When Anthropic asked people what they hoped for with AI(opens in new tab), professional excellence topped the list. People get cautious when an AI rollout focuses on what they don't know yet, and they experiment and explore when it builds on what they already do well and applies those strengths to new tools. Whether a rollout can do that comes down to three levers: the mindsets people hold, the skills they build, and the access they have. If you lead that work in your organization, use this framework to set your strategy; the last section gives you three ways to start the conversation with one team.
How AI fluency is changing
Early work on AI fluency catalogued behaviors you can observe in an AI conversation(opens in new tab), like clarifying the goal or identifying the audience. While those behaviors remain decent coaching pointers for someone getting started, they're no longer an effective way of measuring AI fluency at scale. Over the last few months, model intelligence has improved so much that AI can now handle some of those behaviors without the user. Organizations seeking transformation should consider three key levers on their journey to AI fluency: mindsets, skills, and access.
The AI fluency ecosystem
Mindsets are what people believe about AI: what it can and can't do, and its place in their work.
Skills are defined as someone's ability to complete complex tasks with AI. You see this as task success: how often the task comes out right. As a person learns to manage context, use skills for repeatable tasks, and automate the right things, their task success goes up.
Access is what people can use, from the models themselves to the features that are switched on in the organization.
The AI fluency ecosystem isn't a rubric and there isn't a finish line or a “good rating.” Mindsets, skills, and access are interconnected parts of a system.
Mindsets shape people's appetite for experimentation
Mindsets shape whether people experiment. Understanding your organization's mindsets tells you where to focus your efforts. Seven mindsets to look for:
Given how fast AI is changing, consider this a working list. People often hold several of these at once and move back and forth between them over time. If you only focus on one, start with the third: today's AI is the worst you'll ever use. When people truly believe that, they keep coming back to retest old assumptions and try new things. Where that belief hasn't taken hold, people's ideas of what AI is capable of get frozen in time and innovation stalls.
Skills help people accomplish hard tasks with AI
As people use more tools and strategies, their task success rises. The chart follows one demanding task over fifty attempts: success climbs as the team refines the prompt language and adds a skill so output formatting runs reliably, then again as it adds a rubric so the AI checks its own work and adds adversarial checks for extra validation.
There is no fixed list of skills. The goal is fixed: the task succeeds.
Access is a compounding stack
Access is the lever an organization controls most directly. Models come first: which ones people can reach, and how capable they are. Harnesses are the scaffolding around a model, such as agents and tool permissions. Surfaces are where AI shows up, from a chat window to the tools people already work in. Features are what's switched on, such as memory, projects or connections to your data. Each layer complements the rest, and mindsets and skills can't compensate for missing access.
Reflect on your organization
Take a few private minutes with one team in mind. Then use what you notice to pick one of the conversation starters below.
Where to go next
Here are a few ways you can start the conversation around AI fluency:
- Mindsets: show the mindsets at a team meeting and discuss which ones are most important to your work
- Skills: ask people for the hardest task they can complete with AI, and how
- Access: check what information your AI tools can reach