culture Amp employee. Entries were scored partly on how easily another Camper (yes, that's what we call ourselves) could pick the workflow up and reuse it.

The library we collectively built is the show and tell. Around sixty entries came through across support, customer success, sales, engineering, talent, and people science. Each entry logged what got built, what problem it solved, and what it saved (be it time, resources, or otherwise). Our Chief People and Customer Engagement Officer, Justin Angsuwat, even has an entry in there, which tracks with what the benchmark told us about who the power users are.

Kissing the Paul Hollywood technical challenge goodbye

On Bake Off, the missing information is deliberate. Paul Hollywood removed certain details, knowing exactly what the end bake should look and taste like. The withholding itself is the test.

That isn't what's happening in your organization, though. Leaders don’t have the full recipe either. Nobody knows exactly how AI will play out. But if you aren’t explicit about the plan, employees will see it as a technical challenge.

Why? Because "I'm holding something back to see what you can do" and "I don't have it either" produce the same silence. About a quarter of employees neither agree nor disagree that their leaders have explained the plan, which usually means people can't tell which one they're witnessing.

What can leaders do now?

Start by doing your own “show and tell.” You're using AI more than almost anyone else in the building. Share what you've built, what you abandoned, and what you still can't figure out. Employees will likely take it far better than they’d take silence from you.

You’ve measured AI adoption; now go beyond that. Ask your employees about their perceptions of AI at work in a formal standalone survey, or sprinkle some AI questions into your existing engagement survey. Hearing where they’re at and identifying where the gaps are in your company gives you the information you need to accelerate AI transformation. 

Once you’ve begun tracking AI usage (via survey or otherwise), identify your power users. Treat them as a unique group, paying attention to their recognition, advancement, and retention figures. They're the group converting capacity into output fastest, which makes them both your strongest lever and your biggest flight risk.

Most importantly, share whatever vision or plan you have, even if it’s still a work in progress. When your most advanced bakers AI power users understand where you’re headed, they can align their efforts with your goals, ensuring their extra capacity drives real value for the business.