Skip to content 
post examines the skills required to work effectively with AI, mainly focusing on consumers of AI systems. In the text below, I will dissect the AI skills for the Business Competency Framework developed by The Alan Turing Institute, demonstrate how the framework’s foundation is rooted in timeless skills, and recommend areas for upskilling among non-technical individuals.
My impression is that we entered the global pandemic of hearsay by spreading headlines and 1000-character-long-AI-generated summaries (or as much as LinkedIn permits) on topics that concern us all.
Opinions pile on top of opinions about the future of the workspace and topics such as education, security, or even human extinction in the AI era. Supported, unfortunately, often, by the most recent non-peer-reviewed research, which was superficially red and understood. In some cases, understanding is not even the goal one wants to optimise its function for. The goal is to earn hundreds or thousands of likes and get dozens of new followers.
Panem et circenses are available with every new feed refresh, fresh (mis-) information served, so we don’t need to engage our grey matter in finding the “truth.” Whatever this means today, when basic research efforts are getting outsourced to AI, and the good enough truth is slowly creeping toward becoming a new standard.
Nonetheless, the market demands that we get a proper set of…
AI Skills
For most of us working closely with AI developments, when we step out of our IT circle, we realise people don’t talk or care as much about generative AI as we (would like them to) do.
But, one thing they do care about is the correctness of the outputs produced by AI: is it good or not? Or to reframe it in my sister’s, aka math teacher, words: “What should I use it for? It gives me wrong results from the prompted math equations.”
And yet, a few days ago, it was reported that Gemini with Deep Think achieved a gold-medal standard at the International Mathematical Olympiad.
So, where’s the gap here, or more precisely…
Let’s begin with the concepts that everyone is trying to re-package now, and that is — a skillset framework mixed with some version of the responsibility assignment matrix.
Although these frameworks are questionable classifiers, as they tend to “box” the people and their abilities without a proper assessment, they provide a useful starting point for orientation.
That said, I’ll use an example of an AI skills for Business Competency (Meta-) Framework developed by The Alan Turing Institute, which outlines four skill levels targeting four main learner personas across five dimensions representing a set of competencies, behaviours, and responsibilities👇🏼.
Diverging slightly from the post topic, I need to note my top-of-mind, evident shortcomings in the framework’s mapping of skill levels to personas, such as:
- It is disconnected from the market’s need for M-shaped professionals from the “AI Worker” persona, where the designation of a “Working” level for dimensions like “Privacy & Stewardship” or “Evaluation & Reflection” falls short of real-world requirements. This is especially true in regulated industries, where every employee handling sensitive data is expected to have strong knowledge of GDPR and compliance frameworks — a mandate that will probably extend to understanding AI risks and biases.
- Or, how framing the “AI Leader” as an “Expert” in the “Problem Definition & Communication” dimension is misleading, as it suggests they should possess deep technical expertise. However, this is often not the case; many leaders depend on their AI-savvy teams to bridge the gap with hands-on technical insight when making decisions.
And, there’s more to it, but let’s focus on the AI competencies. To do so, I will share one more table to complement the necessary understanding of the learner personas:
Now, we’ll assume how we all managed to find our “spot under the Sun” and map ourselves to one of the above-presented personas. The next question that comes up is…
Which skills are timeless, and where are the gaps in the current skills vs. AI skills?
The proof to the first question is (somehow) straightforward: if we analyse Image #2 without a focus on the term “AI”, it becomes clear how the listed AI competencies are the application of existing, timeless ones, such as:
- Critical thinking,
- Risk management,
- Ethical judgement,
- Strategic planning,
- Communication and collaboration,
- Continuous learning,
- Digital literacy,…
However, the novelty comes from applying them to AI. The context of AI introduces different challenges, which require these skills to be adapted and deepened. For example:
- “Risk management” is not new, but addressing the risks of biased language models or autonomous decision-making presents a new set of challenges to mitigate.
- “Ethical judgement” is not new either, but applying it to identify model (mis-)use, or job displacement due to automation, presents entirely new dilemmas.
Therefore, the gaps lie in the foundational, domain-specific nuances that allow a collective to effectively leverage AI as a tool rather than be “used” by it.
With this in mind, there are already learning paths being offered to acquire the AI “nuanced” skills, and these can help you kick off your learning journey.
My recommendations for both non-tech and tech people who don’t primarily develop AI solutions would be:
- Master high-level understanding of different language models (e.g., LLMs vs. SLMs vs. other specialised models, “thinking” vs. “non-thinking models”, etc.), how to prompt them and when to use them (what are the pros and cons of using AI). Get an understanding of what AI agents are and where we stand on the AGI path, so you get a feeling of what kind of tools you are dealing with.
- Understand “failure modes” and learn how to evaluate outputs. Learn the ways models can lie and manipulate, such as bias, hallucinations, or data poisoning, so you avoid resolving problems AI created in seconds. For this, you’ll need to develop an evaluation checklist (from input to output) for specific (types of ) problems and ensure that outputs are critically reviewed and tested before they reach the masses.
- Create, don’t just consume AI products. While soft skills are a great asset, building practical hard skills is just as important. I believe everyone should start mastering the AI features accessible in the tools we use daily, e.g., AI tools in Excel. From there, I would recommend you start learning no-code and low-code solutions (e.g., Copilot Studio or AI Foundry) to develop custom AI agents with a simple “clicky-clicky” method. Mastering these workflows will boost your performance and AI domain knowledge, making you more competitive in the future job market.
To end this post, one takeaway I hope you’ll get is that we all need to put in the mental effort to enrich our current skills with AI ones.
Because AI effectiveness depends on how thoughtfully we interact with it, and that requires the same critical thinking, risk assessment, and ethical judgment we’ve always needed, just applied to new challenges. Without these foundational skills to evaluate outputs and avoid over-reliance, we risk being led by AI (or by people who know how to use it) instead of using it to our advantage.
Thank you for reading!
If you found this post valuable, feel free to share it with your network. 👏
Stay connected for more stories on Medium ✍️ and LinkedIn 🖇️.
This post was originally published on Medium in the AI Advances publication.
Source link
#Skills #AISkills #Data #Science
