Transcription
GPT5 is probably the biggest AI model announcement from OpenAI since it was first launched. And now ChatGPT is like a super AI assistant. It has so many capabilities, but how you can make sense of these features. So in this video, I'm sharing five real-world ChatGPT workflows and use cases that show you exactly how to leverage ChatGPT for work, and that can save you hours every week.
Perhaps the most obvious highlight from ChatGPT5 is the consolidation of different models. There are now only two core models: the basic core GPT5 model and a thinking model. It used a unified reasoning system with an internal router. So ChatGPT triggers thinking mode based on your task complexity. This is what is called automatic thinking, but it will use less reasoning effort, so you can also manually select the thinking mode or use the Think Longer option, though this will count towards your thinking usage limit. So for complex strategy work, I recommend explicitly choosing the thinking mode to get the deepest analysis possible.
Another highlight I personally observe is the writing capability, although I know there are diverse opinions in the ChatGPT5 writing. From my own experience, the writing from ChatGPT5 is more natural compared to GPT4. It is using less wording that is obviously generated by AI, less explanatory and what I call the fluffy words. It's more close to how we communicate in real situations, and also the response speed is almost much faster than the previous model. When thinking mode is not turned on, the response is almost immediately being outputted.
Now let's talk about the real use cases and start with what everyone loves: Deep Research. But instead of just running a standalone deep research, this research workflow will leverage the project features. So ChatGPT5 can retrieve all your important business context and then transform it into ready-to-share deliverables using its canvas mode. For example, we are the business strategic research team at Airbnb and we are doing business strategy brainstorming. And here is our annual report that covers the core financial performance, strategic pillars, outlook. And we can use it as a strategy document.
So first, set up a ChatGPT project, as it will allow all of your conversations to have the same business context, and then upload this annual report and set up the custom instructions that define role and the task details. Now we are ready to run the deep research. So use this prompt to research the topic about the experience economy, which is a hot trend right now, and identify the top opportunities and threats specifically for the Airbnb business. So ChatGPT will ask you some questions, answer all of them, and start the research process. So after 11 minutes, we have this really comprehensive report covering the three opportunities, like enhanced user engagement and brand loyalty. The three threats, including the regulations and financial implication, and also the comparative positioning against other competitors like booking.com, Expedia with a table summary, which is great, and also the market opportunities.
Now we are going to turn this report into a comprehensive dashboard for team sharing. So turn on canvas mode, which allows you to create interactive documents or coding, and use this prompt to create an interactive dashboard to visualize the research findings, including the specified areas. So it will automatically trigger the thinking modes for better answers. And I would say the coding speed is really fast. And so here is the preview. We can first click share to create a link for better display. This is definitely not bad, although I see some layout issues that some text is now underneath the callout boxes and also some header text should be better in white text. And the dashboard is highly interactive and you can pick different regions to highlight them in a chart and with the strategic recommendations and executive notes. So you can now use the same method to quickly build a deck, dashboard, and shareable insights for teams using deep research all within the same ChatGPT platform.
Now, as a bonus, let's also test the ability of the highest-end model, ChatGPT 5 Pro, for a reasoning task. So I will upload back the deep research report and then use this prompt to ask it to create three strategic scenarios for the next 18 months: the best, realistic, and worst cases. So now it's triggering the reasoning process. And we can also click the details to see what are the reasoning tasks it is doing. And finally, it has taken around 30 minutes to complete this task. So the output is very detailed with the executive comparative scenario summary with the risk indicator, and then it will break down the details by each scenario, like the best use case with the key trigger events, recommended resource allocation. Also the priority action plan and timeline. So how to approach if we need to capture this trend. So for this same prompt, I have also used the ChatGPT thinking. So the output quality from ChatGPT thinking and ChatGPT Pro is quite similar. So unless you are doing really hard, niche research study or topic, I don't think you need to use ChatGPT Pro for now.
And not just for researching. As I mentioned, the content writing in GPT5 has significantly improved. It's more natural. So together with the agent mode and the thinking mode, we can build a powerful content creation workflow that helps you make your content creation process much easier, and most importantly, more scalable. For example, I want to streamline the newsletter creation process and I love Justin Welsh's newsletter structure. It's punchy, good use of word choice. So first, on ChatGPT, turn on Agent Mode. First, ask it to script 10 of his recent newsletters for our later analysis. Of course, to fully maximize the agent mode, you should definitely consider scraping more than that. So the agent will start checking all the recent newsletters from Justin's site and grab the most recent 10 newsletter content. Note, sometimes some websites may block the agent access, so in that case, you might need to use APIs, scraping tools, or manual methods as the alternatives.
Now the agent has returned a Word document, but it has included citations which impact the formatting. So my tip for you is just ask ChatGPT to remove all the in-text citations, and then the final Word document will be much better with all the 10 articles grabbed with formatting. Now upload this Word document and ask ChatGPT to analyze the tone of voice of Justin. And output it as the newsletter content template. So this time, let's try the Think Longer option. So this option is useful whenever you just want a particular prompt to be run with more reasoning. So it has output a PDF document. It's quite impressive with the key patterns, writing style, and voice identified from Justin's newsletters, the structure and rhythm he used. And I really like this reusable newsletter template section with the successful structure that ChatGPT has analyzed from the samples. So you can use it as a starting point and tweak it further. And also with the opening formulas, transition phrase library. So this is perfect and I've also tried the same prompt without turning on the thinking, and the output is definitely less detailed and less precise. It only has got two pages versus the four pages when I turn on the thinking. So I would definitely recommend using Thinking whenever you are doing any in-depth analysis.
So here we have a ChatGPT project set up for newsletter writing with the custom instructions. I also set a command newsletter, so whenever I type this command, this workflow will be triggered. And also make sure to upload the newsletter template we just generated. Then turn on thinking mode. Let's say I have this YouTube script about career growth and then upload to the chat and type the command newsletter, so it will now trigger the workflow. You can see it's following the template with the hook, story, and reframe. So this is a great starting point without you thinking how to start from scratch, but always fine-tune it to make sure it sounds like in your voice and do not just directly use it. And now we can even ask it to generate a banner image for this newsletter content. And I really like it. I have intentionally asked ChatGPT to make it more authentic. It's just my personal preference, so I love it. Now you have an initial newsletter draft ready with the copy and images all done by ChatGPT, and you can even use the same approach to build a custom GPT as well for scalability. So check out this video for more.
*****
Another area ChatGPT is a great help is data analysis. AI is really good at data processing and finding patterns, but instead of just generating insights, it can be also used to enrich data as well. So the coming workflow, ChatGPT will transform the raw data into useful data sets, get more meaningful insights, and identify quick opportunities.
**********
For example, I'm doing research on a topic about AI marketing, and I have exported this list of commonly asked questions by real users about the topic from Ahrefs. So first, upload this list of questions to ChatGPT. A tip here is to turn on thinking mode whenever you are doing data analysis like this, because I found that if you don't turn it on, there might be a high chance for hallucination. Now we can ask it to analyze each question and assign them into the corresponding search intent: awareness, commercial, or transactional. So this will be very helpful for us to understand what content we should create in order to capture these search intents. So quickly, ChatGPT will help to generate the CSV file with the enriched datasets by adding the search intent column, and this will save you a lot of time. Imagine you are dealing with large datasets.
So with these enhanced data, we can now ask ChatGPT to generate a visualization to summarize the data, to show the question type, pinpoint content opportunities, and make sure you turn on canvas mode whenever you want to build some visualization dashboard. So ChatGPT, build it really quick, summarizing the questions, insights, intent type, and there is only awareness and consideration, which is correct. And what are the most common pain points and what audience are looking for related to AI marketing like informational? It's more about learning how to implement. Users want more step-by-step guides, tool directories, basics. While for commercial type questions, it's more about roundups and comparison, specific use cases. So using ChatGPT, you can quickly turn any enhanced dataset into shareable insights using its canvas option. Then we can even ask it to propose two to three topic clusters with related questions that we can create content about. And it also gives me three cluster suggestions, and I would say they all make sense to me, like AI Marketing tools, AI Marketing agencies, and the related questions from the dataset. So this way you are maximizing value from this dataset, not just getting insights, but understanding what actions to take.
But we can even take a step further to connect ChatGPT to more than one data source because in the real world, we often deal with multiple data sources. So this coming ChatGPT workflow will leverage its connectors for external data, doing analysis, and to automate the whole process using the task scheduling capability. For example, we have the raw traffic data already uploaded to Google Drive, and I need to prepare a monthly traffic report. But instead of just doing a simple automation on the monthly report, we want to maximize the power of ChatGPT to synthesize from different data sources and give us more meaningful insights. So we will also ask it to access the campaign calendar on Notion to do the data synthesis and correlation analysis.
So first, select the source from the connector apps. Make sure you have already used the building connectors to connect your ChatGPT to your Google Drive and Notion account or any data source you need for the task. If not, you have to click "Connect More" to do the connection first. Then we can ask ChatGPT to retrieve the data from both sources and find a correlation from these data and to create a single briefing Word document that summarizes the findings and an email summary for the internal team. So you can see it is now reading both Notion and Google Drive sources, mapping the campaign timeline with the traffic data, and after just two minutes, it outputs some high-level insights, questions, and also outputs the two deliverables. And I think this briefing document is nicely done with the key takeaways, the chart generated with comparison using the traffic data, and it also includes the observed correlations of how the campaigns might impact traffic data and the recommendations. So this is one of the best ways of using ChatGPT, because instead of just asking it to summarize data to insights, also give it different data sources to find connections and to inspire your own thinking. So this is great. And here is the email in a text format, and we can also just ask it to just generate the email summary directly in the chat so we can just copy that.
So assume we have tested the workflow and we are satisfied. So we can ask it to create a scheduled task to run the whole workflow on a monthly basis, and we can click edit to refine the instructions or workflow details, like where it should get the data from. So in this case, it's Google Drive folder and the Notion calendar, and we can further adjust the frequency of this scheduled task as well. So this workflow would be super useful to run any repetitive tasks and use together with the ChatGPT connectors for automatic data retrieval.
Another thing that ChatGPT has emphasized is the coding ability. Anyone can be a builder nowadays, although I'm not going to ask it to build a full-stack app. I think the most useful ways you can use it immediately is to have it research, grab insights, and quickly turn them into functional prototypes so you can communicate your ideas much faster with your stakeholders or clients. For example, I'm revamping the Loom website messaging. And so I first start with the ICP research, and I'm going to use their customer case studies and also some industry reports for the research process.
So now on ChatGPT, upload these research reports and this time, instead of just the deep research, let's try the normal web search. So turn on web search mode, as we want to make sure it will search from the open web, and then use this prompt to ask it to search for the top five video messaging competitors to Loom and identify their target customers, pricing model, and key messaging angle. Again, it's triggering the reasoning mode, and then it pulls out the top five Loom competitors with sources like Vidyard, Vimeo, Hippo Video with the primary customers, pricing, and key messaging with quotes, which is great, and you can always click on the source to verify that, and from what I see, it is accurate. So this will give you an immediate overview on the messaging and positioning against the competitors and where we should emphasize.
Now use this prompt to further analyze the customer reviews and testimonials for these five messaging competitors from review sources like G2, Trustpilot. And then it will give you the breakdown of the frustrations and the desired outcomes from the real users, which can be useful to build up the messaging for Loom with the source links. So besides an in-depth search using the deep research or the agent mode, ChatGPT search also gives you pretty solid research findings and a nice overview.
Now switch to canvas mode. So from these ICP research findings, we can now ask ChatGPT to create a high-converting landing page with a punchy and memorable headline and use messaging that resonates with your target audience. And also embed a functional ROI calculator tool to keep the audience engaged. So after some small tweaks, here is the landing page, and we can preview it with the link. So it's using Loom's brand colors as requested. So what I like is the headline, so it speaks directly to the ICP, and this is why I think the GPT5 has much better writing than before. It's more creative and sounds less robotic. And I like this interactive tool. It is fully functional and makes the page much more engaging. As for the layout capability, it is pretty standard, so the most obvious upgrade is actually the speed. You can now build anything like a prototype web app, interactive tool from research really quickly using the ChatGPT canvas, and you can also download the source code and then run it on your domain by uploading it to your server hosting. So what are you launching new features, entering new markets, or responding to competitors? You can instantly turn customer research into optimized pages or prototypes that actually speak your audience's language.
Now, ChatGPT is powerful, but like every model, it has its own limitations. First, ChatGPT is still hallucinating more than other models, especially compared with Claude, which I use every day. This is especially obvious when you run a large dataset analysis where the web app doesn't support large context handling. Since for now, you only enjoy the 400K context window using the API option. So make sure to turn on thinking mode whenever you do a task that requires high accuracy.
The second limitation is building ability. Yes, it has improved significantly with the UI and the speed, but perhaps because of the limitations of the context window size using the web app, when the chat is getting long, it has limitations in maintaining the code quality and consistency. So you have to carefully craft your prompt in the first attempt.
Join my community if you want to learn how to be more thoughtful of using AI in marketing or online business. You will get access to all the actual prompts I use in all my channel videos, our regular live sessions, and other extra content. You can find the link in the description to join, and before you go, also watch this video about ChatGPT agents to learn some other real use cases for your inspiration. I will see you next time.