Transcripts
1. Class Introduction: Hello. Hello, everyone. My
name is Ronnie Hermosa, and this course is for non technical teams who feel
a bit overwhelmed by AI. Maybe you have been
using Chat CPT. You've been using it to rewrite emails or maybe brainstorm
your social media post. But if I ask you,
could you plug AI into your everyday workflow and confidently say that
you are saving time, you probably would hesitate. And that's what I want to fix
right here with this class. This is a practical course. Not going to overwhelm you with 50 different tools or
bore you with theory. We'll focus on a small
set of workflows you can actually use on
your business tomorrow, like finding new markets, communicating more clearly
with international buyers, and showcasing your products
under their best light. I'll also share hands on prompts and real use
cases you can plug straight into your day
to day work and that even if you don't consider
yourself tech savvy. The goal is simple. Work faster, reduce busy work, and make better decisions
with less effort. And ideally, you will also
save money because you stop outsourcing things you can
handle in quick context, the core of this class comes
from a live workshop I delivered last week to
30 plus W FTO members. So that is ethical and
fair trade brands. Most of them were in fashion and retail selling products
like clothing, home decor, and toys. I decided to turn that workshop
into this class because these workflows are too good to stay within that one Z honestly, because they apply to pretty
much any retail business. By the end, you'll know how to use AI as a daily assistant, how to do better research, how to get AI to write
strong prompts for you, how to review your website through the eyes of
your target buyers, and how to improve
product photography using an AI image generator. In the course, I'll
be using Leonardo AI. To generate my photos. But if you prefer, you can
use any other tool like HGPT or the Google AI models
to generate your images. All you need to get started is access to your
favorite AI chatbot, like HGPT, Gemini, or Cloud, plus an AI image generator. Alright, let's jump in.
2. How I Use AI for Everything: I want to start this course
with a practical example that illustrates how I
think about AI day to day. The more I use it,
the more I figure out ways to integrate it
in my daily work. Not as one magical prompt, but as a habit. Small decisions, small task,
but repeated constantly. So what you'll see
next is the actual way I opened the workshop
with the 30 organization. I decided to keep
that passage in the course deliberately
just to make a point. Let's get into it and you'll
understand what I mean. Making AI work for
your ethical business. Before we jump in, you might notice something
on this slide. Okay, this slide is kind
of missing the photo here, and I did this on purpose. It's not a mistake. It's missing a photo because I wanted to figure out which is the best photo together
with you guys right here. So I'm going to go for the
next slide and ask you to use the little chat that we have
right here in the Zoom core, the webinar to tell
me which one of these four photos do you
think best represents WFTO? Curious to see your opinion. I think this is a nice
little icebreaker, as well. So yeah, just use the chat and tell me number one,
two, three, or four. Which one of these four photos best represent WFTO
in your opinion. Four, three, four. Okay, a combination. One, four, one. Okay. I've got all of them. So you know, in order to actually get an answer
to my question, I decided to also ask CHAT GPT. Okay? So I took a screenshot
of my four slides. Which one of these
four cover slides is best for the first page of my presentation for the webinar. I'm trying to find the
most appropriate photo. The audience is a bunch
of fair trade companies, part of the WFTO network. And then I added the link
to the WFTO website. Here is the short answer. Short answer, I had a
longer conversation, but the short answer is
for WFTO specifically, I would go with
slide number four. Why? Because your promise is making AI work for
your ethical business. The visual should then feel
relatable and not too flashy. Slide number four feels
like someone actually running a small ethical
enterprise, figuring things out. My slide for this webinar, my intro slide should
be slide number four. I also agree that most of
these would actually work. And if you look at my
presentation right here, you see those pages. I had them hidden, but you
can still see them here. But I agree that the first one, this one right here or actually, this one right here should come all the way here
to my first page, and I'm going to go
ahead and delete my first slide and now delete
all of these, as well. And we now have with AI
figured out the best option. Why am I showing you this
and opening up with this? Well, I want to show
you at what level AI has kind of penetrated
my everyday work. I use it for everything. I use it in my
everyday workflow, and I use it to take
all sorts of decisions, little decisions, important
decisions, big decisions. And this is really
today's topic. I want to pass on some of the practices
I've been testing, some of the workflows I've been experimenting
with so that you can start applying them in
your fair trade businesses, in your ethical fashion businesses because
they are important. Okay, so that's really
the topic of today making AI work in your
everyday workflow.
3. Making AI Work in your Everyday Workflow: This lecture is the big picture. I try to break down the AI tools a small retail business should be using to speed up
their everyday workflow. I'll focus on areas
where I actually do have the most expertise
like everyday task, content creation, branding,
and visual communication. So the way we are
going to do this, I'm going to try and hit as
many nails on the head about what fair trade
organizations actually do and actually need in
terms of their workflow. Okay? So I've tried to list here a bunch
of different tasks that an employee or a person working at
a WFCO organization, like Fair Trade Organization actually does on a day to day. Obviously, you do
much more than this, but this is kind of where
I can bring my expertise. So everyday tasks like email,
research, client proposal, sending proposals,
product descriptions, negotiations, and also
asking for feedback. In terms of content
creation design, we have, yes, content
creation for social media, editing photos,
generating mockups, creating consistent
brand, writing reports, documents,
marketing campaign. And then on the visual side, where we need to take
product photography, lifestyle photography, visual storytelling,
all of these things. So what I put here on top of
each of these columns are the tools that I am currently using to achieve these
different things. Okay? And we are going to cover some of that in the workshop. See how they work, see how
we can start using them and see what kind of prompts
we should be using to use. Now that we have the
overall approach, let's get practical. In the next lecture,
I'll show you how I use AI as my text support, and I believe this is the
easiest win for most teams.
4. Which AI Model Should I be Using?: In this lecture, I answer one question from the
webinar's participants. And I have decided to
keep this question in the course because it's a very common question
when working with AI, beginning to really implement the AI tools in
your everyday work. So I figured if this person in the
webinar had the question, you might have the
exact same question. Now, the question
was do we know which AI took to use,
hGPTGmin et cetera? It's really a question
of preference. I think all of these, HAGPTGemini or Claude,
they are all very good. So there is no fundamental
difference for people like us. I mean, we are not
rocket scientists or we are not data engineers or we don't need to use the models to their
very last frontier. If we are using it for
everyday business use, you won't notice
much of a difference between the level of
intelligence of these models. The difference is when
the user experience, so how the interface looks like and the
features that they have. Open AI, so the creators of
CHANGPT have been very good at making HGPT sticky
by giving it a memory. So because it has
this memory feature, it knows you, right? And because it knows you, it feels comfortable
to just keep coming to the same tool because
you don't have to explain all of
your life story, so to speak to the model. It already knows you,
it knows what you do, it knows what
projects you've been working on because of that so that's kind of what keeps you locked in a tool
once you start using it. So if you are at the very
beginning of your journey, I would say, I don't know, like build for example, Gemini is more integrated with
the other tools you might be using like Gmail or Google
Sheets or Google Docs. If you buy one of the
cheapest plan for Google AI, you will have access to
the power of the LL. Of Gemini in those
different platforms, like your sheets, your
Gmail, everything. It will be integrated. So it's better integrated
because it's Google. On the other hand,
HAGPT is kind of like the preferred
consumer AI platform because it has this memory
and it's easy to use, and it was one of
the first one or the first one to kind
of breakthrough. Businesses prefer anthropic and cloud because of the trust and the kind of
ethical consideration this company is kind
of focusing on. So it's the preferred
business solution. Yeah, I don't have me. Personally, I use HAGPT because I'm locked in
because of that memory, because of all the context. I've been finding myself also
experimenting with Gemini. Now, let's move forward
with the course and see how I use AI as my
everyday tech support.
5. Using AI as your Tech Support: If you only use AI for
one thing this week, make it this tech support. Anything confusing on your
computer, on your phone, a feature you can't find a
bug that is happening and preventing you to
move forward or tool you keep hearing
about and you want to try, just ask your
favorite AI chatbot. Let me show you exactly
how I would do it. The main benefits
for me of using AI, sometimes people don't
realize really what AI brings them or what AI can
bring to an organization. For me, it's very clear me
work smarter and faster. So smarter because
I feel like I have the super intelligence
at my fingertips. These models know everything, and so every time
I have a doubt, I have a question, there
is something I don't know, there is something not
working on my computer. I rely on AI to give me the answers or find the
answers or teach me. So that's why I also put
feedback in coaching. I use these models to
get a lot of feedback on my work and to kind of
get some free coaching. Let's see some of the
examples of me doing that. Using ChRGPT is my tech support. Just going to give
you a few concrete examples of my stuff, and then we'll dive
into practical stuff, practical examples for
fair trade organizations. But first, the first pillar, how do I use it in my
everyday workflow? Well, recently, I noticed this new little button in my
ChatGPT app on my desktop. So you can either use it
as your browser, ChatGPT. That's why most people
use it in your browser. But there is an app
or that you can download and start using it
locally on your computer. And they added
this little button right here. Didn't
know what it was. So I just took a screenshot. What you see here is the
screenshot and prompted. What is this new feature record in the Chan GPT Mac desktop app? Okay? And I can click here
to go to that conversation. I want to show you the
actual conversation. Yes. Zoom in a bit. So that's my prompt. And it started giving me
basically the answer. So when you hit
that record button, you can start using the
new voice capture mode, and it just gave me a bunch of information about what it is. And then I continue to prompt or to ask more questions asking, for example, can it record a Google met or a Zoom
call? Can it do that? And can you also tell me what other participants are saying in this call? And
then I got my answer. No, actually, I
was trying to find the best solution for
an AI note taker. I've noticed in
several Google Meet or Zoom coors that I was part of it was this little
butt that was just recording everything that
everyone was saying. And after that
meeting, I receive an email with all the notes, everything that has been
said, action points. And I thought, This
is super convenient. Like, I would love to have something like this for myself. So I was trying to
find that product. So I continued asking, What is the best solution for AI note taking during meetings? Give me some ideas,
and then continued, I need the best tool that allows me to record both
Google Meets and Zoom coors that I organize and also other cores that
other people invite me to, et cetera, idea here is to
show you how you can start a conversation with a
simple question right here was more like a tech
support kind of question. What is this?
Sometimes something is not working on my iPhone. And I ask, how do I
get to this feature? How do I make this work? And then it kind of shifted
to something more like in the realm of advice or
product review or whatever. But it gave me my
answer at the end, and I started installing
this fathom app, so fathom right here,
which is great. It's just an AI note taker. Free. You just invite it to your Google Meet
or your Zoom Coll and it will start recording, taking notes, and send every
participant a summary. That's just one
little example of how making me work smarter. It would have taken
me a long time to actually go there and analyze the market for an AI note taker and compare
them with each other, try a few different ones, didn't want to do
all of that stuff. So just got my
answer very quickly. Okay, tech support
is, in my opinion, kind of your way of putting
your foot in the AI door. Once you trust it for
solving small problems, you start to use it for
bigger kind of work. Next up, I'll show you how I use AI to do all sorts of
research and also give you a few tips on how to get precise answers instead
of generic fluff.
6. Using AI to do Research: This lecture is about
research, not Googling, but real deep research where the AI will go out,
explore the Internet, explore various sources,
and come back with a structured report
that could be formatted in whatever
format you prefer. And with all the sources it has used cited in your
research paper. And you'll see I'll be doing
a research for my own work, so it's something
related to Canva, but the method can
work for any industry. So obviously, feel free to
apply this to your own. Something else is
really good at. I'm referring to the LLMs, the large language models. So that is the likes of a ChatGPT anthropic Cloud
or a Google Gemini. So these are what we call LLMs, large language
models or chatbots because they materialize in the form of a chatbot,
where you go, you ask so these LLMs are
very good at research as well because they can work
super fast and they have access to all the
information that is online. They've been trained on that. So most of them have this functionality
called deep research, which will allow you
to ask the models to really go out and
research a specific topic, and it will come back
with after maybe 10 minutes or it
depends on the models, and it depends on how long and how hard you ask them to think. But it takes on
average 10 minutes, and they can produce really in depth reports that
you can use about any specific I want to show you this one example
and then I'll move on to some more practical
fair trade stuff. But this one is a prompt that I have used for my own work. So my work as an educator,
Canva expert, influencer. Okay? So I said, act as
a research assistant. I need a comprehensive database of the questions people
ask about Canva. And then research sources, the task, the output. So I told model, I need you to research what people are
asking about Canva. I have been very specific about the task I wanted it to conduct. So build a topic taxonomy
for Canva learning, extract the actual questions, phrasing people use,
etcetera, et cetera. And the output, how do I
want this output delivered? I want a table of 200 to 500 i questions with different columns
in that table. I launched this prompt, took about 10 minutes and I
got a massive report with questions that
people had about Canva from all sorts of places. I want to show you this report by going back to the
conversation right here, which is this one right here. Yes. This is the report. So it's super long. I'm not going to go over this. This is the table I asked for, and it goes more on This
side is more columns. Obviously, you can open this. You can open this
directly into a, and this is all the
sources that it has used or identified to
conduct that research. So I'm showing you this
to show you, okay, the possibilities
are almost infinite, right, about what they can do. So the deep research so
I think it's important to understand how to ask the best questions.
So this one right here. You might wonder, Okay, Ronny, this is nice, but the
prompt that you use, this part right here is
something I don't know how to create because
with these tools, the quality of the output will depend on the
quality of your input. If you ask a question that is not well structured,
detailed, precise, if you ask something like
you would ask Google search, you won't get this
kind of a result. Deep research with
AI is powerful, but just like for anything else, the quality of that output
will highly depend on how you ask the AI to work to do that research
in the first place. And in the next lecture, I
will give you the shortcut. Today, you don't
need to give the AI very elaborated prompt
for good results. You just need to tell
it exactly what you want and some context
to work with.
7. Asking AI to Write Prompts for You: What I'm about to
show you might be the most important takeaway
from this entire course. Stop trying to write
the perfect prompt. Instead, use AI to write
the prompt for you. In this lecture, I'll show
you how I use ChatGPT to turn a messy brain dump into
something well structured. Two years ago, it
is true that you needed to know how to
write a prompt like this. But today, you don't
this is my point. Like, this is this
slide right here. You don't need to be a prompt
engineer, prompt engineer, meaning the person who
engineers the prompt, the questions you ask the you need is to know how to ask
the model to help you. Okay? So what you see
here on this page is the actual prompt that I input it in order
to get to this. Okay? So let me
explain. In ChatGPT. So let's just go to the
ChatGPT conversation. All right. So coming back up. Yes. So what I did to start this conversation is
this prompt right here. And this prompt,
I didn't type it. You see this little
guy right here, the little microphone
that says dictate? When you click on that, I'm not going to do it now because it's going to
prompt something else. But when you click on that, then you can use your microphone to give some context.
That's what I did. If you read this,
you will realize this is mostly spoken language. I said, I'm trying
to build something kind of like a
dashboard where I could monitor what other
creators who talk about Canva are publishing on
different social media platform. So the goal for this tracker is to keep track of
what's going on. I started the conversation
like this and it gave me, this is the structure you
should have in your tracker. Okay? Then I asked it for the prompt that I would use for a deep research.
I did the same thing. I didn't know how to structure this well structured prompt of the task, the
output, and everything. I just tell it, help me with a prompt for a deep research, both on ChatGPT,
Gemini, and perplexity. These are all different
kinds of LLMs. I want to do a deep
research to find out what people are asking about
Canva and how to tutorials. I'm trying to create
an exhaustive list of everything that a beginner or intermediate Canva user can
to learn on the Internet. Not going to waste your
time and read everything, but this part is important. You don't need to be a
prompt engineer anymore. You just need to know
how to use a microphone and explain with your own words what
you are trying to do, what you are trying to build, what kind of outcome
you are trying to you can do so in a messy
all over the place way. It was relatively
well structured, but sometimes some
of my audio prompts are all over the place that it's okay because
the mothers are so intelligent that
they will understand. I launched this prompt and I got the prompt that I read
to you previously. Tim previous page, this one, the very well
structured prompt that was optimized for
my deep research. I went from this c
audio brain dump to a well structured
prompt because I was able to use the model to
create the prompt for me. Okay? Once I had this, I could use that little copy
button, which is what I did. I used that, and
I came to Gemini. So the other model. So Gemini is from Google. There you go, so I came here, and I inputed that prompt somewhere here and
it gave me the whole thing. So once you copy it, you can simply paste. I didn't paste here, but, like, so you paste your prompt
and you're good to go. Just make sure you use
the deep research tool. So under the tools right here, deep research, make
sure it's selected. If that's what you
want, if what you want is a more in depth kind of yes, research with an outcome
that will be a report, large spreadsheet or
something like this. Okay, enough time spent on this. This is
an important one. You don't need to be the
master prompter anymore. Just ask the models
to write your prompt. We are ready to apply all of this to real business assets. So in the next lecture,
we'll see how we can use AI to review your
website and improve it. And to do so, we'll
try to review it through the eyes of
your potential buyer.
8. Improving your Website with AI: If you sell to
international buyers, your website is either
building trust or killing it. In this lecture, we use AI to
audit a website's homepage, like a serious buyer. And I'm going to use a
real example that is the website of one of the organizations that
attended the webinar, and I want you to treat
that as a case study. And then you can repeat the same process to
your own website. Let's start with websites. I've been given a few examples when I was preparing
for this webinar. One of them was Mahaguti
I decided, Okay, let's see how we can maybe have a look at Mahaguti's website, use AI and see how we can
maybe make some improvement, get a nice feedback
and understand how we can work with AI to
improve our website. The first task was, how do
I figure out the prompt? Just like I told you,
not a prompt engineer. I started with my microphone
and my audio prompt. I asked HGIPTT is ha have
the conversation right here. If you scan this, you
should be able to access that same conversation and
continue it for yourself. Maybe scan it if
you want the link. So took my microphone,
and this time, I'm going to read the
whole thing because I'm going to take you with
me on that journey of using hATIPT to review Mahguti's website and give me some suggestions
on how to improve it. So when you analyze
a company's website, specifically a fashion
retailers website, would you evaluate the website? Okay, that's the beginning
of what I said my prompt. And then I give it some context. So I said, I mean, what are the different
criteria that are important? Because I'm doing
some website reviews and feedback for ethical
fashion producers, fair trade producers,
and most of their website are
not super good. So I would like to help
them by giving them some constructive feedback on how they could improve
their website. Of the time, these companies
are looking for finding international clients who
would be distributors, importers of their products. Sometimes they also sell
direct to consumers. So they want to present their collections under
their best light. So product photography
is important, but also stories like
impact stories of how their products actually empower local communities, et cetera. So stories, product photography, they need to show
professionalism so that international clients are
willing to work with them. So all of this is kind of
reflected by their website. You see how this is
spoken language. You can see that when you
read this prompt because, yeah, it transcribes everything
I say in my microphone. So your task, I need
you to give me a list of criteria for my
website analysis. What should I focus on? What should I pay attention to, and what kind of feedback would be helpful for
these organization? That's how I started
this entire project, and it's right here on
this HGPT conversation, so I'm going to go back to it. Alright. So the answer. Alright, if I land on a fashion brand website
and they tell me, we want international distributors
to take us seriously, I'm not looking at colors first. I'm asking one brutal question. Would I trust this company
with a 50,000 euro order? That's the is how
I would break down a proper website review for ethical and fair trade
fashion producers. Not going to read
the whole thing, but there is a few
different points here. One, first impression
and positioning, what it calls the
five second test. So do I immediately understand what they
sell? Who is this for? Retailers, wholesalers,
consumers. Where do they operate? What makes them different? Then it gives me
a common problem. So the homepage, in most cases, the homepages talk about empowerment for
three paragraphs, but never clearly
shows the product. So if that's the case,
feedback you can give Professionalism
and trust signals. This matters a lot for
international buyers. Look for clear
company information, location, year it was founded team, et
cetera, certifications, WFTO, fair trade, GOT, et cetera, red flags,
no physical address, no company registration info, generic stock photos
everywhere, broken English. So that's kind of like some of the red flags I should
be paying attention to. Product presentation. This is huge for fashion. Photography. Are these
images high resolution, consistent background,
consistent lighting? Are they style in a way
that matches the brand, multiple angles for the photos, close up of fabrics or
texture, et cetera, et cetera? Fb2b clarity, very
often missing. If they want distributors, the website must support that. So check if there is a
dedicated wholesales page. If it's easy to request a catalog if there is a
downloadable line sheet, if there is a clear
call to action like request our
wholesale catalog or become a retail 1.5,
storytelling and impact. Now, we talk about the heart. Ethical brands have
powerful stories, but often they are
told in a messy way. I should evaluate, is
the story structured? Do we understand the problem
that the company is solving? Are there any real
numbers, job created, percent women employed, wages, above local average, et cetera? Are there real faces and
names? It gives me an example. What's a weak version? We empower community. What's a stronger version? We work with 64 women
artisans in Waaka. Their income is 2.3 times
the regional average. It's much better than
we empower communities. Feedback, six, navigation
and structure. Is the menu clear? Are
there too many items? Is it logically grouped, easy to find products mobile,
friendly, responsive. So common issues, overloaded
menu with vision, mission, heritage, value,
craft, philosophy. It should be more
simple than that. Simplified to shops wholesale
about impact contact, reduce, clutter, continuing
point number seven, conversions and call to action. Do they want visitors to do? Buy, request a catalog,
book a meeting, sign up for a newsletter, language and copy quality. This is sensitive,
but important. Is the English natural?
Is it too emotional? Is it too vague, too activist? For B to B, especially,
buyers need clarity. So replace abstract phrases
with concrete details, shorter sentences, clear
benefits for retailers. The technical basics,
quick checklist, brand coherence, et cetera, doesn't give me an analysis
of Mahaguts website. So far, I haven't showed
Mahaguts website. All I've done here is ask for some criteria for me to
evaluate the website. If my goal is to
use AI to evaluate Mahaguts website and then make some improvement,
that's the whole task. I need to break down this
task into smaller task and the first smaller
step is to ask for some solid criteria for
evaluating this website. That's a common and that's something I would encourage
you to do if you have a project that is relatively complex,
improving your website. Break the task into smaller
steps because if you ask the AI just I make my website better
and you paste the URL? You will get some of this, but you will not get
this level of precision. Here, we have ten
very detailed points and each point could
be an entire webinar. You've got something solid. So break the task into smaller chunks and
move progressively. So now that I have all of that, what I did is simple. So I figured out the prompt, right? That's what I just did. My next slide here is. So, this one was a good one. Like, would I trust this
company with a 50 K or? Really the core of what we
are trying to solve here. Like, the website should
answer yes to this question. If your website can do that, then it means it's
actually working. If your goal is to
sell internationally, find some new customers, which I know a lot
of you and a lot of your companies are
actually working toward that goal.
I had my prompt. I had my contacts, my criteria. Now it was time to leverage it. And I did this by
switching tools. I'm not sure why
I went to Gemini, not continued in Chat GPT. But just also to show you that the tours are not really
the differentiator. I came to Gemini. I asked it because I wasn't
sure how to do this. Like, how would it
see my website? So I asked, If I give you
some URLs of websites. Are you able to access them and actually evaluate the
content of these websites? I'm talking about visuals. I'm talking about structure of the website copy of the
website overall structure. So basically, I'm looking for feedback about website design, copy and efficiency only
by providing you a URL. Right now, I'm just looking for yes or no answer.
Can you do this? So it's a very
straightforward question. I give you a URL. Can
you analyze my website? It said yes, I can access
the text, the structure, and the code of the website
to evaluate the copy, the SEO, search
engine optimization, and information architecture. While I cannot see the website in the way human does
as rendered image. So I can analyze the
underlying CSS and layout the data to provide detailed feedback on design principles, and user experience. Basically told me, yes, I can analyze your website, but I cannot see your website. I said, Okay, understood. Now, if I give you a screenshot or screenshots of
the landing page, could you see it or if I export it as a PDF,
could you then see? And then upload
that to the chat. Can you give me some proper
design feedback about visuals and visual consistency and the visual look of the website? So I'm trying to
find a work around, like, Okay, you cannot see it, but what if I give
you screenshots or I exported my website as a
PDF and just give you that? Can you then do what I'm asking? Absolutely. In fact, if you
provide screenshots or PDF, that's the best way for me to give you high quality
design feedback. Give me more blah,
blah, blah here, but I didn't know how to
export my website as a PDF. So I asked, Do you know a good tip to export
a website to a PDF? Scroll, because I'm not sure how to do that.
And it gave me. So this is me going back to the tech support kind of way
of using DAI. I need a tool. So it gave me a few options
like the no Install, browser, trick browser
extension thing. He told me there's a browser
extension for Chrome. It's called Go Fool
page. It's highly rated. It's very simple. You
click a camera icon. It scrolls for you and you can download the result
as a PDF or a PNG. So I thought, Yeah,
let's give that a try. So what I did, I just came to
the Chrome extension page. So let me show you something
like this, manage extension. And here in the
Chrome web store, you can download new extensions. So I went ahead, find
the Go full page. I installed it took 1 second, and now it's installed
on my browser. You can see the little
icon right here. So what I did, I came to
Mahagut's website right here, and I use my little
extension right here. So I clicked it and you see
what's going on, GofoPage. It's scrolling the page.
I'm not touching anything. It's just scrolling the page. You see the little
Pacman thing going. And once it scrolled all the
way down, it gave me this, which is a full scan
of my website page, the entire page from the top, the navigation to the footer. And here I can save that as
a PDF download as a PDF. Remember the long conversation I had with the website review, all of these ten criterias, I copied all of that stuff from hat I came to my prompt here. I uploaded the Maga
agutiHmepage PDF. I just created with my
tool, so it's here. So now you can see it and I
pasted all of that context. This is a super long prompt,
but I didn't write it. I just asked ChaGPT to
create it, then I run that. I gave me, here is the comprehensive framework
your b2b BTC audit, applying your brutal question. Would I trust this company with a 50 K order to the
MahagutiHme page and PDF? The answer currently
is no, not yet. So again, I'm doing
this respectfully. I don't want to
offend anyone here. I'm just doing the
reviews as, like, in a way that we can
all learn from this, we can all improve from this. Let's get into the reviews. Wilder Craft and mission
are clearly authentic. And maybe we should
have a look at the homepage for just a few sec. Okay, so we have a
slider right here, Artisans made ceramics or home decor. We have
different slider. Then we have a
message from Sunil, the CEO of Mahaguti. Then we have some top sellers, some nice and cool animation. Some links to direct product
categories, I imagine. You can shop by
categories. Okay? So we have the categories. You can navigate y Mahaguti. So these are kind of like the selling points
featured offers, pashmina, ceramics, home
decor, new arrivals. Again, it's a
carousel, Blog posts. They're still loading,
but some quotes, more products than statements about Mahaguti and the footer. Okay? So that's what we
are evaluating right here. Okay, so the first impression, the five second test. The hero slider mentions
handmade toys and figurines. However, the navigation and categories suggest a
massive range of product, home furnishing, clothing,
wearables, dining. So the problem, it
lacks a sharp headline. Is Mahaguti a toy shop? Is it the furniture shop?
Is it the textile provider? By trying to show everything, they position themselves as a general gift shop rather than a specialized producer
for ethical retailers. Also, it picked up
a visual conflict. Hero image shows pottery. But the text talks about toys. This creates immediate
cognitive friction. So actually, I
noticed that as well, Artisan made ceramics
for home decor, and we see toys. Then when we go to the next one, handmade toys and figurines, but we see the ceramics. So they kind of
mixed up the images together with the tag
line in that slider. And the AI picked that
up the visual conflict. Moving down here, professionalism
and trust signals. Credibility gaps,
the CEO statement is a wall of text that
is difficult to read. International buyers want data, not a manifesto in
the hero section. So I tend to agree with this. This is nice for explaining maybe the vision and who is
kind of navigating the ship. But for an international buyer, this doesn't bring much. Now, maybe that's
not the goal of this website to talk to
an international buyer. I'm not sure. Missing
B to B there is no wholesale or distributor
link in the primary header. A 50 k buyer doesn't want to use a contact us form to get
in touch with the company. They want to see a wholesale portal or download
catalog button. Certifications. While Mahaguty
is a known WFTO member, the logos are not prominent. Above the fold, they
need to be front and center to establish
instant trust. Okay, moving on to
product presentations, photography, the product shots
the notebooks that cause. Are inconsistent. Some have harsh shadows and
the actual framing, the cropping of
the photo varies. Styling, the featured
offers images are much stronger than the
individual product listings. However, the mix of high end
lifestyle photography and basic product on white shots makes the brand feel
disorganized. Technical specs. On the homepage, there is zero information on
material or sizing. A buyer needs to know
if it's 100% lock up paper or handmade hand
loomed cotton immediately. So it gave I would say a
very thorough analysis of the website storytelling and impact vague
versus concrete. The copy uses phrases like preserving Nepal's rich
cultural heritage. Improvement. They
should replace the CEO long paragraph with mahaguty
by the numbers section. Artisans supported
percentage of women led WFTO certified since
why, why, why? I agree with that. Quick wins. Fix the typography. So this one it picked up a
question mark somewhere. I'm not going to talk
too much about this. B2b called to action, had a wholesale
international trade button in the top right corner. The header, so a
wholesale section or international buyer, hero alignment, ensure
the text handmade toys matches the images of
toys, not the ceramics. So a few things to fix here. Go to leave it here for
this example because me, honestly, I was
quite blown away. It picked up the misalignment of the slider with the
text and the image, which for me, that
was showing me that it's a deep analysis
that went on. Your website done. Next, we move to the one
thing that I believe will make the difference
between people clicking that By button or
closing the tab, and that is your
product photography.
9. Improving your Product Photos with AI: This lecture will
teach you how to use your favorite LLM to improve
your product photography, how many images you
actually need per product, which angles matter,
what to add, what to reshoot, and what most
retail sites are missing. I'll show you the
framework first and then we'll see
how to apply this to a real product page with some real product photos from one of the
webinars participants. But again, you take
that knowledge and apply it to your own
product photography. The AI models have
become sparingly good at creating real images. There's obviously all of the
ethical considerations here, so I want to open with this. What I'm going to show you, I'm not saying that
you should per se, use that all of the time, but I believe it's useful. But it should be used ethically, meaning we can pretty
much create images of whatever we want nowadays
with the right prompt. It's important that in
the framework of being a producer organization
producing products that the photos that you will present to your buyers and your clients, they need to be faithful to
the reality of the product. So it's tempting and easy to embellish the product
with these technologies. But I encourage you to use
the technology responsibly. So that being said,
let's dive into it. Let's see how we can produce smarter visual
production with AI, and we are going to focus
on product photography. Alright, so the next use
case I'm going to use is KTS Combahar Technical
School organization from Nepal Kathmandu which I had the pleasure of
visiting and work with. So the specific pages
I'm looking at, let me get to KTS website here. So that's their website, and they have on the product, they have a bunch of
different product categories. So I think the one I
analyzed was the carpet. So this is what we see. They produce these beautiful
carpets hand woven knitted. So that's the product
photography. Okay. So there's only one
photo per product, and they're mostly all
taken in the same room, same kind of angle. But I wouldn't say
it's really presenting the product under its best light because this is
natural lighting. I mean, best light in in the literal sense and also
in the metaphorical sense. Product photography
could be improved. Let's say there's room
for improvement here. The specific use case that I use for demonstrating this section is this particular
carpet right here. What I did here,
I right clicked, I saved the images, and I just save that
image as a PNG. I had the image of the carpet. Now let's come back
to my presentation. Started from my favorite LLM, which is Chachi PT to try
to figure out the prompt. Let's jump into
that conversation. All right. So I said, with
my microphone, no writing. I'm trying to help the
fair trade organizations improve their
product photography. I'm going to provide some
screenshots of their website, the part of their website
where they sell their product. They showcase their products. Okay? I need you to give
me a detailed analysis and feedback on how they present this product
via photography. What can they improve? What kind of shots would better present these products,
how many shots, how many photos per product would be a good
average for presenting one specific product kind of product shots angles would present the product
under their best light. Give me some
detailed suggestions on how to improve their
product photography. You hear me rambling
a bit in this prompt. It's not a perfect prompt. It's just a description
of what I'm trying to do. Okay, so chatty here, perfect. This is exactly the
kind of things that can make or break a
fair trade brand. I agree. When you send
me the screenshots, I'll give you specific feedback based on what I actually see. But before that, here is a
solid framework you can use to evaluate their
product photography in a structured way.
And this is gold. It gave me a framework. So first question. Does this look like
a brand I would trust with a wholesale order?
Remember your audience. Most of these factual producers are trying to attract
distributors, convince international buyers, sometimes sell
direct to consumers. That means the photos must
communicate professionalism, consistency, quality,
story, scale. If the photography looks homemade, inconsistent
or random, the brand looks risky and
distributors don't take risks. This is a very
thorough analysis. Point number two, how
many photos per product, minimum acceptable four
photos per product. Idea range four to eight. Less than three photo
looks incomplete. A strong structure
per product could be like it gave me
seven different shots. A clean hero shot,
the main image, alternate angle, a close
up or detailed shot, lifestyle or in context,
scale reference, optional but powerful, a packaging shot or back or
underside when relevant. Manufactured side only
show one or two photos. That's not enough. It gave me some actual examples
of these shots. So the seven shot types
they should be using one, the clean hero shot. I didn't insert any
URL or anything. This is just knowledge
that ChatGPT gives me, like showing me
actual product shots. I haven't put any
fair trade URL yet. So this is the most
important image it's showing me the
clean hero shot. And it's doing this for every single one of these
seven category. So this one is the
scale reference, so I can see how big or small
the product is, et cetera. Just that is already super good. Lighting, the hidden killer. Most fair trade producers
shoot near a window, use mixed indoor
lights, don't control, white balance, et cetera. Background choices. They have three good
options, pure white, light beige or textured, but consistent artisan surface, like wood, linen, clay. Avoid patterned fabric, cluttered rooms, visible
messy environments, mixed surface position
problems to look for story integration,
et cetera. The story I like this point. Unlike Zara, H&M fatrad
brands have artisans, communities, impact
stories, handcraft process. But often that story is not visually connected
to the product. So smart improvement
would be to add one slide per product that shows the artisan making it
or the weaving process, the dying process, the workshop. Okay, so I have my ten or
actually nine criteria for evaluating websites here. Okay? So what I did come back the next slide,
yes, the analysis. I actually ran this analysis
in a separate chat. But you could do it
in the same chat. I think I wanted to
keep it clean so I can share the first
one with you guys. So let me come back
to this slide. If you scan this QR code, you will gain access
immediately to this nine point with the
examples of the shot. So I wanted to keep
this chat clean, just that so you
can use that and continue prompting from
there if you want, but I wanted to give you
a clean version of now, let me go to the
actual conversation. I uploaded three screenshots here from the Cuboi
Technical School website. Here you don't see them. That's because it's a
shared conversation. So here are three
different pages directly taken from the
product stab on Cuboi. Okay, I think I had, the product page for the
scarves, the carpets, and the sweaters.
Upload these visuals. And now, please go ahead, diagnose specific issues, tell me what to keep,
what to reshoot. So this is the end of the prompt that was given
to me by CHATPT before. That's just that.
I copied that and I pasted it again
in my prompt here. And he said, Alright, I went through the
three product pages. Okay, I lists them. I'm
going to be direct. The craftsmanship looks real. The photography does
not support it. Let's break this down properly, and let's have a look
at what we are talking. We're talking about
this page right here. We're talking about
the sweaters. So this page right here, just keep this in mind
what we are talking about. So let's break it down. The hand knitted sweaters, page one, what's working, clean white background,
consistent mannequin style, grid layout, feels organized. The lighting is relatively even. Already better than
many small producers. But what's weak? Everything is front facing. There's no close up, there's no model shots, no
texture, storytelling. All images feel like catalog documentation,
not brand presentation. Again, I'm going
to show this for a couple of seconds so we can relate to what we just said. Sweaters have beautiful
knitted patterns around the colors
and chest area, but you can barely appreciate
them at thumbnail size. So what I would reshoot
100% of the hero shots, no, but I would add depth. So keep one clean front
shot on the mannequin and add maybe a 45
degree angle, backview, detailed shots of
color, patterns, close up of kit texture, lifestyle image
with a real person wearing it, et
cetera, et cetera. So it's giving me this analysis, what we would keep, but
what we should also add. Right. And it will do this for all of the different pages, the three pages, the carpets, the sweaters, and the scarves. Once you know the short list, then the next question
is pretty obvious. How do you produce these product photographies cheap and easy? Well, the answer to that
is in the next lecture, and it's related to Leonardo AI, more in the next lecture.
10. Generating Images with Leonardo AI: Let's create your images. In this lecture, we'll be
generating some images, some product photos with an image generation tool or an aggregator
called leonardo.ai. I'll be using Leo because it's
fast and because it gives me access to all the latest
models like nano Banana Pro. And you'll see how one single image reference of your product plus a solid prompt can produce an entire set
of product shots. Is the original image downloaded from the
*** Bei website. With this, I went to an
image generation tool. The one I used is this one. It's called leonardo.ai. This is part of
the Canva family. Leonardo is an image
generation tool and an AI aggregator. You can generate more
than images here. You can generate images,
you can generate videos. You have different
apps that you can use. I encourage you to explore it. Here. I went to the image
little button right here. And I started prompting. But basically, I
needed prompts for each of these shots that my analysis told
me I would need. Okay? So I asked it, give me six precise nano
banana prompts to help me generate the following
product shots based on a single
image reference. So the image I took from the KTS website,
that one carpet. Required shots per product. Okay? So I need
front view centered. Basically, I pasted all of the information it gave me
of what I need to reshoot. So clean product shot, the hero shot, the alternate
angle, the back view. So the six or seven shots it
told me I should reshoot. And it gave them to me. These are the prompts for the image generation
because I ask it, give me some nano
banana prompts. That means it knows that nano Banana is the image
generation platform. So how can I generate the
based on one reference image, you always need that
reference image, not always, but it's better if you want consistency, and
you need the prompt. So here's the first prompt, the clean front product
shot, the hero image. And here's the prompt. Use
the uploaded reference image as the exact product. Generate a professional
ecommerce hero shot, front view, perfectly
centered, straight alignment, entire product fully visible, no cropping, neutral,
white background, soft diffuse studio light, no harsh shadows,
no reflections, no. High resolution, ultra sharp, true to original colors,
maintain exact shape, proportion and details
from the reference image, commercial catalog
style export ready. I didn't invent this prompt, the model did for me. Took this and I came back to Leonardo and I pasted
this prompt right here. That's the exact
prompt I pasted. This is the first
image that came back. We can see the same carpet. You could analyze
them side by side. I did that yesterday.
It's the same carpet. I gave me this
shot, which is what I asked for the hero shot, the center, the one
with the entire. The first additional photo. Then I ask it for prompts for all of the other
shots that I needed. Okay? The close
up, the back view, the 45 degree angle, and I'm going to
show them to you. They are all right here. So let's get back to
the presentation here. I have added all of the shots
here for you to appreciate. That is the hero shot
with the prompt. That is the 45
degree angle shot. This is the back shot. This one is the one that
surprised me the most. So if you took a photo from
the back of the carpet, I'm not sure it looks like this. So this is where you need to
be careful about realism. Ethical consideration. If it doesn't look like that, then don't use that
photo, obviously. The detailed close up,
the scale reference, I ask you to generate the
shot with the carpet held by a Nepalese lady artisan
and the lifestyle photo. And this is where
we need to talk about ethics because
the tours are powerful enough that it's kind of easy to cross the line without
you even noticing. So in the next lecture, we'll talk about
ethics and gardwails.
11. Ethical Considerations: Before you publish any AI generated
images, you need rules. If your product photos
generated with AI end up looking better than the reality or different from the reality, you are setting yourself
up for complaints, returns, trust, or
even legal issues. I don't know what you
think about this, but this blew my mind
because you don't need a photography
studio need anything. You just really need
one photo and to have strong enough ethical
considerations to decide whether or not you
are blurring the lines here, whether or not it's ethical to present such a photo to a buyer, does it correspond
to the reality, or is it too far away from? Okay. So this is why I
want to finish with this. Like, you should if
it's not close enough, you can try to refine, by adding these kind of
lines to your prompt. Keep product identical
to the reference image. Do not alter the design, pattern, shape, or color,
maintain full accuracy. Because the models are
becoming so powerful nowadays that they are able to
reproduce identical objects, even person this
is where we are. So let's approach
this technology in an ethical manner. All right. Now, you have all the
workflows and we have talked about the ethical consideration and
the guardrails. It is now time to
wrap it all up with four habits that will make
it all stick together.
12. 4 Tips to Make AI work for You: Before we go, let's recap the most important
takeaways from this course, the four habits
that will make AI useful and practical for
your retail business. This is my last slide. Four
tips to make AI work for you. Let the AI create your prompt. Don't play the prompt engineer. This is not 2022 anymore.
We have this now. So let's do the work. Give plenty of context with voice mode. Use your microphone. Brain dump everything you
need, it will structure it. Break complicated
tasks into steps, make it easy for the model to just do one thing and then
move on to the next thing. Don't ask it for the final
outcome immediately, and then save or
organize your chats, your best prompts, what
has been working for you. Save that into projects
inside the LLMs, ChatGPT lets you
organize by projects. The other platforms also
have similar kinds of features or simply copy and
paste them into a slide deck. I don't know, and
share them within your organizations so
that everybody can use. And you now have the full loop, daily workflows, research, prom building, website review, product photography upgrades,
and ethical guardrails. It is now your turn to play. Just pick the thing
that resonates with you the most
from this course and start applying it to your
business today or tomorrow. My tip is do not try to do
everything all at once. Just pick one thing, apply it, and then come back and
try something else. And you'll see if you
work like this with AI, small wins will start to
compound and stack up fast. Thank you for learning with me. I wish you all the best
with your business, and I'll see you
in the next class.