I am not sure how you are feeling, but I subscribe to fifty-plus quality knowledge sources on AI/ML, Data Management, and diverse digital business transformation sources, and I usually have about ten books on the go by my bedside.
It is easy to feel overwhelmed as our AI/ML world is advancing so rapidly. It is virtually impossible to keep up unless you are sourcing the right knowledge sources and you have access to trusted networks.
Too many people are relying on Generative AI sources like Gemini, OpenAI (ChatGPT), Anthropic (Claude), and Perplexity for solving problems. These sources can certainly get you started, but it would be irresponsible to think you have the “most relevant and quality” knowledge that will demonstrate critical thinking and ensure trustworthy outcomes.
How are these tools being used?
As these tools are in their early evolutionary stages, we cannot dismiss their value, as OECD recently reported on Generative AI on Innovation and Productivity:
“Generative AI has proven particularly effective in automating tasks that are well-defined and have clear objectives, notably including some writing and coding tasks. It can also play a critical role in skill development and business model transformation, where it can serve as a catalyst for personalised learning and organisational efficiency gains, respectively. In creative industries, generative AI aids in producing novel ideas and designs, helping diversify thought processes and speed up the innovation cycle. In R&D, it assists in problem-solving, study design, and product development. In academia and industry alike, it enables faster knowledge recombination and cost reductions, accelerating the pace of discovery. Entrepreneurs are also using the technology to automate routine tasks, optimize resource management, and free up time for strategic decision-making, thereby possibly leveling the playing field in new business ventures. Early-stage start-ups have particularly benefited from generative AI’s ability to expedite processes like refining business ideas or funding.”
However, what are some of the problems with over-reliance on ChatGPT (prompt interfaces)?
There are many problems that leaders must understand to ensure their employees are using Chat (LLM) tools efficiently.
- False Information is Rampant in LLMs – Hallucinations are pervasive in LLMs, and Chat GPT can generate answers that sound plausible but are completely false, because it predicts text rather than verifying facts.
- No Source Citations – Without traceable references, it’s hard to trust the data and verify claims, or validate insights.
- Lack of Real-Time Data – Unless connected to the Web, ChatGPT’s knowledge has a cut-off date; new regulations, leadership changes, and emerging competitors can easily be missed
- AI Bias – If the AI’s Training data contains bias, recommendations may unintentionally reinforce stereotypes or poor practices
- Confidentiality Risk– sensitive customer or company data entered into Chat GPT may be a risk if proper privacy controls are not in place
- Regulatory Compliance Risk – In sectors like government or healthcare, inaccurate advice will cause compliance breaches.
What is also key to understand is that there are deeper knowledge generation requirements, as the shelf-life of knowledge is continually reducing. Depending on the profession, it can range from 5-7 years to literally minutes. Until the LLMs have citation sourcing integrated to reputable content sources, leaders must be vigilant to ensure there is not an over-reliance on these toolkits, as professional development well beyond access to LLMs is crucial to build knowledge robustness and relevancy.
Below is a table summarizing some key professions and their estimated knowledge half-life with some implication notations. What is important to appreciate is the speed of knowledge compression in terms of relevancy. Everyone must be learning constantly and be vigilant about this priority to have a successful career.
Knowledge Shelf-Life
How should you look at ChatGPT toolkits?
These chatbot tools, based on LLMs (large language models), provide tremendous value for ideation and structure, but they are not a replacement for human expertise, live data, or critical thinking.
If used alone, there are a number of risks.
First, accuracy suffers and can lead to poor decisions. Second, context can easily be missed, resulting in irrelevant or risk-prone advice, and third, research citations of trusted sources can be easily lost, increasing the risk factor with data not curated from trusted sources that are acknowledged.
Research is showing that the productivity hype is overrated in using ChatGPT tools, and ROI value outcomes are often disappointing, as reported recently by The UK government’s Microsoft 365 Copilot trial, which found staff satisfied and tasks eased, but there were no clear productivity gains, raising doubts about AI’s workplace impact.
A 2025 Harvard Business Review study also found that while generative AI boosted performance on specific tasks, it caused a decline in workers’ intrinsic motivation and an increase in boredom for subsequent, unaided tasks.
The ideal model is “human-in-the-loop,” leveraging AI collaboration toolkits where these chat prompting tools help accelerate work, but humans validate, contextualize, and own the decisions.
Humans are responsible for the integrity and quality of their work products, and demonstrating leadership accountability for the quality of their work.
So despite all the excitement of having new toolkits to use to solve problems, the tacit knowledge is often the richest that humans have in this world, and it is often not recorded anywhere – so the dialogue and communication with relevant people will always enrich perspectives. Tacit knowledge is the deeply personal, experience-based knowledge that people carry in their minds but is difficult to articulate, document, or transfer to others. It’s what someone “just knows” because of years of practice, intuition, and hands-on experience — even if they can’t fully explain how they know it.
As Michael Polanyi said so eloquently, “We can know more than we can tell.” (1966).
In closing, his words perfectly capture the essence of tacit knowledge — that human expertise often resides beyond what can be fully articulated or codified, which is exactly where tools like ChatGPT face major challenges.
This is one of the most strategic conversations that we must start having as leaders, and not to get too lost in the AI hype, although it is extremely immersive, rich, and fascinating – as long as we stay grounded with what constitutes quality knowledge, supported by trusted facts.
In my mind, it’s all about substance and appreciating what a human already has behind his or her ears that is most precious to keep cultivating with new knowledge from trusted sources and using these innovative AI toolkits, to spark ideation, versus being seen as the breakthrough productivity booster.
With the high risk of retirements and extreme tacit knowledge loss, as some experts say that it is over 80% of knowledge that is tacit — and this is NOT in our LLMs.
Internalize this point deeper and recognize that Humans Really Matter!
Learning Sources:
- Tacit Knowledge – The Most Important AI/ML Knowledge Sources for Equations
- Tracking LLM Leaderboard Positions – Shifts Daily.
Definition of Term: GenAI
Generative AI refers to AI models or systems specifically designed to produce content, such as text, program code, images, videos, or sounds, in response to human language queries or prompts. This type of AI differs from other non-generative AI applications that focus primarily, e.g., on analysis or classification tasks. A well-known example of generative AI is ChatGPT — a tool that uses large language models (LLMs) to generate content based on given inputs (prompts). LLMs are trained on large amounts of data and use algorithms to predict the most likely next word or phrase in a sequence, allowing them to produce coherent and contextually relevant content. The development of generative AI has seen key milestones, notably the introduction of transformer-based architectures in 2017. This advancement made AI models more efficient and scalable, with models increasingly growing in size and complexity, now having billions of parameters. These breakthroughs have expanded the use of generative AI across industries, enabling its potential as a general-purpose technology (OECD Report, 2025)
Research Sources:
Boden, Margaret A., Ed. (1990). The Philosophy of Artificial Intelligence. Oxford: Oxford University Press.
Fodor, Jerry. (1981). “The Appeal to Tacit Knowledge in Psychological Explanation.” Representations. Cambridge, MA: Bradford/MIT Press, 1981.
OECD, The impacts of Generative AI On Productivity, (June 2025).
Polanyi, Michael. 2002 (1958). Personal Knowledge: Towards a Post-Critical Philosophy. London: Routledge.
Ryle, Gilbert. (1949). The Concept of Mind. Oxford: Oxford University Press.
Ticong, Liz. (Sept 8, 2025). Microsoft Co-Pilot Study – UK Government Finds no Evidence of Productivity Gains. Tech Republic.
Turing, Alan M. (1950). “Computing Machinery and Intelligence.” Mind 59.2236: 433-60. Reprinted in Boden 1990: 40-66.
