Chapter 2 - AI chatbots and prompt engineering.pdf

Chapter 2
AI Chatbots and
Prompt Engineering
Copyright©2025 Strategic Data Science Limited
Dr. LAM Yat-fai (林日辉)
Chief Data Scientist, Strategic Data Science Limited
Adjunct Professor, City University of Hong Kong
E-mail: quanrisk@gmail.com

Declaration
❑ Copyright©2025 Strategic Data Science Limited.
❑ All rights reserved. No part of this presentation file may be
reproduced, in any form or by any means, without written
permission from Strategic Data Science Limited.
❑ Authored by Dr. LAM Yat-fai (林日辉),
Chief Data Scientist, Strategic Data Science Limited,
Adjunct Professor, City University of Hong Kong,
Doctor of Business Administration,
CFA, CAIA, CAMS, CFE, FRM, PRM, MCSE, MCNE.
2

❑ Artificial intelligence
❑ Conservational prompting
❑ Prompt engineering
❑ Banking applications
❑ Afternoon workshop
Outline
3

❑ Customer experience
▪ Recommend to view
▪ YouTube
❑ Loan application
▪ Approval in 15 minutes
▪ WeLend
❑ A super key
What is artificial intelligence?
4

Turing test
5
Reasonable man Judge Machine

Weak AI vs Strong AI
6

Consumer AI
❑ Free to use
❑ On a personal computer
❑ Over Internet
❑ With a Google account
❑ Through standard prompt interface
❑ In simple human language
❑ No programming
Copyright©2025 Strategic Data Science Limited 7

AI chatbot
8
Can read, write,
listen and speak
But cannot action

Prompt interface (PI) +
Large language model (LLM)
9
Features
Smartness

Top tier generative AI platforms

OpenAI ChatGPT
❑ OpenAI
▪ Microsoft being a major shareholder (49%)
▪ Elon Musk once a founding investor
▪ Valuated at USD 500 bn in August 2025
❑ ChatGPT
▪ One PI
▪ Several LLMs at
different smartness
11

Google Gemini
❑ Transformer
▪ Attention is all you need
▪ A seminal research paper released in 2017
▪ The basis of ChatGPT
❑ Bard
▪ Released in March 2023
❑ Gemini
▪ Released in December 2023
▪ A replacement of Bard
▪ Now Gemini 2.5

Google Gemini free
❑ Very generous
▪ Free to use
▪ Lots of features
❑ Multimodal, multimedia, deep research, connectors
❑ Easy to use
▪ Best suit beginners
❑ Less stable
▪ A large number of active users

First task
❑ Please write an outline for a course titled
“Generative AI for Banking and Finance”.
❑ Please translate into Chinese.
❑ Please translate into Japanese.
14

Outline
15

Prompt
❑ The way to instruct a chatbot to deliver
❑ Give smart prompts
❑ Get smart deliverables
16

Simple vs conversational prompting
❑ Simple prompting
▪ New chat, one question and one answer
▪ New chat, one question and one answer
❑ Conversational prompting
▪ New chat
▪ One short question and one answer
▪ One follow up question and one follow up answer
▪ Another follow up question and another follow up answer
▪ ……….
❑ Always start a new chat for a new topic to
▪ Speed up
▪ Save cost
▪ Avoid interference from previous chats
17

Prompting rules
❑ Basic
▪ Tell the chatbot what you know
▪ Request the chatbot what you want
❑ Advanced
▪ Ask the chatbot what you don’t know
▪ Feedback the chatbot what you have been told

Simple prompting
❑ Generalist
▪ Knows everything but far below professional level
▪ Reply primarily according to public knowledge
❑ Each prompt
▪ Background information
▪ Simple instructions
19

Knowledge prompting
❑ Knowledgeable on a subject matter
▪ Text
▪ Web link
▪ Private knowledge file
❑ Reply according to the subject knowledge
20

Text
❑ Entered directly into the prompt
❑ My portfolio has 500 shares of Microsoft, 400
shares of Apple and 300 shares of Telsa.

Web link
❑ Web link pointing to a web page
❑ Limited to first layer
❑ Public but specific knowledge
❑ Warning
▪ A web page may change tomorrow
▪ A web page may not be readable

Private knowledge file
❑ A file containing private information
❑ For jobs beyond public knowledge
❑ To be uploaded
23

Outline
24

Prompt engineering
❑ Professional role
❑ Work style
❑ Language
❑ Writing style
❑ Knowledge
❑ Action
❑ Audience
❑ Purpose
❑ Structure
❑ Zero shot
▪ No example
❑ One shot
▪ Example
❑ Few shot
▪ Example 1
▪ Example 2
▪ Example 3
▪ ……

Professional role
❑ Government official
❑ Compliance officer
❑ Teacher
❑ Script writer
❑ Accountant
❑ Salesman
26

Role with further description
❑ You are a compliance officer working in a
retail bank in Hong Kong
▪ “Compliance” then means meeting regulatory
requirements from the HKMA
❑ You are a compliance officer working in a
listed company in Hong Kong
▪ “Compliance” then means meeting listing rules
from the HKEx
27

Work style
❑ Conservative
▪ Your work should be as deterministically and factually
as possible, without any speculation or creative
interpretation. Provide only direct concise and
objective responses based strictly on known facts.
❑ Creative
▪ Your work should be as full of imagination. You will
craft fantastical stories, dreamlike art, and surreal
experiences, effortlessly blending reality with
imagination.

Language
❑ 100 languages, incud
▪ English
▪ Chinese
▪ Japanese
▪ French
▪ German
▪ Russian
▪ …
29

Writing style
❑ Simple
▪ Easy to understand
❑ Legal
▪ Official but difficult to understand
❑ Tough
▪ No room for any negotiation
❑ Academic
▪ Theory, deduction, conclusion and validation
30

Knowledge
❑ Text
❑ Web link
❑ Knowledge file

Duties
❑ Summary Make it short
❑ Explain Make it long
❑ Rewrite Make it professional
❑ Bullet For presentation
❑ Extraction Show key points
❑ Relation Linkage between any two
32

Audience
❑ Primary school students
❑ University professors
❑ Corporate sales
❑ Computer programmer
❑ Medical doctor
❑ Lawyer

Purpose
❑ To select courses
❑ To assess a customer’s risk tolerance level
❑ To highlight the key features of a new product

Structure
❑ Course outline
▪ Learning objectives
▪ Target participants
▪ Major topics
▪ Class activities
▪ Assignments
▪ Test and examination
35

Example
❑ What will the deliverable look like?
Stock
Name
No. of
Shares
Latest Stock
Price (USD)
Stock Value
(USD)
Microsoft
(MSFT)
500 503.32 251,660
Apple
(AAPL)
400 211.16 84,464
Tesla
(TSLA)
300 313.51 94,053

Example prompt
❑ You are a college professor of finance. Your work should be
as deterministically and factually as possible, without any
speculation or creative interpretation. Provide only direct,
concise and objective responses based strictly on known
facts.
❑ Using simple English, in bullets and 200 words, write an
outline on a 30-hour course “Generative AI for Banking and
Finance”.
❑ This outline is prepared for a student to assess whether the
course matches his/her learning plan.
❑ Please include, among others, the title, learning objectives,
class contents, class activities, assignment and test.
37

Spelling and grammar review
❑ You are a professional English editor. You will
be provide an article. Please correct the
spelling and grammar. Show the final version.
❑ What changes have you made?
38

English article
❑ Here is the English article:
❑ Generative artificial intelligence (generative AI, GenAI, or GAI) is artificial intelligence capable of
generating text, images, videos, or other datas using generative models, often in response to prompts.
Generative AI models learn the pattens and structure of their input training data and then generate
new data that has similar characteristics.
❑ Improvements in transformer-based deep neural networks, particularly large languege models
(LLMs), enabled an AI boom of generative AJ systems in the early 2020s. These include chatbots
such as ChatGPT, Copilot, Gemini and LLaMA, text-to-image artificial intelligence image
generation systems such as Stable Diffusion, Midjourney and DALL-E, and text-to-video AI
generators such as Sora. Companies such as OpenAI, Anthropic, Microsoft, Google, and Baidu as
well as numerous smaller firms have developed generative AI models.
❑ Generative AI has uses across a wide range of industries, including software development,
healthcare, finance, entertajnment, customer service, sales and marketing, art, writing, fashion, and
product design. However, concerns have been raised about the potential misuse of generative AI
such as cybercrjme, the use of fake news or deepfakes to deceive or manipulate people, and the mass
replacement of human jobs.

Limitations
❑ Heavy human interaction
❑ Lots of follow up efforts
❑ Many prompts for complex tasks
❑ Subject to both human and chatbot errors
❑ Difficult to re-use prompts
❑ Difficult to replicate results

Prompt library

Outline
42

Stablecoin regulations
❑ Regulatory analysis
❑ Presentations

Regulatory analysis
❑ Summary
❑ Presentation slides
❑ Frequently asked questions
❑ Profit and return analysis
❑ Top three impacts to the banking industry
❑ Industry implementations

Gamma AI

KYC for wealth management
46

Outline
47

Use case
❑ You are a fresh graduate reporting duty to a
retail bank today
❑ Your first assignment is to conduct a
presentation on stablecoin regulations to a
group of senior executives at 4 pm today
❑ You are provided a stablecoin guideline issued
by the HKMA
❑ No other colleagues can support you due to
their tight schedule

During the presentation session
❑ You show a set of professional quality slides
❑ You conduct a 30 minute oral presentation,
including Q&A
❑ You provide insights from a banker’s
perspective
❑ You response to any relevant questions from
the senior executives

Starting from next
Monday, you work
10 times faster

Chapter 2 - AI chatbots and prompt engineering.pdf

More Related Content

Similar to Chapter 2 - AI chatbots and prompt engineering.pdf (20)

Recently uploaded (20)

Chapter 2 - AI chatbots and prompt engineering.pdf