There's a concept in algorithmic advertising called the "conversion signal floor." Below 50 conversion events per week per ad set, Meta's algorithm can't optimise effectively. CPMs spike 15-40%. Performance craters. Not because the platform is broken — because the algorithm doesn't have enough data to learn from.
The exact same principle applies to AI adoption in your business. And almost nobody is talking about it.
The adoption floor nobody mentions
Every AI tool — from ChatGPT to Copilot to custom-built internal solutions — gets better with use. Not in a vague, hand-wavy way. In a measurable, mechanical way. The more your team uses the tool, the better it learns their patterns, their language, their workflows. The better it gets at delivering useful output.
Below a certain usage threshold, the tool can't learn enough to be useful. It gives generic answers. It misses context. It feels like a novelty rather than a productivity multiplier.
This is where most businesses are stuck. They've bought the licences. They've announced the initiative. And they've got 20-30% of the team using the tools sporadically, getting mediocre results, and confirming their suspicion that "AI isn't ready yet."
AI is ready. Your adoption rate isn't high enough for it to prove it.
What this looks like in practice
Imagine a team of 20 project managers. You've rolled out an AI meeting summarisation tool. 5 of them use it regularly. 8 tried it once and went back to manual notes. 7 never logged in.
The 5 who use it regularly? They're getting increasingly good summaries because the tool has learned their meeting patterns, their terminology, their action item format. It's saving them 3-4 hours a week.
The 8 who tried it once? They got a generic summary that missed context and included irrelevant details. They concluded the tool doesn't work.
Same tool. Same technology. Completely different experience. The only variable is sustained usage.
The compounding effect of adoption
This is where it gets interesting — and where the profit impact becomes real.
When adoption crosses the threshold, something compounds. The tool gets better, which makes people use it more, which makes it get better still. The team starts sharing tips. They discover use cases nobody planned for. They start solving problems you didn't even know you had.
Below the threshold, the opposite compounds. Poor results lead to lower usage, which leads to worse results, which confirms the belief that AI doesn't work. The tool becomes shelfware. The budget is wasted.
The difference between these two outcomes isn't the technology. It's whether someone did the change management work to push adoption above the floor.
How to get above the floor
Three things move the needle, and none of them involve buying better technology.
First, pick one team and one workflow. Don't try to get 500 people using AI. Get 15 people in one department using one tool for one specific task, every day, for 30 days. That's enough to get above the signal floor for that team.
Second, make it the default, not an option. If meeting summaries are still being done manually alongside the AI tool, people will default to manual. Remove the fallback. Make AI the way this workflow happens.
Third, show the numbers after 30 days. When the team can see that they've saved 60 hours in a month — specific, measurable hours — the conversation changes. It stops being about whether AI works and starts being about where to apply it next.
The bottom line
If you've invested in AI tools and you're not seeing results, the diagnosis is almost certainly adoption, not technology. Your tools aren't broken. Your team isn't using them enough for them to work.
That's not a criticism. It's the most common pattern in AI adoption. And it's fixable — with the right approach to change management, the right focus on specific workflows, and the patience to push past the signal floor.
The businesses that get above that line see compounding returns. The ones that don't see compounding disappointment. Same tools. Same technology. Different adoption. Different profit.
Related case study: How an accountancy firm increased operating profit by 35% — what happens when AI usage crosses the adoption threshold.