Mind Readings: The TIPS Framework for AI Task Identification

In today’s episode, you’ll learn a practical framework called TIPS for identifying tasks best suited for generative AI. You’ll explore the four key dimensions of Time, Importance, Pleasantness, and Sufficient Data to evaluate tasks and determine their suitability for AI implementation. Additionally, you’ll gain insights on leveraging generative AI to streamline workflows, improve efficiency, and enhance overall productivity.

Mind Readings: The TIPS Framework for AI Task Identification

Watch this video on YouTube.

Can’t see anything? Watch it on YouTube here.

Listen to the audio here:

Download the MP3 audio here.

Machine-Generated Transcript

What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for watching the video.

Christopher Penn: How do you know which tasks are best suited for using generative AI? Of all the things you do in your job, which things are the best candidates for generative AI? No surprise, we’ve got an acronym for it called TIPS.

It stands for Time, Importance, Pleasantness, and Sufficient Data.

So let’s talk about what all these mean and how you would use this.
Generative AI is all about saving time, saving money, making money by making things faster and better.

So the four dimensions you want to look at any task in your office with are:
Time: How much time does a task take? If a task takes you two seconds, it’s probably not a good candidate for generative AI.

It’ll probably take you longer to run the prompt than it does to do the task right now.

And so that’s not a really great candidate.

If a task takes you ten hours, yeah, that’s a good candidate.

If you can chop that down, even just in half with generative AI, that’s a really good candidate for use of it.
Importance: How important is the task? The more important the task is, I would argue the less you should consider it for generative AI because it’s going to require more human review.

For example, should you replace, say, a radiologist with generative AI? Replace? No.

Second opinion? Yes.

But replace? Absolutely not.

Should you replace your lawyer with generative AI? Absolutely not.

Should generative AI maybe help do a first draft? Yes.

But think about all the other tasks that you have, some of which you know aren’t important.

“Oh, I got to do expense reports.

I got to write this memo.

I got to do this summary.

I do this.” You know which tasks you’re like, “Does anyone really care?” The less important a task is, the better a candidate it is for generative AI.

Just hand that stuff off.

Get rid of it.
Pleasantness: How pleasant is the task? How enjoyable is the task? The less pleasant it is, the better a candidate it’ll be for generative AI.

This one’s really important for executive buy-in.

If you can show someone who is maybe skeptical about the benefits of AI, show them, “Hey, here’s this.

I know you hate this task.

I know you hate doing it every month,” whether it’s writing your speech for your all-hands staff meeting, whether it’s writing sales pitches, whatever the thing is.

You hate it.

And you can show someone how to replace it with generative AI.

They’re going to be like, “I can stop doing that.

Great.

Where do I sign up? Here’s the credit card.

Go get me some AI.” So the less pleasant a task is, the better a candidate it is for generative AI.
Sufficient Data: How much data do you have to do a task? The more data you have, the better the candidate is for that task to be done by generative AI.

So if you, for example, were trying to do an expense report, and you don’t have a template for that for your company, generative AI might or might not do a good job.

It might or might not be sufficient.

But if you have a template, and the last five expense reports you submitted, you’ve got sufficient data to hand to a model and say, “Hey, here’s this month’s receipts, go and do the thing and make it look like these other examples.” The more examples you have of any task, the better a candidate is for generative AI because you’re going to be able to hand that to a machine and say, “Here you go, do the thing, make it look like this.” This is one of the reasons why I use generative AI so much for things like clients—we tell them this is generated by AI, and it’s even a little label on it.

So we’re not hiding anything.

But things that the customer expects in a specific form anyway, so let’s let the machine do it.

Let’s let the machine not miss details.

So one of the things for example I do is with a secured model, I will take the contents of an inbox, or client over the last month, and dump it all in AI and say, “Hey, summarize the month, what did we do?” Because I forget, I forget all the emails going back and forth and me—and no, it’s also—I forget, I don’t have to forget.

I can have AI read all that and say, “Here’s what you did this month.

And here’s what’s still outstanding, or here’s what’s”—if you have sufficient data, great tool.
So that’s the framework that I would recommend.

How do you put it into play? Make a list of tasks and put TIPS at the top, hit the columns, and put scores for each of these things.

You know, how much time does it take a task? Take a scale of one to five, or five is the most time.

Is a task, the least important tasks should get a five, right? Because we ultimately want to add all these up.

How pleasant is the task? The least pleasant tasks should get a five.

And then how do you have sufficient data? The most data should get a five.

And, you know, the least data has a one.

And let’s just add these all up.

If something gets even a ten, you might want to use a—15 should definitely be considered an AI.

If it’s a 20, hand that off to AI as soon as you possibly can get it out of here.
That’s the very straightforward framework for identifying which tasks, no matter what department you’re in—HR, finance, marketing, you name it—do this, use the TIPS framework to decide what tasks to hand off to AI and then work with a competent engineer to build a solution for that.
So that’s gonna do it for today’s episode.

Thanks for tuning in.

We’ll talk to you next time.
If you enjoyed this video, please hit the like button.

Subscribe to my channel if you haven’t already.

And if you want to know when new videos are available, hit the bell button to be notified as soon as new content is live.

Mind Readings: The TIPS Framework for AI Task Identification

Machine-Generated Transcript

Comments

Leave a Reply Cancel reply

Pin It on Pinterest