Bosses rate their AI rollout 17 points higher than their staff do. What that means for your office

What's new

On Oct. 13, Asana, the work-management software company, released its Australian AI Execution Index. It surveyed 1,000 people through the research panel Dynata: 300 senior decision-makers and 700 knowledge workers, as reported by IT Brief Australia.

Almost everyone is using AI. 94% reported meaningful AI use in their organization, and 91% reported at least one measurable improvement in the past year. But only 22% met Asana's bar for an "AI Executor," a company that pairs advanced use with measurable gains across several business outcomes. On average, respondents saw improvement in 3.1 of the 10 outcomes the survey measured.

Two findings matter for a mid-size company: the people at the top see a different picture than the people at the desks, and smaller firms are doing better than big ones at the part that counts.

The numbers

Smaller firms score higher on putting AI to work

Asana index scores out of 100, by company size. Smaller means 10 to 999 employees; large means 1,000 or more. Self-reported survey answers, not an audit.

1. AI execution Smaller firms10-999 employees64.9
2. AI execution Large firms1,000+ employees61.7
3. Workflow integration Smaller firms10-999 employees63.5
4. Workflow integration Large firms1,000+ employees58.8
5. AI agent readiness Smaller firms10-999 employees61.2
6. AI agent readiness Large firms1,000+ employees56.3

Source: Asana Australian AI Execution Index 2026, as reported by IT Brief Australia, Oct. 13, 2026 (1,000 respondents).

The size finding is the surprise. Large firms were more likely to say AI is used across the whole organization or fully embedded: 50% versus 42% for small and mid-size firms. Yet the smaller firms scored higher on every index IT Brief reported. Asana's own summary makes the same point: the companies using AI most widely aren't always the ones getting the most out of it.

The view from the top and the view from the desk

The gap between bosses and staff was about 23 points at companies with 1,000 or more employees and about 12 points at firms with 10 to 999, according to IT Brief. That fits what you'd expect. In a big company, the person approving the AI budget may be four layers away from the person using it. In a 150-person distributor, the owner can walk to the order desk.

Training shows up too. At the companies Asana counts as AI Executors, 59% of workers said they'd had thorough training in practical and responsible AI use. At the companies Asana classes as AI Adopters, 29% said so.

Leaders see the purchase. The people at the desk see whether it helps.

What the other side has right

This survey has real limits, and they're worth saying plainly.

• It's vendor research. Asana sells software for organizing work, and its conclusions point toward better workflow tools.
• It's Australian. The companies, labor market and tools mix aren't the same as Dallas-Fort Worth.
• It's self-reported. IT Brief notes the scores reflect respondents' own answers, not an outside audit. The executives and workers weren't from matched companies, so the 17-point gap compares two groups, not bosses and their own staff.
• Executives may know things workers don't. A leader can see projects still being built, or results in finance that never reach the floor. A low worker score doesn't prove the rollout failed.

Asana's page doesn't give fieldwork dates. We'd treat the exact numbers loosely and the direction more seriously, especially since it matches U.S. data. In the Federal Reserve's survey of small employer firms, 46% said they use AI, but only 7% of users said it was fully integrated into their business. The top problem users named was accuracy, at 46%.

Our take for a 40-400 person company

The finding that smaller firms execute better isn't surprising to anyone who has worked in one. You have fewer layers, fewer committees, and the owner often knows every customer's PO format by heart. That's an advantage, but only if you use it. A license bought from the top and handed out with a one-line email gets you the same gap the big companies have.

Asana's general manager for the region, Jo Gaines, put the point simply: "Buying AI tools is the easy part." The hard part is one workflow at a time: the order desk, AP, quoting. That means writing down the steps, connecting the tool to the systems the work lands in, deciding who approves exceptions, and timing it before and after. We wrote about why generic rollouts stall in why AI pilots fail.

The unapproved-tools finding deserves attention too. If 72% of executives in the survey know staff use tools nobody approved, assume it's happening in your office. That isn't a reason to ban AI. It's a reason to find out which tools people picked, what customer or pricing data goes into them, and whether a supported option would serve them better.

What changes, and what doesn't

What to do this month

1. Ask the desk, not the managers. Five minutes with each person who handles orders, invoices or quotes: what AI do you use, for what, and does it save time?
2. List the tools people already use, including the free ones, and what data goes in. Decide what's fine and what isn't, and write it down.
3. Pick one workflow and time it. Minutes per order or per invoice, for a week. That's your baseline.
4. Train the people on that workflow on the actual task, with your documents, not a general AI course.
5. Measure again in 60 days. If the people doing the work can't feel the difference, it isn't working yet, whatever the dashboard says. For the bigger picture, see ChatGPT vs. custom AI for business processes.