Among people who say their team doesn't, it's 10%.
The finding
Since 2021 we've surveyed thousands of staff and leaders in about 100 organizations. This finding comes from our newest survey, The NoW AI Team Index, which we run with staff before and during our work with an organization. People rate statements from 0 (strongly disagree) to 10 (strongly agree). Two of those statements are:
- "My team has practical examples of how AI can improve our work."
- "AI has become a regular part of how I complete my work."
We grouped people by how they answered the first (agree is 7 to 10, disagree is 0 to 3) and looked at how many agreed with the second.
Team has practical AI examples
61%agree their team has examples (7 to 10)Team doesn't
10%disagree (0 to 3)Source: The NoW AI Team Index, July to October 2026
The gap is roughly six to one. People in the middle, who scored their team 4 to 6 on examples, sit at 37%. The more clearly a team has examples, the more often AI is part of the day.
The same split shows up on value. Of people whose team has examples, 71% agree "AI is helping me produce better outcomes." Of people whose team doesn't, 30% do.
Team has practical AI examples
Team doesn't
Source: The NoW AI Team Index, four of five survey groups, July to October 2026
Across everyone who took the survey, the rank correlation between team examples and regular use is 0.56. That is stronger than any of the rules-and-safety statements in the survey, such as knowing what not to enter into AI tools or knowing when output needs human review, which sit between 0.31 and 0.36.
The pattern holds organization by organization. Leave any one organization out and the split stays between 59% and 62% for people with examples, and between 8% and 13% for people without.
What it means for a leader
Only 31% of staff agree their team has practical examples of AI improving its work. It is one of the lowest-scoring statements in the survey. So roughly two in three people are missing the thing that sits closest to everyday use.
In the organizations we work with, the early money and attention usually go to licences, an approved-tools list, a data policy and a training session. Those matter, and people in our data mostly know the rules. But a policy tells people what not to do, and a licence gives them a tool. Neither shows a claims adjuster, a project coordinator or a finance analyst what the tool is for in their own job. An example from someone on their own team does.
A policy tells people what not to do, and a licence gives them a tool. An example from someone on their own team shows them what the tool is for.
This is our reading, not something the survey proves: examples are cheap, and they are the part of the on-ramp most organizations leave to chance. One worked example from a colleague costs an hour to prepare and show.
What this does and doesn't show
Method: The NoW AI Team Index survey, 333 staff in four organizations working with The NoW of Work (five survey groups: one organization ran two), fielded 27 July to 8 October 2026. Group sizes: 104 with team examples and 125 without for regular use; 68 and 76 for better outcomes, a statement in four of the five survey groups. Statements rated 0 to 10; agree = 7 to 10, disagree = 0 to 3. Correlation is Spearman's rank correlation. No organization is named.
The move: put an example in every team
This is the second step in the playbook our data points to (first, state the outcome you want from AI; then put examples in every team; then train on real work).
Ask each team lead for one worked example a month
, from the team's own work: the task, what AI did, what a person checked, and what changed in quality or time.
Have the person who did it show it
in the team's regular meeting. Not IT, not a vendor, not a slide from head office.
Keep the examples in one place, by team
, so a new hire can find three from their own function in five minutes.
Give each team an owner for this
In our programs that is an AI Champion: someone who collects examples and helps the next person try one.
Measure it
Ask both statements again in 90 days and watch whether the share of people with examples moves, and whether regular use moves with it.

