This informal CPD article, ‘AI fluency: more than knowing how to prompt’, was provided by Heather Baker, founder of The AI Edit, a B2B AI training, coaching, and consulting business.
Most professionals are using AI in their work. Few are genuinely fluent in it.
Being fluent is not the same as being able to type a prompt into an AI tool. Fluency is the judgment that sits around the tools, not just the ability to use them. It is the difference between AI improving professional work and AI quietly undermining it.
Adoption of generative AI in professional and workplace settings has grown rapidly since 2022, yet engagement is uneven. Most users concentrate on basic prompting and content generation, while practices related to risk, governance, and critical evaluation remain underdeveloped (1).
Genuine AI fluency develops across six related areas of knowledge and practice.
1. Learn the concepts
The first area is how the technology actually works. Not at an engineering level. At a decision-making level.
Generative AI predicts. It does not know. Each output is a probabilistic guess at the next plausible word, based on statistical patterns the model has absorbed from a large body of training data. There is no underlying comprehension and no inherent reliability (2).
That mechanic is only the start. Genuine fluency also requires an awareness of where this technology fails and how it carries risk. Hallucinations, where models generate plausible but false outputs, are an inherent feature of generative systems rather than a defect that will be patched out (2). Training cut-offs limit what a model has seen and therefore what it can reliably discuss. Context windows limit how much information can be considered in a single conversation. Agents, which take actions in the world rather than only generating text, change the risk profile significantly. Prompt injection, where malicious instructions hidden in external documents or webpages manipulate AI behaviour, is now recognised as one of the leading security risks for any AI system that processes untrusted input (3).
Without these foundations, evaluating AI output critically becomes very difficult. The likely outcome is either over-trust or under-trust. Neither qualifies as professional fluency.
2. Learn the tools
The second area is genuine competence with the tools themselves. Fluent professionals do not simply know that AI exists. They use it regularly on real work, until they know what each tool can do, where it falls down, and how to get useful output from it.
The current landscape includes several major general-purpose AI chat tools and a smaller number of AI-powered search tools. Each has its own strengths, its own privacy posture, and its own appropriate use cases. Choosing between them is a professional judgment, not a matter of preference.
Beyond standalone tools, AI features are now embedded in mainstream office productivity software, collaboration platforms, and customer relationship management systems. Many organisations are paying for AI capabilities that employees have not yet activated (4).
Competence develops through regular use on real work, not from watching tutorials.
3. Learn how AI is changing the profession
The third area is how AI is changing the specific profession a person works in. Adoption is uneven across sectors. Legal services are moving quickly on document review, contract analysis, and discovery. Medicine is wrestling with diagnostic models and patient data governance. Education is rethinking assessment and academic integrity. Financial services is reshaping advice, fraud detection, and compliance workflows (1).
Deep professional knowledge remains a fundamental asset in this context. Years of practice in a field cannot be replicated by any generative model. The combination of domain expertise with AI fluency consistently produces stronger outcomes than either capability on its own.
This area also requires knowing where to look for signals about what is changing. Peer practice. Client expectations. Regulatory direction.
The European Union's AI Act
4. Learn the macro picture
The fourth area is the broader landscape: who is building these systems, what they are incentivised to do, and how they are being regulated.
Major AI developers operate as commercial enterprises, not neutral parties. They have investors to satisfy, strategic interests to protect, and particular product visions they are advancing. Free training, free tools, and free guidance from any commercial AI provider will reflect those interests, intentionally or otherwise.
Regulatory frameworks have advanced more quickly than many practitioners expected. The European Union's AI Act establishes obligations for high-risk AI systems and prohibits certain uses outright (5). The United Kingdom government has set out a principles-based approach to AI regulation through guidance to existing sector regulators (6). The United States is developing AI legislation at both federal and state level. Liability for AI-related decisions in professional work typically rests with the practitioner rather than the technology vendor.
5. Learn the ethics
The fifth area is professional ethics, treated as operational practice rather than as a lecture.
When AI generates an error in professional work, accountability remains with the practitioner. The algorithm cannot be held responsible. The professional retains ownership of the outcome.
Every professional setting requires clear, written red lines drawn before AI is deployed. Which tasks are off-limits for AI assistance entirely? What categories of client or patient data must never enter a public AI model? Which professional judgments must remain wholly human regardless of model capability?
Bias in AI outputs (7), transparency about AI involvement in client work, environmental impact of training and inference, and the security of confidential data are not abstract concerns. They are operational risks. Like all operational risks, they require governance.
6. Build the right mindset
The sixth area is the one most easily overlooked, and arguably the most consequential.
AI systems trained on human feedback tend to agree with their users. They reflect a user's framing back, sometimes dressed as analysis. This behaviour has a name: sycophancy. It is a recognised property of large language models trained through reinforcement learning from human feedback (8). Without accounting for it, users mistake validation for insight, and reinforcement for thinking.
Overconfidence is also widespread among new AI users. A short period of use is often mistaken for capability. A small amount of knowledge is genuinely dangerous in a field that is changing this quickly.
Fluent professionals resist the pull. They retain curiosity while holding scepticism close. They slow down in moments that reward speed. They treat AI capability as a habit to maintain, not a course to complete.
What this means for CPD
AI fluency is no longer optional for any profession that involves judgment. It belongs alongside core professional competence, not as an extra. For CPD purposes, that means treating AI fluency as a continuous area of development. Not one course, completed and set aside. A standing thread within professional growth.
A short self-assessment can help locate the right next step. For each of the six areas, a score from one to five offers a rough benchmark.
- Concepts: how well do you grasp how AI works, including hallucinations, agents, and prompt injection?
- Tools: how competently are you using the major AI tools in day-to-day work?
- Industry: how clearly can you describe the way AI is changing the specific profession?
- Macro: how well do you grasp the regulatory and commercial picture?
- Ethics: how clearly have you defined professional red lines for AI use?
- Mindset: how consistently are you treating AI capability as a continuous learning habit?
A score below three in any area indicates where the next piece of CPD belongs.
We hope this article was helpful. For more information from The AI Edit, please visit their CPD Member Directory page. Alternatively, you can go to the CPD Industry Hubs for more articles, courses and events relevant to your Continuing Professional Development requirements.
References:
(1) Stanford Institute for Human-Centered Artificial Intelligence. Artificial Intelligence Index Report 2025. Stanford University, 2025. https://aiindex.stanford.edu/report/
(2) Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y., Dai, W., Madotto, A., Fung, P. Survey of Hallucination in Natural Language Generation. ACM Computing Surveys, Vol 55, No 12, 2023. https://doi.org/10.1145/3571730
(3) OWASP Foundation. OWASP Top 10 for Large Language Model Applications. 2025. https://owasp.org/www-project-top-10-for-large-language-model-applications/
(4) McKinsey and Company. The State of AI: How Organizations Are Rewiring to Capture Value. McKinsey Global Survey on AI, 2025. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
(5) European Parliament and Council of the European Union. Regulation (EU) 2024/1689 of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union, 2024. https://eur-lex.europa.eu/eli/reg/2024/1689/oj
(6) UK Department for Science, Innovation and Technology. A Pro-innovation Approach to AI Regulation. UK Government White Paper, 2023. https://www.gov.uk/government/publications/ai-regulation-a-pro-innovation-approach
(7) Schwartz, R., Vassilev, A., Greene, K., Perine, L., Burt, A., Hall, P. Towards a Standard for Identifying and Managing Bias in Artificial Intelligence. NIST Special Publication 1270, National Institute of Standards and Technology, 2022. https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.1270.pdf
(8) Sharma, M., Tong, M., Korbak, T., Duvenaud, D., Askell, A., Bowman, S., Cheng, N., Durmus, E., Hatfield-Dodds, Z., Johnston, S., Kravec, S., Maxwell, T., McCandlish, S., Ndousse, K., Rausch, O., Schiefer, N., Yan, D., Zhang, M., Perez, E. Towards Understanding Sycophancy in Language Models. 2023. https://arxiv.org/abs/2310.13548
