Yesterday, I’ve done an AI Workshop for young PR professionals as part of an event.

In this mail, I’d like to share with you what I learned first hand about our market, about how people use ChatGPT, and what areas to focus on when educating people.

I’m writing this to you with the goal of passing down my learnings so you can train people better.

Because, chances are you’re or will be responsible to train people on how to use AI.

If you are a business owner or manager, you’ll need to train your team.

If you are an AI consultant, you need to train your clients and guide them in the right direction.

And if you deliver AI automations to clients, you also need to train them how to use the thing you made.

So here we go:

Adoption at work is crazy high

The attendees on the workshop were between 20-30 years old. So GenZ and late Millenials.

I began the workshop by asking, show of hands if you used ChatGPT ever, last month for work, last week for work, daily. The answers were 95%, 80%, 70%, 50%, roughly speaking, out of 25-30 people.

Now, they were already somewhat pre-selected because you wouldn’t attend an AI workshop if you’re not using it actively, but this was the highest adoption number I’ve ever seen on any workshop. Half a year ago, these numbers were all halved.

Prompting and conversing is intuitive.

We began the workshop with some questions about AI, ChatGPT and they knew the difference between the two, and had an understanding of how it works.

I taught them the 6 key elements of a prompt by doing group exercises and doing tasks of their choosing, by writing each element at a time.

Some groups already started engaging in longer conversations with ChatGPT, sending follow up prompts, and went ahead of the workshop.

They also got the hang of things really fast, and asked great questions.

Tokens and Context Window are still foreign.

When we got into the part of talking about Tokens and Context Window, they didn’t know what that was, so unlike the AI and ChatGPT questions that they could answer, here I had to chime in and simply explain what these are.

So this hasn’t really changed since last year. This is still a concept people don’t quite understand generally.

I’m assuming it’s probably because ChatGPT doesn’t give errors related to Tokens and context windows, and if your prompt is too long, you might think it’s because of a character limit and not a token limit issue.

And it’s not even really necessary information, especially now with GPT-4o, 128k context window is more than enough for 99% of work cases. And as most people will not touch the OpenAI API, they don’t necessarily need to know and understand these concepts.

We covered them because in the second part we got into building automations, so it was foundational there.

If you are only teaching ChatGPT and CustomGPTs, you can probably leave these out.

The 6 key elements of a prompt are still life-changing.

Especially the first two, the persona and the task definition. Most people said to me that this was the most useful part of the workshop.

Now they know what kind of tools they have in their toolbox and when to use them.

I noticed that they were already using these elements sometimes, but not consciously.

One student told ChatGPT in a follow up prompt “make the sentences more spicy.” which is basically the “what is considered good” element in a costume. To make it more obvious, what this prompt means is that “A good response to me is more spicy”.

This again confirms that prompt engineering is nothing more than effective communication.

Identifying steps of tasks is difficult.

Most attendees were struggling a bit with clearly articulating how they do what they do, and this is a fundamental roadblock in them automating their own responses.

I have a highly analytical brain, so thinking about processes is easy for me.

For example, on the workshop we had to go down to the level of like what are the steps you take to write an email.

1. Open browser or mail client
2. Add address of recipient(s)
3. Write subject line
4. Write email body based on certain information
5. Write signature
6. Send email.

And in order to automate your work with AI, you need to be able to think in such a granular way.

This is made harder by most user interfaces getting simpler or even entirely disappearing.

I think the Baby Boomer and Generation X has a lot better systematic thinking, because they grew up in a world where you had to tell the computer almost every step when using it.

The younger generations don’t even know how a mobile phone or how the internet works. They just know how to use it (usually better or faster than the older generations.)

I think the successful Prompt Masters or Automation Engineers will be successful because of this ability. To clearly articulate how a certain process is done, step-by-step.

No-code tools are still unfamiliar

Unlike how familiar attendees were with the ChatGPT interface, they got very intimidated with Make’s no-code interface.

It’s understandable, as ChatGPT is the same chat interface they you use to talk to your friends and family, and its familiar, whereas the no-code platform is totally different.

They haven’t heard of Make or Zapier before, so the adoption curve of no-code tools is a lot lot slower than the adoption of ChatGPT.

This makes total sense, as only a small percentage of ChatGPT users will think to automate their work with no-code tools, but I wanted to emphasize this as you might use Make or Zapier as often as ChatGPT, and it’s easy to forget that people don’t know what you know.

CustomGPTs > Automations

I haven’t talked about CustomGPTs on the workshop, but if I were to redo it, I would have switched the second part of building an automation to instead configuring Custom GPTs.

CustomGPTs are still the next logical step from ChatGPT Prompt Engineering, and I think the attendees could’ve taken more home if I spent the second part of the session on the CustomGPT world.

I don’t have regrets, as this learning was only possible to learn through this obstacle.

Maybe our workshop offerings need to be the following:

1. AI prompting workshop (for beginners, first part intro and ChatGPT prompting, second part configuring Custom GPTs)
2. AI automation building workshop (for those already familiar with CustomGPTs, first part workflow theory and documentation creation, second part building automations in Make)

Reply to this email if you want me to do a workshop for your team or event, and let’s talk!

Radical time savings for knowledge work

One of the best moments, and I left it to be last on purpose was when we went through the first few prompts they send and they were picking their jaws off the floor.

The exercise was only 3 minutes long, so they had to write a persona and task and then send the prompt to see what happens.

All tables got a good enough result in those 3 minutes. When I asked them how long would it have taken you to get to this level of completion on your own, they said between 30-120 minutes.

So right there on the workshop, they experienced what it really feels like to “think faster”.

They’ve written a personalized pitch to a specific journalist in 3 minutes instead of 30 minutes.

They’ve written a set of “spicy” questions in Spanish and English for a provocative interview about a recent event in 3 minutes instead of two hours (if we calculate the research of the latest event as well).

They’ve written a press release for a cupcake store in 3 minutes instead of 30 minutes.

It gave me goosebumps when they were saying these numbers, goosebumps so strong that you could even see on my bald head. 😂

This is what using AI is about. These time savings on tasks, doing something that would’ve taken you 30-60 minutes, and doing it instead in 5 minutes. Writing a prompt: 1-2 minutes. Waiting for GPT response: almost instant. Correcting a few minor things: 2-3 minutes.

Add this all up for everything you do, and you can go home an hour earlier, or you can go to your son’s baseball game.

Prompt Masters can do this. If you want to become a Prompt Master, click here and check out our Prompt Master AI course with bonus consultation session.

I hope you enjoyed reading what I learned by teaching. The age old saying is still true: “When one teaches, two learn.”

Talk soon! All the best,

Dave Talas