The Big AI Secret · ongoing research

Everyone can use AI. Who gets to shape the work?

The excerpts from agency interviews distinguish permission to use AI from participation in shaping how it works. They describe ways to bring ideas into shared practice, but do not establish whether those routes spread influence across the team.

6 November 2026 · 4 minute read

An open invitation does not necessarily produce ideas from across a team. In an agency with under 25 staff, a team lead said they encourage suggestions. Yet new AI ideas mainly come through them, while colleagues tend to follow what is already provided. The revealing detail is not a restriction. It's the gap between being allowed to contribute, and actually doing so.

“It's mainly to do with me because my team works for me and I'm actually always encouraging them to bring up new ideas, but they usually like to follow what's given to them.”Team lead, agency, under 25 staff

That account leaves an important question unresolved. Do colleagues prefer an established method, lack the confidence to challenge it, or simply have other priorities? The excerpt can't tell us. It does show why permission alone is a weak description of how an organisation develops its AI practice. An invitation can be genuine while initiative remains concentrated.

Use is not the same as influence

Another small agency account made a related distinction in unusually few words. Asked about use across the business, the speaker described broad participation alongside their own deeper involvement.

“All, but I use it deeply.”Agency, under 25 staff

Depth of use is not proof of authority over everyone else's work. Nor does uneven depth necessarily signal a problem. Some roles may need more experimentation than others. But this distinction matters when a director hears that everyone uses AI. That answer says little about who chooses the tasks, tests different approaches or decides which methods colleagues adopt.

A test needs somewhere to go

A separate small agency describes a route from an individual's experiment to a possible change in shared working methods. The person tests an idea, brings back the results and allows the team to consider whether it belongs in their work.

“But then, yes, usually that person would go off, test it, and then share the results and see whether it was successful. And if it is, then, you know, we potentially all build that into our workflows.”Agency, under 25 staff

The useful detail is the return to the group. An experiment is not left entirely with its author. Results become available for others to examine before adoption. The wording also preserves uncertainty: a successful test may enter shared practice, rather than automatically becoming the new method.

This is a reported process, not evidence that contributions are evenly spread. We do not know whose ideas are tested, what counts as success or how often colleagues adopt them. Still, it identifies something more concrete than an open invitation: a route through which a staff suggestion can affect how other people work.

Making uneven development visible

Elsewhere, a small agency described collective development sessions in which ideas are listed and developed into standardised work. Its stated aim is not simply to encourage more experiments. It is to prevent parts of the business from moving ahead while others fall behind.

“So they're not silos of the business that are really advanced with AI and others that are atrophying. We make sure that there's balance across what we're doing.”Agency, under 25 staff

That is a claim about intent and practice, not a demonstrated result. The excerpt offers no measure of balance. Its value is that uneven development is being treated as an organisational concern, rather than something individual enthusiasm will necessarily correct. Shared sessions create a place to notice which areas receive attention and which do not.

An agency with 100+ staff describes a different arrangement. People can experiment, while specialist support helps refine their work. The accompanying account says an AI Capabilities Lead helps prioritise workflows that benefit more people.

“But what it does is it means that everyone's got autonomy to try things, but we also have a roll-up positioning where we can then get the real experts to really help refine it.”Agency, 100+ staff

Look at whose ideas become shared practice

These accounts don't establish a preferred structure. They describe ways to connect individual initiative with wider adoption: sharing test results, developing ideas collectively and providing specialist help. None proves that quieter staff contribute, or that the most useful ideas win attention.

For a director, the useful next enquiry is therefore about actual proposals. Where did they originate? Who tested them? What changed for colleagues afterwards? Following an idea through those decisions would reveal more about participation than asking whether staff are free to experiment. The distinction is between a team that can use what it is given and a team that helps decide what gets built.

Evidence note: this reading draws on the supplied anonymised agency call excerpts and their accompanying answer summaries for The Big AI Secret. They are qualitative accounts, not a representative estimate of agency practice. Speaker roles are marked where they were not recorded; no job titles have been assumed.