At a recent roundtable, one participant asked me, “How do you develop creative briefs today?” She said she had a Google Doc with a nine-page prompt for developing the brief and pasted the whole thing in at the start of every session. Then she mentioned, almost in passing, that she couldn’t get through a full article without running out of tokens.
Then she said she was the only one using this document. Ugh.
Nine pages describing the brand, the audience, the formats, the constraints, and her editor’s pet peeves. She found the time to build this, and it worked well enough that she reached for it every single day. And yet…nobody else at that company had it.
A few minutes later, someone else said the same thing had happened in three separate business lines with different marketing managers. There was no common language, no common skill, and as he put it, no “common brain.” Like many organizations, there was no shortage of people figuring things out. But what they figured out had no way to be shared.
The power users hold the keys to progress…and aren’t sharing
When I look at usage data inside client organizations, I see the heaviest AI users aren’t necessarily the technical staff. At one financial infrastructure company, the top 10 token users were almost entirely non-technical. Marketers, operations folks, customer support. The people closest to the messy, judgment-heavy work were inventing the most. Does that surprise you?
That means the most valuable AI knowledge in your organization is being generated at the widest, thinnest part of the org chart, by people with no mechanism for telling anyone about it. Your marketing manager is not going to file a ticket about her nine-page prompt. Why would she? To her, it is not an innovation. It’s just how she gets through her work.
AI enablement is built as a one-way street
Most AI enablement programs are designed to push knowledge outward. A central team chooses the tools, develops the training, runs the sessions, and publishes the playbook. Knowledge flows from the center of the organization to everyone else — but rarely in the other direction. When was the last time your AI program brought what employees were learning back to the people running it?
To be fair, this approach solves a real problem: most people don’t have time to teach themselves, so giving them structured training is a worthwhile investment.
The world’s most powerful SEO platform, purpose-built for Enterprise.
But those central teams are not where the discoveries happen. If knowledge only travels outward from the center, then everything your front-line people learn stays with them, and you are paying for the same lesson over and over.
AI literacy must be a bi-directional flywheel rather than a program. Its guidewords: Spark, spread, scale, sustain.
How to get the flywheel going both ways
Doing this requires three things, and none of them are software.
1: Somebody whose job includes noticing. I call this the AI Lead, and noticing is the part that often gets left out of the job description (if there even is one). When your marketing manager mentions her Google Doc in passing, someone needs to recognize the signal, capture it and share it. Who on your team is doing that right now? And what happens to the information once they have it?
2: Somewhere for the noticed signal to land. A discovery that gets noticed but stays buried in a Slack thread isn’t useful to anyone else. Someone needs to take it from there, decide whether it should become standard practice, and turn it into something others can use. That requires a named function with the authority to take something like a nine-page prompt, refine it, and put it into a format that can be spread.
3: A horizontal path for patterns to spread. If your consumer marketing group solves a problem in March and your B2B group solves the same problem again in July, the problem isn’t either team. It’s that there was no path for what one group learned to reach the other.
What the AI flywheel gets you
Winning marketing orgs learn faster than the world around them changes, and that requires channels for learning to spread. You need the wheel turning in both directions, so that when one person figures something out, everybody gets it.
I walked through how the full flywheel turns in a short video, if you would like to see the mechanics.
In the meantime, there is a version of that nine-page Google Doc somewhere in your organization. Someone built it, it works, and you don’t know about it. How will you find it?
The post Turn scattered AI wins into an organizational advantage appeared first on MarTech.