The technology works. It has for a while now. Yet most AI projects in business still go nowhere. An MIT report in 2025 found around 95% of companies’ generative AI pilots were delivering little to no measurable return. BCG puts it another way: under its 10-20-70 rule, 70% of the work in AI transformation is people and processes, and only 10% is the algorithms.
That matches what we’ve seen. A business buys an AI tool, hands it to the team and waits for results. Six months later there’s a subscription nobody uses, a few people producing average content at high volume, and a leadership team wondering what happened.
What happened is they treated AI like a software rollout. It’s a culture change. So we wrote a playbook about it.
AI amplifies what’s already there
The most important idea in the playbook is simple. AI doesn’t replace judgment. It amplifies it.
Give AI to someone with good judgment and you get a force multiplier. They know when the output is good and when it’s rubbish, and they get faster and better. Give it to someone without that judgment and you get the opposite: more work, faster, and worse. The playbook borrows a name for these people from writer Dan Hockenmaier: slop cannons.
His simple two-by-two maps people on two questions. Do they use AI? Do they have good judgment?
- Turbo Brains have good judgment and use AI. These are your stars.
- Steady Hands have good judgment but haven’t picked up AI yet. This is where your effort should go.
- Slop Cannons use AI enthusiastically but can’t evaluate what comes out.
- Dead Weight have poor judgment with or without AI. Don’t start here.
The question isn’t who should get AI access. It’s who has the judgment to use it well.
We learned this the expensive way
The playbook includes our own mistakes, because they teach more than the success stories.
One team member with no coding background tried to build a software project with AI coding tools. The AI wrote plenty of code. Nobody could tell whether it was any good. Two days and $2,000 in usage costs later, there was nothing usable.
Another spent hours trying to get an AI to click one button in a desktop app. Someone with the technical context spotted the problem in minutes: the click was off by a few pixels. Same tool, same problem, completely different result.
The lesson in both cases is the same. AI rewards expertise. It doesn’t replace it.
The champion model
So what do the few successful projects do differently? The playbook lays out a five-step approach we use across R6 Digital, R6 Automate and now with R6AI clients:
- Map your people. Work out who has the judgment before you buy a single tool.
- Pick two or three champions. Not ten. Give them access, time and a real business problem with a dollar figure attached.
- Let them build. Prompts, workflows and guardrails, tested and measured. “We saved 12 hours a week” counts. “The team likes it” doesn’t.
- Roll out through champions, not mandates. People trust their colleagues more than a company memo. Show, don’t tell.
- Document everything and keep iterating. Update what works, remove what doesn’t, and add guardrails as you learn.
The 90/10 rule
If you take one thing from the playbook, make it this. Around 10% of the value in AI comes from building the thing. The other 90% comes from rolling it out: the training, the documentation, the permission to experiment, and the iteration.
Most businesses get this backwards. They spend months choosing tools, then launch with a Slack message and wonder why nobody uses it.
Download the playbook
The AI Culture Playbook is free. It covers the Judgment Matrix, our own war stories and some very public ones, the champion model, the guardrails every rollout needs, and a five-step checklist you can use this week.
Download The AI Culture Playbook (PDF)
Every recommendation in it was tested on our own businesses first. If AI isn’t the answer for yours, we’ll tell you.
