Getting to Grips with AI
Plus. justifying marketing attribution...
I’m working on a new website for Solent Strategies, and I have a photo that I really want to use for the home page banner. The problem? It’s a vertical image, but it needs to be landscape to fit the space.
Luckily, my designer’s always willing to experiment! He uploaded the image to Adobe Photoshop, tried a few AI prompts and - as if by magic - the image had ‘grown’ on either side. It now has a believable coffee shop background on either side of my smiling face.
We’re hearing a lot about AI. Some of it’s useful, a lot of it is just “hot takes” (aka ‘hot air’). So I’ve been thinking about how AI is likely to affect the day-to-day of how we get stuff done in marketing.
The Photoshop example may be trivial, but it’s indicative of how it’s already making life easier. I didn’t have to set up a new photoshoot (saving me money), didn’t have to wait 2 weeks for the new image (saving me time), and didn’t need a deep knowledge of AI techniques. Just a talented designer using his standard tools.
Then, a few days ago, I was talking to a business owner about his early attempts working with ChatGPT, and specifically, how we can write prompts that produce useful outputs.
Except, I don’t think that’s how it’s going to work.
For most of us, the real value of AI tools is going to be the apps layered on top of it. And to illustrate that, let me tell you about my podcasting process…
But first, did you listen to last week’s episode yet?
Podcast - Marketing Attribution
Attribution is a fancy term for understanding when and where a customer interacted with you during their decision process. The more you know about their actual buyers journey, the better you can decide where to invest your marketing dollars.
At least, that’s the theory. But - as we discussed on this episode of the podcast - does it work, and is it a good use of resources for a small, growing company? Steffen Hedebrandt’s take: can you afford not to?
“ How do you not justify it? Because you're spending a lot of time and money every single week…on activities that are supposed to drive attention about what you do.
You can only spend a dollar once in your company. So you want to be able to put that dollar where it looks like you get the most impact out of it.”
Tough to argue with that!
So we talked about the best approach - first touch vs last touch, and even multi-touch - and the importance of collecting data wherever possible, even if you’re not sure how to use it yet. And, since sales cycles can take months or even years, you should set up the data collection early, before you’ve built the attribution model.
It’s a big topic, and one I’ll dig into deeper on next week’s episode too. I’ll be talking about the complexities of attribution in B2B and multi-level models; and how to measure meaningful attribution when you can’t attach a marketing metric to actual revenue.
AI Tools vs AI Models
I promised to tell you how my podcasting process shows a different path for how we’ll use AI in marketing.
Recording the interview with my guest is the quickest part (and the most fun). Then the real work begins.
A transcript is generated automatically by Riverside a couple of minutes after the recording ends. That lets me quickly review the conversation and highlight the best parts, and any sections that need editing. It’s not perfect, but it’s good enough for the job.
I upload that transcript into Deciphr.ai, an app that is built solely to create materials for podcasts. It generates show notes, summaries, timestamps and suggested pull quotes. It’s not perfect, but it gives me a really good starting point. I’m doing the editing, not the grunt work.
I take chunks of the video and upload them to Descript (h/t Jon Udell), which lets me create snippets via a text editor. As I delete words from the transcript, it edits the video automatically...almost magically. That allows me to quickly make short video bites for social media posts. It’s not perfect, but it gets me 80% of the way there.
It’s easy to get caught up in the battle to learn the right prompts to make ChatGPT and Midjourney do your bidding. But in reality, the path to success is finding the right tools that others have built on top of the AI engines.
Don’t wait for AI tools to be perfect. Start using them for the 80% solution, and you can invest your time in the 20% that makes the difference.
Drink of the Week
I’ve spent the last few days in England, and whenever I’m back “home,” cocktails take a back seat in favour of a few pints of ale. The winner this trip was a pint of Butcombe Original.
Cheers!
Marketing that Scales
If you know you want to have more impact with your marketing, but aren’t sure where to start…
If you’re investing in campaigns and promotions, but they aren’t really connecting...
If you want to punch above your weight…