After 50,000 Prompts...
The 12 Levels of AI Fluency That Actually Matter
Most companies and schools haven't seen real business results from AI yet. McKinsey found that 92% of companies plan to increase AI spending over the next three years. But only 1% call themselves "mature" at actually using AI [1].
The MIT research (could) explain why. The researchers tracked 54 students for four months, splitting them into three groups: ChatGPT only, search engines only, and no tools at all. Brain scans showed a clear pattern. Students working without tools had the strongest neural networks. Search engine users fell in the middle. ChatGPT users? Their brains barely worked [2].
But here's the finding that inspired me to write the original research reflection: when students who had never used AI got access to ChatGPT in session four, their brain activity increased. The brain scans showed "network-wide spikes in alpha, beta, theta, and delta bands" - their brains worked harder, not less. Meanwhile, students who started with ChatGPT and switched to working alone showed "under-engagement of alpha and beta networks." Their brains had learned to expect external support [2].
The timing matters enormously. Students who built thinking skills first maintained cognitive ownership while gaining efficiency. Students who relied on AI from the start struggled with reduced capabilities even when the tool was removed. Think of it like muscle atrophy - your brain reallocates resources away from tasks it doesn't regularly perform.
“Prior cognitive investment creates a foundation that enhances AI collaboration. Students who build thinking skills first can use AI as a genuine tool.”
After testing over 50,000 prompts across Chagpt, Gemini, Grok, and Claude, I've learned something important. Most AI fluency frameworks miss the point. They're either too technical or too basic to help people do better work.
The real question: How do you get enough people to Level 5-8 fluency (see below) so AI actually changes how work gets done?
How Fluency Actually Works
Employees who use AI frequently report big benefits: 65% say it makes them faster, 49% say it improves quality, 48% say it makes them more accurate, and 46% say it boosts creativity [5]. But these gains concentrate among a small group who've learned to have conversations with AI systems.
The 12-level framework I am developing, trust me, this is a 2+ year work in progress, shows something you might not expect. The biggest productivity jumps don't happen at the extremes.
In my opinion, ‘Foundation levels (1-4)’ are basic awareness and simple queries, basic digital literacy skills. High Innovation levels (9-12) require technical knowledge, and a developerr mindset, most people don't need.
The sweet spot sits in the middle: Competency levels 5-8.
Foundation (Levels 1-4): Where Most People Start
Level 1: AI Awareness You know AI exists. You've seen people use it. You're aware of basic issues like accuracy and bias.
Level 2: Basic Questions You use AI like Google - asking simple questions and accepting the first answer.
Level 3: Context Setting You've learned to give background information to get better, more relevant responses.
Level 4: Role Assignment You can use "Act as a..." prompts to get specialised perspectives and expertise.
Competency (Levels 5-8): The Target Zone
Level 5: Iterative Conversations You treat AI like a dialogue, refining and improving responses through follow-up questions.
Level 6: Task-Specific Prompting You adjust your approach based on what you're trying to accomplish - different prompts for writing, analysis, brainstorming.
Level 7: Project Integration You use custom instructions, saved conversations, or project-specific setups for consistent results.
Level 8: Workplace Integration AI is seamlessly integrated into your daily workflow and strict testing and development is done. You don't think about "using AI" - you work “with AI”.
Innovation (Levels 9-12): The Advanced Zone
Level 9: Multi-Platform Strategy You know when to use different AI tools for different purposes and combine them strategically. Think: Use Chatgpt o3 to create instructions for a Gamma GPT, Gpt creates incredible Gamma workbooks etc.
Level 10: Quality Evaluation You can quickly assess AI output quality and know when to trust, verify, or discard results, you know when to try new models and different approaches, you may even rewrite entire prompts based on knowledge you might have picked up on a reddit subforum.
Level 11: Knowledge Transfer You can effectively teach others and help teams adopt AI tools productively (For the previous levels).
Level 12: Strategic Vision Anything 12 and above becomes strategetic and vision from practice. For example, I have worked with iCamp for over a year, I have worked with Kaiba, Alayna, Masterymate, Solvably, now with a Studyhall project. I have used AI in many different curriculum approaches, I know where AI works better, and where users might experience problems. You understand AI's broader implications and can make strategic decisions about implementation and adoption.
Level 5 marks when you stop treating AI as a search engine and start using it as a collaborative partner. You begin having conversations, asking follow-up questions instead of accepting first responses. Level 6 means you recognise that brainstorming needs different prompts than analytical work or creative writing. Level 7 means you set up systems: custom instructions, saved conversations, and project-specific setups that work consistently.
Anthropic's researchers identify four core competencies they call the "4Ds": Delegation, Description, Discernment, and Diligence [6]. Delegation means deciding what work to do with AI versus by yourself. Description means communicating effectively with AI systems. Discernment means evaluating AI outputs properly. Diligence covers ethical practice, transparency, and accountability.
A workplace study tracked over 5,000 customer support agents using AI assistance. The tool increased productivity by 15%. The biggest improvements came from less experienced workers and skilled trade workers, who also improved the quality of their work [7]. These productivity gains came from basic conversational competence and task integration, not advanced technical knowledge.
It’s Urgent To Upskill
40% of workforce skills will change within five years [8]. AI affects this differently than previous technologies. Calculators removed arithmetic barriers to mathematical understanding. Word processors eliminated transcription limits on written expression. These tools built cognitive capacity while reducing cognitive load.
AI can lower skill barriers, helping more people acquire proficiency in more fields, in any language and at any time [9]. But there's a catch. The most advanced AI users aren't those who've memorised prompting techniques. They're people who've developed intuitive sense for when and how to engage AI in multiple ways, and projects.
Think of phase transitions in physics. Water stays liquid across temperature ranges, but at specific points, small changes produce dramatic state shifts. AI fluency works similarly. The jump from Level 4 (role assignment) to Level 5 (iterative conversation) is a phase change. Users suddenly discover they can engage AI in real problem-solving instead of simple task completion.
Research shows that 50% of all workers will need retraining within five years [10]. But the conversational nature of AI tools suggests something different. Unlike previous workplace technologies that required specific training and certification, AI tools can adapt to users instead of demanding users adapt to them.
This enables what you might call "synthetic polymathy" - the ability to engage meaningfully across disciplines by using AI as an extended cognitive system. People with Levels 5-8 fluency can work effectively in areas outside their core expertise. They're not becoming fake experts. They're developing a new form of human-machine hybrid intelligence.
McKinsey found that redesigning workflows when implementing AI has the biggest impact on company earnings [11]. Companies that focus only on tool deployment without developing conversational competence among their workforce capture minimal value. Those that build Levels 5-8 fluency while redesigning workflows around human-AI collaboration get compound benefits from both technological capability and human adaptability.
What Next
I read thought leaders every day, I read blogs every day, I read articles every day, and I read what people write, every day. Most organisations are optimising for the wrong outcome. They're trying to create AI experts when they need conversational partners. The future belongs to the 5-8s - people who've moved beyond basic prompting into genuine collaboration, but haven't disappeared into weird technical complexity.
In five years, Level 8 fluency will be as fundamental as email competence was in 2005. The question that should keep leaders awake: "How quickly can we get enough people to Level 5 that our competitors can't catch up?"
The window for building this capability systematically is closing. Not because the technology is advancing too quickly, but because the organisations that achieve critical mass of conversational competence first will reshape their entire industries around human-AI collaboration.
Where are you on this scale? More importantly - where is your organisation?
I help schools and governments work this out, and my programs have incredible adoption and build confidence. Feel free to get in touch to know more.
Phil
phil@pblfuturelabs.com
References
[1] McKinsey & Company (2025). "The state of AI: How organizations are rewiring to capture value." Survey of 1,491 participants across 101 nations showing only 1% of companies describe their AI rollouts as "mature."
[2] MIT Media Lab (2025). "Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task." Brain connectivity analysis showing increased neural activity in students introduced to AI after developing cognitive foundations.
[3] Aura AI Blog (2025). "Workforce Skill Gaps: AI-Powered Strategies for 2025." Executive survey data on skill gaps and planning.
Betterworks (2025). "How AI Fluency Enables Workplace Innovation." 2025 State of Performance Enablement report findings on AI user perceptions of untapped potential.
Betterworks (2025). "How AI Fluency Enables Workplace Innovation." Survey data on AI benefits reported by frequent users.
[6] Anthropic (2025). "AI Fluency: Frameworks and Foundations." Framework for AI Fluency describing the 4Ds competency model.
[7] Microsoft Research (2024). "Generative AI in Real-World Workplaces." Study of 5,000+ customer support agents showing 15% productivity increase with AI assistance.
[8] World Economic Forum (2025). "Future of Jobs Report 2025." Projection of 40% change in required workforce skills within five years.
[9] McKinsey & Company (2025). "AI in the workplace: A report for 2025." Analysis of AI's potential to lower skill barriers and expand access to knowledge.
[10] World Economic Forum (2020). "The Future of Jobs Report 2020." Estimate that 50% of workers will need retraining by 2025.
[11] McKinsey & Company (2025). "The state of AI." Survey finding that workflow redesign has the most significant impact on EBIT when implementing generative AI.