Transcripts
1. Class Introduction: You've tried ChatGPT, but still feel you're not quite
getting the best out of it, then this course is for you. In this course, I'll
guide you step by step through how to write effective prompts
that save you time, whether that's for work, creative projects
or everyday tasks. You don't need any
coding knowledge or technical background, just curiosity and a
willingness to learn. Hello, I'm Nancy
and I've been using ChatGPT since it
launched in 2022. I've also worked for an
AI company where I helped design system prompts
for chatbots including ChatGPT, Claude and Gemini. We'll be focusing on
ChatGPT for this course. But what you learned
will help you to write prompts for any other
chatbots you might use. AI is now built into tools many of us use almost every day. So if you're ready to unlock
ChatGPT's full potential, let's get started, and I'll
see you inside the course.
2. How ChatGPT works - why prompts fail: In this lesson, we'll take a brief look at
how ChatGPT works, so you can get the best results. Most people use ChatGPT
by typing in a few words, the way they search on Google and hope that something
useful comes back, sometimes it works, but it can also be a bit hit and miss. The reason is that AI tools like ChatGPT don't actually
understand what you they recognize
patterns and predict what tet should come next
based on your words. Think of ChatGPT as an incredibly fast,
literal assistant. If your instructions are vague, it fills in the gaps with whatever seems most
statistically likely. Here we have an
obvious week prompt write about marketing. This is very general
and gives ChatGPT too much freedom to write
in whatever way it wants. Compare it to this
strong prompt. Write a 200 word email to small business owners
explaining why customer reviews
matter for local SEO. Use a friendly,
encouraging tone, and include one practical
action they can take this week. This is specific and explains exactly what
we want ChatGPT to do. When we talk about
prompt engineering, it's simply learning how to communicate clearly
with AI tools, so you get exactly
what you need. The key is giving ChatGPT four essential pieces
of information role, task, tone, and format
or RTTF for short. When any of these
pieces are missing, the responses you get back
won't be the best they can be. Now that you have
some understanding of how ChatGPT works, let's move on to the next
lesson where we'll put RTTF into practice and learn how to fix common
prompt failures.
3. How RTTF fixes bad prompts: Even when you understand
how ChatGPT works, some prompts will
still fall flat. Maybe the answer sounds vague, robotic, or
completely off track. The good news is
that you can fix almost any weak prompt by
checking four simple elements. Role, task, tone, and format. Here's a quick overview
of what each does. Role, this sets the
role and the audience. Task this defines the goal. Tone shapes the personality
and style of the response, and format controls
the format and length. Now, let's look at the four most common
prompt failures and how RTTF fixes each one. Number one, lack of role. It's important to tell ChatGPT who it is and
what perspective to take. Without a role, it has to guess, and that's usually
when responses come back generic or just
not what you expected. Let me show you what I mean. I I type into ChatGPT, write an introduction for a
presentation on teamwork, ChatGPT has to guess everything, the audience, the tone, the purpose, the setting. And because it's
guessing, the responses, again, will usually
feel pretty general. Watch what happens
when I add role. You're a project
manager introducing a new teamwork initiative to the sales department
in a computer store. Write a confident, friendly, opening paragraph
that encourages collaboration and
highlights one key benefit. Now I've given it a role, a specific audience, a situation,
and clear expectations. The response becomes much more specific and much more useful. A second common mistake
is an unclear task. Sometimes we think the goal of our prompt is obvious,
but it isn't. This is a weak prompt
about leadership. Now, the response we get
back is good overall, but again, it's very general. Here's a stronger prompt. Write a 150 word
linked in post about leadership challenges
for first time managers. Oh This is great because ChatGPT has
adapted it to the purpose, which is a linked in post
and for the audience and the length appears to be about
150 words like I asks for. Now, let's look
at something else that can cause problems
with prompts. Missing tone. If you don't specify tone, ChatGPT invents one,
and what it gives you might sound too
informal or bland. Here's an example of a
prompt missing tone. Explain our new flexible
working policy to the team. Now, let me paste in a revised
prompt to add to the tone. Explain our new,
flexible working policy to the team using a casual, friendly tone that addresses common concerns and makes people feel excited
about the changes. The tone shapes personality. This is so much better and
specific to what I want. Et's move on to the final
common mistake for format. If you don't tell ChatGPT
how to format the answer, it again, decides for you. And you might end up with long paragraphs when all you wanted was just a
quick checklist. This is weak because it doesn't specify any format
for the response. Give me productivity tips. Now let's make it strong. List five productivity tips as bullet points with one
practical example each. So here we have five
bullet points of productivity tips and with one practical example,
as was asked for. Perfect. Always tell ChatGPT exactly how you want the
information displayed a list, table, script, or summary. In the next lesson,
we'll build on this with some advanced
techniques that'll make your results even more
consistent and professional.
4. Advanced prompting techniques: I All right, you've
mastered the RTTFFramework, role, task, tone and format. That alone is going to make a huge difference
to your results. But if you want
Chachi PT to produce consistently professional
polished responses, here are three advanced
techniques that will take your prompting
to the next level. We'll cover number
one, using examples. Two, refining through iteration and three test in
variations. Let's dive in. Number one, example
based promptin. If you don't provide
any examples, then that's called
zero shot promptin. That means Cha GPT has
nothing to work with, but will do its best to give you a response it
thinks you want. Give it one example, that's one shot promptin. Provide two or three examples
that's few shot promptin. Whenever you add an example, it helps Chat GPT to better
understand your desired tone, structure, and the level
of detail to provide. Let me show you what I mean
without examples, zero shot. Create a social
media caption for a coffee shop with examples,
few shot prompting. So I would type into Chat GPT, write a social media
caption for a coffee shop, then say, here are the two
examples of the style I want. Example one, Monday
motivation starts here, fresh beans, warm vibes and your favorite corner table
waiting for you. Example two, that 3:00
P.M. Slump hitting hard, we've got the perfect
remedy brewing. Come grab your
afternoon pick me up. Write a caption for our
Tuesday morning post. So remember to provide
Chant GPT with some examples if you want it to have a particular
tone and structure. This gives it a clear
reference to mimic instead of forcing it
to guess what you want. Let's move on to number two,
refining through iteration. It's likely your first prompt
won't always be perfect. Sometimes it's missing a bit of detail or the tone isn't
quite what you wanted. Instead of starting over, you can just build on what Cha GIPT has already given you. Here are some simple
follow ups you can use to refine
Chachi PT's responses. Make this more conversational, add more technical detail. Shorten this to 100 words. Change the tone to be
more professional. This process is
called iteration. Think of it like co writing. Chat GPT gives you a
draft and you guide it, and each adjustment brings the result closer
to what you need. I use this all the time when
I'm polishing an email, tightening up a blog
post or generally shaping ideas into something
clearer and easier to read. Let's now look at number
three, testing variations. Once you've refined a response, try making small changes to see what gives you
the best result. You don't need to rewrite
the whole prompt. Just change one thing at a
time and compare what happens. You can vary things like
the examples you include, the role or perspective you
want Chan GPT to adopt. The tone and the
format you asked for. Let's use tone in this example. You can ask Cha EPT for the exact same response
in different tones, for example, friendly,
neutral, and expert. Then just compare them. Notice which version feels
closer to what you need, and keep using that
tone going forward.
5. Fixing common mistakes: By now, you've seen how powerful RTTF can be when you
use it consistently. But even when you're
doing everything right, sometimes Chat GPT will still
give you something vague, too formal, or just not
quite what you meant. So in this lesson, we'll look at two simple things to keep in mind when a response doesn't
turn out the way you hoped. So common mistake one is
asking for too much at once. Sometimes we try to get Chat GPT to do
everything in one go. Now, Chat GPT doesn't
know what to prioritize, so the result can feel confused. Instead, try breaking it
down into clear steps. One, summarize the
article in 120 words. Two, write a compelling
title for the summary. Three, expand the summary into a two minute
YouTube script. This will give a
much cleaner result. And common mistake two, forgetting to check facts. Chat GPT can sometimes produce confidence sounding information that isn't
actually accurate. This is what we call
hallucinations. If you're working on something
where accuracy matters, simply add into Chat GPC, use only verifiable facts and flag anything that
may be uncertain. And if it's going to
be shared publicly, just double check
key details with other sources. Now
it's your turn. Take one of your earlier
prompts that didn't quite work, rewrite it using
RTTF and if needed, break the task into
smaller steps. You'll see an improvement
straightaway. So you're ready. You now have everything
you need to use Chat GPT effectively a
clear framework R TTF. Techniques to refine
and improve responses. A simple way to troubleshoot
when things feel off, and I've included a downloadable
quick reference guide with five ready to
use prompt templates. You'll find it in the
course resources. Also, do have a go at
the class projects. I want to thank you
for taking this class. If you found it useful, please leave a review. It really helps others
discover the course. I hope to see you in my
next one. Bye for now.
Nancy Sealy, AI & Digital Skills Instructor
