Intro
As we continue to delve into the world of generative AI at work, I’ve been chatting with various people about how to integrate GPT-style tools into their day-to-day lives. There are all sorts of interesting ways to tease out use cases, test and measure their impact in the workplace (my day rates are reasonable, corporate clients…), but on a basic level, it’s struck me that tons of people think of them as glorified search engines, and they often don’t have clear use cases in mind.
It’s sometimes difficult to advise people, both friends and in a corporate environment. Colleagues have chosen a particular career path, and socially, everyone’s into different things. I don’t necessarily understand what they do day-to-day in any real depth. I’ve worked hard to get into the details of specific areas of construction (ask me how to do an engineering grade survey or what the issues were on our big problem projects, and I’ll send you to sleep with details… ), but nobody can advise everyone all the time.
However… this struck me as a good use case for ChatGPT. Using generative AI to help people use generative AI is my current favourite thing (the self-referential irony makes me smile all the time, every time).
On the basis most people are using ChatGPT, I pulled together a prompt that:
- Asks ChatGPT to act as an expert in ChatGPT training.
- Triggers it to ask for your job description.
- Generates some potential high-value use cases based on what you do.
- Gives the user some tips on how to delve into those use cases, and talk to the gen AI tool in an effective way.
A new ChatGPT Assistant is Born
I turned it into my first custom ChatGPT bot, which makes it dead easy to use.
If you’re a ChatGPT Plus user, you can access it here.
In keeping with my newfound love of giving these things slightly pretentious names relating to ancient Greece, I called it
“PragmaPilot – A Work Focused AI Coach, using ‘pragma’ (deed, affair, matter of business) and ‘pilot’, suggesting a navigational tool for practical affairs.”
And yes, I used ChatGPT to help me name a ChatGPT bot. Of course I did.
And for the record, that line in quotes above is the only AI-generated text in this post.
The Useful Bit
If you’re not a ChatGPT Plus user (and you really should be – it’s cheap and high value), you can copy and paste the prompt below into the free chat window, and you’ll still have a high-value experience.
Other AI tools are available – swap out ‘ChatGPT’ with ‘Google Bard,’ ‘Perplexity AI’ or whatever you’re using, and the outputs should still be useful. You’re mileage may vary, depending on the model you use. A brief note for the gen AI model nerds: I’ve tried it with various versions of Llama v2, with some interesting results.
That sent me down a rabbit hole around corporate (and general) L&D use cases, the corporate value of these tools, business cases and all sorts that I’ll gloss over here.
The prompt, ready for you to copy and paste:
I want you to act as an expert in coaching people how to use ChatGPT.
You’ll ask me to share my job description in the next prompt, then, based on that information, I’d like you to create a list of versatile use cases for ChatGPT that are specifically applicable to my role.
This list will be organised into distinct sections, each highlighting how ChatGPT can be utilised in my professional context. The sections will include practical examples and direct requests you can make to ChatGPT to enhance your work efficiency and productivity.
Pick one of the examples that’s very specific to my role, and coach me on how to iterate through one of the use cases, highlighting to users how they can ask more questions to improve the quality of ChatGPT’s outputs.
Hopefully this clearly illustrates a high value point… you can ask these tools to help you use them. This is the first time we’ve ever being able to meaningfully talk to the technology, and we should leverage that capability everywhere we can.
In a corporate environment, this means that first-line support should be the tool itself.
Is it any good?
Spoilers: it did an alarmingly good job.
It’s not perfect (a bit like that image ^ – you got so, so close, DALLE.3), but it really highlighted for me these things *will* be able to do parts of my job better than me, further confirming that leaning into talking about technology is probably still a decent career move…
To explain my personal experience with the prompt / GPT assistant… it suggested various things I already do, which was both great and gently alarming at the same time. While iterating through the prompt and testing it, it suggested a project idea eerily similar to one I’m actively working on. I’m not going to share any details here, although if you work in Balfours, there’s a Yammer post floating around that includes the full transcript of one of the chats.
The process demonstrated loud and clear that these tools are very capable in their base state (without any significant data or extra training). In technical terms, we’re using a ‘foundation model’ and it gave me some hope that no matter what you do, using this approach will guide you on where to start with these tools.
In the interests of full disclosure… it also highlighted the benefits of a particular AI project type that I’d never thought of before. Despite this gently offending my ego (I’m an expert!?), I’m sharing this point to highlight the value of these tools.
No matter how good you are (or how good you think you are!), if used in the right way, ChatGPT-style tech will trigger thoughts and give insights that have never crossed your mind before. They can (and will) send you down high-value paths you would never have explored otherwise.
Be careful what you share
A word of caution: I’d strongly argue against sharing sensitive or proprietary business/personal information with public tools. “No top-secret stuff for our public AI friend” is a basic rule for me, and I tend to use a rule of thumb…
If you wouldn’t say it to a friend, or talk about it in a public space (coffee shop, pub, wherever), then don’t say it to publicly available AI tools.
There’s a risk your information get’s incorporated directly into future version of the model. There’s an argument to be made that we all benefit if these tools get better, but equally, there are many risks that go with that that I’ll write about another time.
Given there’s a reasonable chance your CV is public on LinkedIn and you probably share that stuff freely anyway, I don’t see any real risk of harm here.
The value of experimentation
If you’re curious about generative AI tools and haven’t used them like this before, I highly recommend giving the bot or prompt a go. Describe your role, or outright upload your job description, and see what it comes up with.
I’m hopeful this article will encourage people to experiment with these things more, as I’m convinced that they’re crazy powerful in a corporate and personal productivity context. I worry that lots of organisations are thinking about them in traditional “data” terms (that’s probalby a whole post in itself) and getting distracted by the unhelpful noise in the media.
I’m (no exaggeration) doing the work of 2, maybe 2.5 people at the moment, and that wouldn’t be possible without some of the custom AI’s I’ve built, and using techniques like those above. Imagine what the world will be like when everyone has these capabilities, and it’s common as having a TV, a home internet connection or car?
The thing that blows my mind is that the whole exercise—from conceiving the idea to having a working prompt—took me about 15 minutes. It took 10 more minutes to create the custom ChatGPT bot, and 5 of that was doing some identity verification so I could share it with the public. I’m amazed at how well it performs.
Keep experimenting!
Despite using these tools every day for a year, they still endlessly impress and surprise me and it feels like we’re only just starting to explore what they can do for the wider world.
Because I refuse to use AI to write these articles, the bulk of my evening was spent typing these words into Evernote like a 1950’s ape with a typewriter. 🤷♂️ 🤣 Make of that what you will…
Hopefully, this has been useful. I’m going to stop bashing keys now.
Take care, and keep experimenting!
David