6 Great AI Ideas

It was such a pleasure to present ’35 Great AI Ideas in 35 Minutes’ with Tamara Gielen, Materialise, at Email Innovations World 2024!

For those of you who missed our presentation, I thought I would present 6 of my ideas here, and another 6 on my Email Optimization Shop blog; I’ll write-up the final 6 next week for my blog. I’m going to use our actual slides, for visual appeal and to keep the copy minimized.

Let’s jump in!

1. LLMs are less like magic… and more like SmartCompose

It’s important to understand how LLMs work if you’re going to use them.

At a conference last month one of the speakers said that ChatGPT had ingested the entire internet – and that, as a result, it was very smart.

Yes, LLMs have ingested a lot of content. No, they aren’t very smart.

LLMs gather information and they are able to pull from that content, in context, to provide output. But it’s more like SmartCompose than it is actual smarts – it’s looking at what the next most likely word over and over again. You’ll see some examples in these blog posts that illustrate that well.

LLMs are a valuable tool – but they aren’t magic, they’re more like SmartCompose.

2. Your GPT input may not be private

I pay for a ChatGPT workspace account; one of the benefits is that your inputs are private. You don’t necessarily get this benefit from a free or even the lowest level paid account.

But even though privacy is one of the benefits I pay for, I don’t assume that the content I input into ChatGPT will be private. Don’t upload proprietary information; even if you don’t consider it proprietary, don’t upload anything you wouldn’t want to be in the public domain (or if you must, take all company/brand identifying information off of it).

3. Check for errors

By now you’ve certainly heard about hallucinations. They are real. Which is why you need to check all the output you get from an LLM.

The screenshot here is from a sample campaign I built for a conference presentation (Arthur Bryant’s Barbeque is not a client of mine). The inputs had included URLs for the Arthur Bryant Barbeque site as well as a page on the Goldbelly site where an Arthur Bryant’s Barbeque Sampler for 8 was offered for home delivery.

The sampler does not include brisket (it includes other barbequed meat products). But twice during the course of creating the plan and then the campaign ChatGPT included ‘brisket’ in the offering. Both times I asked it to check, and it corrected its output. But I had to ask. So don’t assume it knows – you always have to check.

Side note: remember when we talked about LLMs being like SmartCompose? I imagine this error happens because many if not most barbeque offerings do include brisket. And it’s so intent on providing the next most likely word that it doesn’t go back to confirm the actual offering.  

Looking to up your AI game for email marketing? Join us for a 6-hour hands-on workshop training -- you'll create 3 of your own custom GPTs and leave with a good start on your next email marketing campaign! Learn More

 

4. Use the iron triangle to guide your foray into AI

Not a new idea, but a new way to leverage it.

Time: Incorporate AI into your work

Find a way to use AI every day. I say use it for work, but really, just use it. When I started, I used AI for research; I also used it to help plan my itinerary for a few days in Copenhagen after a planned speaking engagement Denmark.

Cost: Spring for the paid version of your AI of choice

The benefits of the lowest level of paid ChatGPT access are well worth $20 per month. They include the ability to use custom GPTs that are public, as well as create your own custom GPTS for personal use. I believe that access to this technology will only become more expensive – why not learn now, even if your company won’t cover the cost.

Scope: Start by perfecting a single task

AI is like any other skill – you want to start with the basics and build. Don’t try to use it for everything all at once. As I mentioned, I began using it for research for client work. Then I gradually moved into collaborating with it on the rest of the campaign tasks. Work until you can do one thing really well with AI – then choose a second task and perfect that.

Quality: This is the only way to get to a quality result

We’ve all seen those blog posts where you create an email with one prompt and no previous experience. But for really quality work, I recommend you follow these steps. It’s a little more effort for a much better result.

This tip is actually based on a post I wrote in March 2024 for the Only Influencer Blog called ‘The Cost-Scope-Time-Quality Triangle of Generative AI.’

5. Apply Your expertise

Don’t assume that AI knows everything that you do – or that it knows more. It does not.

The best way to work with an LLM is to collaborate – and that means bringing your expertise to the work. In the example you see me trying to educate the LLM on what I consider the basic best practices of subject lines. As hard as I try, it doesn’t seem to get it. But that’s not an excuse for poor quality work. Be sure to bring your expertise to the table – and take it offline if the AI isn’t getting it.

6. Take advantage of free resources

To my knowledge, Only Influencers (OI) hosted the first email-marketing-specific free webinar on using AI to create full campaigns, not just copy, for email marketing messages, back in February 2024. It was structured as a workshop, with the idea that people could follow-along at home and get hands-on training. Tamara Gielen, Materialise, and I led it.

In hindsight, we had much too much content to cover, and it was too much to ask people to follow-along at their offices. We fixed these issues in the four-hour pre-conference workshop we led in Phoenix before Email Innovations World.

But…

The original webinar still has value, and it’s available for free, on-demand, on the OI website. Also available on the OI site, the deck we used for the presentation.

And a companion webinar and deck – on using generative AI for data analysis and reporting – is also available for free, on-demand. I was honored to share presentation responsibilities here with my colleagues on the OI Metrics Committee Paul Christmann (rasa.io), Luke Glasner (Glasner Consulting), and Elizabeth Jacobi (MochaBear Marketing).

So if you are AI-curious, these webinars are a great free way to get started.

Six down. Twelve to go.

So here are my first 6 tips. Visit my Email Optimization Shop blog to get 6 more this week, and then the final 6 next week.

And if you’re looking to learn more AI, watch this space – and my blog. Even better, sign-up for the Only Influencers email newsletter and my Email Optimization Shop email newsletter to stay on top of free and paid resources to up your AI for email marketing game.

And when, not if, you try these tips on your own, let me know how it goes! 

Until next time,

jj

Photo by Danielle Rice on Unsplash