The Executive Study Group (ESG) webinar named “AI Prompt Tips – How to Release
the Hidden Power of GenAI to Boost Your Productivity” was held by Asia Pacific
Institute for Strategy Limited (APIFS) on 7th Nov 2025. The webinar
topics included “Core of AI Interaction”, “The Need for Clarity”, “Introducing
CROFTC”, “How AI “Think””, “Layered Prompting” and “Semantic Precision”.
In the beginning, Dr. Mark Lee explained “Prompt” which were more than
instructions and with multiple forms.
Dr. Lee used the relationship between a director and an actor as a
metaphor for the user and GenAI. You need to provide detailed prompts to teach
GenAI to output the content you want.
Then he mentioned the need for clarity that had golden rule of prompt
engineering. They were “Different Thinking Logic than Human” and “By Patter
Recognition”.
And then Dr. Lee briefed how to achieve clarity. That used specific
scope and detailed description, as well as, break down complex tasks in
sequence.
The Poll 1 surveyed participants’ current prompting habit. Most of us
are in item 2 (39.4%), item 3 (20.4%) and item 4 (23%).
After that Dr. Mark Lee introduced effective pattern for prompting named
“CROFTC” that stands for Context, Role, Objective, Form, Tone and Conditions.
Where Context for setting the background, Role for defining the Persona,
Objective for stating the goal, Form for structuring the output, Tone for
setting the emotional style and Conditions for defining the boundaries.
He also demonstrated the different versions using CROFTC but one focused
on product and the other focused on brand.
And then he asked how AI “Thinks”. Actually, AI doesn’t think but used
sophisticate algorithm and big data.
There are four steps of AI processing from prompt to output. They are “Tokenization”,
“Embedding”, “Neural Network Processing” and “Output Generation”.
So that you should be given clearer and better structures prompt to get
more accurate and relevant output.
The Poll 2 was asked the greatest barrier preventing to use GenAI. Most of
participants selected Training Gap (37.7%).
About information retrieved, Dr. Lee said prompts aided retrieval but
GenAI still had hallucination. One method could overcome the AI knowledge
cutoff was Retrieval-Augmented Generation (RAG).
Finally, Dr. Lee mentioned layered prompting that the concept of the
chain-of-thought was introduced.
He explained a layered approach which enhanced contextual depth with sequential
layering.
Lastly, he discussed the semantic precision that should select words
carefully. Don’t use generic commands but more specific words.
At the end, the Poll 3 was AI prompting implementation and surveyed us
based on today’s contents.
Most of participants selected CROFTC as a Habit (55.1%).
沒有留言:
發佈留言