Thoughts on Marketing

How has AI changed your work day? For many people the answer, right now, will be “not too much”. Maybe you’ve been playing around with ChatGPT and other AI tools. Maybe you’ve even integrated it into some of your daily workflows. But from what I see, most of us are just scratching the surface of how AI might transform our processes.

The aim of this article is to reassure you that it’s not too late to embark on your own AI journey .And to give you some thought starters.

 I’ll talk through some of the options available – things like Chat GPT vs Claude vs Gemini. I’ll highlight some use cases of AI that will be helpful to most people in a marketing or leadership role And share some thoughts on where this goes next.

Prefer to listen? This article started out as an episode of The Marketing Mix podcast:

Why AI, Why Now?

Sam Altman, CEO of OpenAI is quoted as saying that ‘AI will do 95% of the tasks that marketers use agencies for”. Which is hyperbole, and I don’t believe it – at least, I don’t believe the 95% number. It will absolutely replace a lot of the current marketing tasks, and make marketing tactics accessible to people and companies that don’t have big marketing budgets.

 But it will also create opportunities. And I’d like to think that anyone paying attention to the trends – someone who listens to The Marketing Mix podcast maybe – is the  kind of person that’s going to be up to the challenge of taking advantage of those opportunities.

I mention the Sam Altman quote because AI, like any new technology, can  generate equal amounts of excitement and fear. And the easiest way to tap into one, and assuage the other, is to start to use it. So for anyone who asks me how to get started, my answer is something along the lines of “just get started.” Which primarily means, create a free account with ChatGPT and start playing around. But it can also mean taking a closer look at the AI features in tools you already use, like Canva or Salesforce, or your email app

Getting Started

But that’s probably not enough advice to do more than frustrate someone. So I want to give more specific ideas on how you might get more comfortable with using AI in your day-to-day workload.

So I’m going to  walk through some of the initial things to consider when you’re starting out. I’ll look at the pros and cons of the most popular models, and which one you should use. We’ll talk through three use cases for generative AI that most people can start to use immediately in their workflows; and I’ll consider where AI might go next.

And I should say at this point, I believe there are three camps, in terms of how people are approaching AI. Those who aren’t paying attention. Those who are playing around with it. And those who are using it in their day-to-day work. I hope I’ll provide something useful for each one of those camps during this episode.

Before I dive in, a quick word on terminology. I find myself referring to the AI services, like ChatGPT, in many ways. I’ll call them bots, AI assistants, AI tools. I refer to them as “they” and talk about them “thinking,” even though they don’t. It’s not ideal, and it clouds the issue a bit. But let’s put that down – ironically – to me being an imprecise human.And also to our tendency to humanize machines.

I’ll also talk about LLMs, or Large Language Models. These are the engine that the AI bots are built on. You don’t need to know much about them when you get started. For now, just think of it in the same way as the data that Google scrapes when you ask it a search query. It’s important, but you don’t really need to pay attention to it.

What to Consider when you Consider AI

One of the first things people will talk about when you bring up AI is the idea of hallucinations, or – effectively – wrong answers and bad information. And that is undoubtedly a concern, but I think as you use the AI tools more, you’ll start to put that in perspective.  

Jon Udell has been using Large Language Models and generative AI for longer, and more extensively than I have. And he writes about it in his column in New Stack. In a recent article, he talked about the 7 Rules for working with LLMs.

It starts with “Think Out Loud” – encouraging you to have an ongoing conversation with the bots, and to keep asking and digging. I personally find this to be a key component of being successful with AI. Number two is a key one – “Never trust, always verify.” Which taps into the horror stories we’ve all heard of hallucinations and non-existent references. A clear reminder that AI is a resource, not a paragon of perfection. And Jon ends the 7 rules with “Learn by Doing”. And that’s one of my key suggestions to people who aren’t sure. Get stuck in. the more you try, the more you’ll learn, and the less daunting it will seem.

Which AI Tool Should I use?

OK, so you’re ready to get started. Now then. Which of the models to use? Chat GPT, run by OpenAI,  is the main one that people talk about – they did lead the charge and are still, arguably, ahead of the pack. Then there’s Claude, by a company called Anthropic; and Gemini from Google. There are a bunch of others as well, often tuned for specific use cases,. But most people will start with one of these three. 

There’s a few resources that try to help here. Just make sure you’re reading a recent article when you do go looking for help, because the models are constantly evolving and improving.

Timothy B.Lee who writes the newsletter “Understanding AI”  compares the new version of Claude (3.) with Chat GPT 4, and also Gemini. He ran a range of tests on them, to see what they’re good at. You should read the full article, but at a high level he’s telling us:

1. Different models are better at different things. ChatGPT seems to  be better at counting things. Claude might be better at proofreading
2. None of them are great at everything. And he has a suspicion that they are programmed to do well at certain standardized benchmark tests. As soon as they move outside of those benchmarks, performance drop off
3. And he echoes Jon Udell’s caution to always “verify”

Another article, from the AI For Good newsletter, gives insight from a power user on how the three most popular tools compare. There’s a useful table showing how they stack up on features such as ability to submit attachments, analyze images and so on. So that’s another place to do some research.

TL;DR Start with ChatGPT

I suggest you start with ChatGPT, ideally the paid version. Firstly, the majority of online courses and resources will focus on ChatGPT. I also think it’s the most competent overall, although for certain applications, it may not be the best. And, let’s be honest, we just want to get started, So let’s pick the one everybody’s talking about.

The free version is fine, for now. But there are some features of the paid version that I do find valuable. It gives you the ability to attach documents. For example, I can upload a transcript of a podcast, and it will “read” it and be able to summarize the discussion. It also gives you access to DALL-E, which is the image generation model. I’m still struggling with it – this is one instance where I think a dedicated program, like Adobe Firefly is much better –  but DALL-E is a cool tool to play around with. 

And the paid version also has more updated data, and can access the internet. The free version cuts off at January 2022 – that’s the extent of it’s training data, But in certain applications, you may want to have current data – such as writing a sports article that needs to have the latest scores and stats.Or a research task that needs to have access to up to date information. So the paid version makes sure you can get that.

Three Use Cases for ChatGPT

So let’s assume you sign up for ChatGPT. What, exactly, can you use it for? How can it improve productivity, or just make the work day a bit easier?

There’s three tasks that I think are great applications of ChatGPT, that anyone could find a use for. Firstly as a writing companion. And especially in marketing, this is probably the number one use case. Secondly, I use it for summarizing material; and a third area is as a customized search engine – basically, a replacement for certain Google searches.

A writing companion:

If you have to write a new blog, and are struggling for ideas –  staring at a blank page –  start a conversation with ChatGPT. You could ask it to give you ideas for an article on your industry; or prompt it to give you ten headlines or thought starters. If you have an idea for the article, but aren’t sure where to start – ask it to write a three paragraph summary. And then use that as a basis.

Maybe you’ve written the article, and want to polish it. Ask GPT to act as a copy editor and suggest improvements to the article. Or ask it to summarize the main points, as a sanity check that it’s achieved the goals you had in mind. There are plenty of Linkedin posts around this, so I won’t belabor it. But a writing assistant is a really good way to start.

A summarization tool:

If someone sends you a 50 page PDF which you’d really love to read, but you know you won’t find time – upload it to GPT and ask it to summarize for you, pulling out five key takeaways and five quotes.

I personally use this when someone is pitching to be a guest on the podcast. They often have a book, or an ebook,  that they send over to highlight their expertise. So as a first pass, I’ll ask ChatGPT to summarize it. If I like the summary, then I’ll invest some time to read the original content.

A search tool:

Let’s say you’re running your first big tradeshow event, and you want to get an idea of all the tasks involved. In the past, you’d do a google search, which would give you a bunch of links that you’d then have to read through. And a lot of those links are to self-serving blogs from tradeshow vendors. Plus, they are very repetitive.

Ask Chat GPT to provide a list of the key activities to prepare for a tradeshow and it will do the research for you and provide a list of recommendations. Each of which will have a link out to the original resource if you want to dig in further. You can even ask it to create an excel spreadsheet of the key activities and dates, which then forms the basis of your workflow.

And here’s a fun one. If the show is in Chicago, and you’re taking out a client – ask ChatGPT to recommend a few steakhouses within a set budget, and no more than a mile away from the convention center. I wouldn’t take its suggestions verbatim – but it gives you a better starting point than OpenTable or Google.

Getting Better at “Chatting”

Now, the reality is, when you first start playing with this, you’ll get frustrated. The answers will be vague, or won’t really give you the information that you need. So you need to get better at writing the prompts.

A few tips:

• First of all, give it a role. In the case of the Tradeshow research, I’d start by saying “Acting as an Event Marketer, please provide a summary.” This gives it some context and narrows down the output. 
• And then treat these queries as a conversation. If you were talking to someone at a networking event, you won’t get far with your first question. It’ll be a shallow answer But as you ask more questions, you’ll get deeper and more interesting answers. ChatGPT is the same. 
• The more you interact, the more context it has. So when the first answer comes back, don’t give up. Ask for more details, or clarification. Tell it why that wasn’t the right answer, or give it some direction. 
• Style tips are always useful. Say things like “Using a conversational tone; or “write in the third person”

My main point here is that it takes time,. It takes iterations, And it takes practice. Just like anything else.

I really encourage you to play around with Chat GPT. I think it truly can help with productivity. And most importantly, it will start to give you a sense of the potential of AI.So you won’t feel left behind.

What’s Next?

Well firstly, as much as we’re all talking about how to prompt the AI bots, and how companies need to hire “prompt engineers,” the reality is that’s a short term issue. As this evolves, very few of us will be talking directly to AI models. The value will come from the apps that are built on top of them. As I said earlier, starting out with AI could mean playing with the AI features in tools you already use. And that’s probably where things are going to go. AI’s biggest benefit for most of us will be in making our tools more intelligent and more useful.

So that’s something to bear in mind. What we think about AI in business now is completely different to a year ago. In fact, a year ago, most of us were only just starting to think about it. So you can bet that AI – particularly in terms of real world applications for business – will be unrecognizable a year from now.

When I want to dig deeper and think harder about AI, one of my go to thought leaders is Ethan Molik. He’s a professor at Wharton and he’s doing some serious thinking on AI potential and pitfalls. I often share his material on LinkedIn, but it’s worth signing up for his newsletter called One Useful Thing.

One more thought on this. Premium, paid services are starting to gain momentum. All of the AI models have a paid version, strangely, they all seem to be $20 a month… But much of the money around AI will be in need to sign up for the premium tiers of your existing tools – the software companies will use this as a way to upgrade you from a free tier, or up to a higher-paid tier to access their AI features. But it won’t be a smooth road.

Microsoft is running into some headwinds on their $30 per month fee for corporate users for their CoPilot offering, as reported in the Register:

“Juniper Networks CIO Sharon Mandell told the WSJ: “I wouldn’t say we’re ready to spend $30 per user for every user in the company.” Microsoft’s current position is that they believe in teh value for their email and Teams products, but accept that CoPilot isn’t really doing it for things like Excel and PowerPoint – which, I would imagine, is where the corporate folks really want it to help them.”

There May Be Hype. But Get on The Curve!

All this means to me is that expectations have run ahead of the development curve. My friend Marcio Saito would always come back to the Gartner Hype Cycle whenever we got into discussions about tech trends. And although it may not apply to AI in general – since that’s such a broad category – it will for sure apply to certain areas and applications of AI. 

So be prepared to enter the Trough of Disillusionment for some of these tools and services that have been rolled out very very quickly. Meaning – don’t write off things like the CoPilot service. Microsoft has the time and resources to make it successful. But don’t expect it to be an overnight success either.

By the time you’ve read this article, no doubt some of it will be outdated. Generative AI – and AI tools – are on a pretty steep curve. Which is why the initial point of the article – “just get started” – is such an important message. The sooner you get on the curve, the easier it will be to keep up!