Nearly half of desk workers would feel uncomfortable telling their manager they'd used AI for everyday tasks like writing a message or brainstorming ideas. That figure comes from Slack's Workforce Lab, and it explains a lot about how AI adoption in agencies tends to go. When the tools arrive, one or two people usually start using them where everyone can see. Others try them in private, and some barely touch them.
AI is one more new way of working, and when an agency team starts using it for real, the same few patterns tend to show up in roughly the same order. The research backs most of them up.
Who moves first
The first people to move are usually at the top. Gallup's figures for the second quarter of 2026 show 51% of leaders using AI frequently, against 26% of individual contributors, and Gallup says the gap between the two has widened over time. In an agency that often means the founder or MD, plus one or two people who were using ChatGPT on their own account before anyone bought a licence.
That second group is easy to miss. The Microsoft and LinkedIn Work Trend Index found 78% of people who use AI at work bring their own tools, rising to 80% in small and medium-sized companies. So by the time you roll anything out, someone on the team already has their own habits in their own account.
Early movers are great to have, but what they learn tends to stay in their own chat history. Their best prompts and the brief template they rebuilt don't reach anyone else unless somebody makes that happen. And if you're one of the early movers yourself, honestly, it's easy to assume the team is where you are, when the Gallup numbers say they probably aren't.
Where AI adoption in agencies tends to stall
After the first burst, use flattens out. Slack saw AI use among desk workers worldwide climb from 20% to 32% by March 2024. In its next survey, US use moved by a single point, from 32% to 33%, and global excitement about AI dropped from 47% to 41%. BCG's 2025 AI at Work survey found the same shape: regular use among frontline employees had stalled at 51%, while more than three-quarters of leaders and managers use it several times a week.
The stall tends to line up with training time. BCG found regular use is higher among people who get at least five hours of training and have access to in-person coaching. Slack found 61% of desk workers had spent less than five hours learning how to use AI, and 30% had no training at all. So most people haven't had even five hours, which is a low bar to miss.
In an agency, a common version starts with a lunch and learn. People try it on a quiet afternoon, then a pitch lands or a client escalates and everyone goes back to what they know works. Nobody decides to stop. They just lose the reason to keep going, and only 22% of employees in Gallup's 2025 survey said their organisation had communicated a clear plan or strategy for AI.
What the quiet majority does
This is the group owners tend to misread, because a lot of the quiet majority are already using AI and keeping it to themselves. Microsoft and LinkedIn found 52% of people who use AI at work are reluctant to admit using it on their most important tasks, and 53% worry it makes them look replaceable. In Slack's survey, the top reason for keeping quiet was that using AI feels like cheating (47%), followed closely by fear of looking less competent or lazy (46% each). Company policy came last, at 21%.
That matters more in an agency, where the work is the product. A copywriter who drafts a first pass with Claude and then rewrites it might not want the creative director to know, in case it makes the work look cheaper. So the learning stays hidden, and you've no real idea what client material is going into which tool.
There's another reason to care about the quiet majority. In a BCG experiment with 758 consultants, run with Harvard Business School, MIT Sloan, Wharton and the University of Warwick, the people who started with lower scores improved by 43% on a creative product task once they had GPT-4. That was a much bigger jump than the top performers got. Which sounds obvious when you say it out loud, but it's why I'd never judge a rollout on how the early movers are doing.
What makes AI spread through a team
AI spreads through a team when there's a clear process with some tracking behind it, so everyone can see whether it's working. The research points at the same few things.
- Say it's allowed, in writing. Slack found 45% of workers don't have explicit permission to use AI, and people who are comfortable sharing their AI use are 67% more likely to have used it for work. Put it on the agenda of the weekly team meeting so it becomes normal to talk about.
- Give people real training time on their own work. BCG's five hours with some in-person coaching is a sensible floor, and a one-hour demo of someone else's use case won't get anyone there.
- Rebuild one shared workflow, so the wins belong to the team. If the process only lives in someone's head, write it down first.
- Get leaders using it in the open. AI power users in the Work Trend Index were 61% more likely to have heard from their CEO about why AI matters, and BCG found the share of employees who feel positive about AI rises from 15% to 55% with strong leadership support. I've covered why manager support matters so much before, so I won't repeat it here.
- Make room to experiment. Microsoft and LinkedIn call frequent experimenting the number one predictor of becoming a power user, and power users are 66% more likely to redesign how their work gets done.
Where to put the effort
Plan your rollout around the quiet majority. The early movers will keep going whatever you do, while the rest of the team needs permission and proper time on real client work before they'll change their habits, and that's most of your agency.
If you want help bringing the whole team with you, this is how Ideal and Digital works with agency teams.