They Said It

“We defaulted customers into a [generative] AI chatbot experience and they hated it. It was one of the worst-performing experiments we’ve launched during my entire tenure.”

-Etsy CEO Josh Silverman, who’s been in that role since May 2017, on a recent podcast (as quoted in TIME magazine)

Welcome to the premiere issue of The Narrative! I'm glad you're here. You might be expecting a long-winded introduction since this is our first issue, but nah...let's get right to it.

-Matt

The Narrative Deep Dive

I'm gonna put my cards right here on the table: AI has been a disappointment.

Don't get me wrong. My jaw hit the ground when ChatGPT launched on November 30, 2022. Yours did, too, I bet.

It answered questions. It wrote stories. It created code. It did all that conversationally.

It was going to change the way we do everything.

It was the future.

Or was it?

Here we are, 20 months later (an eternity in the tech world!), and IMO AI hasn't lived up to the hype -- especially LLMs (large language models) like ChatGPT, Gemini, and others.

In "conversations," they still get facts wrong.

They're only capable of regurgitating the internet content on which they were trained.

They give bad advice. You probably heard about Google's AI telling you to drink urine to help pass kidney stones. You may have missed the one about the moving company that installed an AI chatbot to talk to customers, only to discover the chatbot started referring them to other moving companies.

Sam Altman, OpenAI's CEO, recently said AI will replace 95% of marketing work, but that day sure seems a loooooong way off, doesn't it?

To me, AI is like the troublesome toddler who sticks his fingers in outlets, fans, and everything else within reach. It's the overactive puppy you don't dare leave home alone.

They're not ready to be left to their own devices and not responsible for what they do. Neither is generative AI.

Like the mom or dad who has to keep one eye peeled on that baby (or puppy...or, God forbid, both at the same time), marketers have to keep an eye on the tools we use to do our work, especially the AI tools.

So, how can we responsibly use AI in marketing?

Disappointment or not, AI is (and should be) an important tool for everyone in marketing.

Despite what Altman says, the key is to treat AI like an assistant, not a replacement. You can't outsource your strategic planning to ChatGPT, have Gemini write all your content, or have Claude build your personas and funnels.

You need strong guardrails in place when you use these tools.

Here are 5 guardrails to help us all use AI responsibly in marketing:

1. Hands-on human involvement

Like the smart parent who puts outlet covers on every electrical socket, let's not leave AI to its own devices. Any content it creates must be fact-checked, edited and/or rewritten, and ultimately improved.

(Bonus: This is good for SEO, too. Google's quality standards include a concept known as E-E-A-T, which stands for Experience, Expertise, Authoritativeness, and Trustworthiness. The more of that in your content, the better. In their current forms, LLM-based AI tools are incapable of putting personal experience and expertise into the text they generate.)

We like to use AI at the beginning of our work–to help us refine ideas and organize thoughts. Sample prompt: Here's an outline of an article/presentation/webinar/etc. on [TOPIC]. What concepts are we missing?

2. Bias awareness and moderation

LLMs are only as smart as the online content they're trained on, and that content reflects the biases of whoever created it. So, it's reasonable to assume that "outputs may be biased if the data used to train an AI algorithm is not diverse or representative," as Chapman University says in this overview of how and where bias can creep into artificial intelligence.

As marketers, it's our job to make sure our own implicit and explicit biases don't show up in the prompts we write, nor the outputs we get in return.

3. Transparency in identifying AI

Fooling or tricking people is pretty much never a smart marketing tactic. Ask Sports Illustrated about the backlash it faced last year when it was caught publishing AI-generated articles under the bylines and portraits of AI-generated authors.

If you're going to publish AI-generated content, labeling it is a Good Idea. It's also either suggested or required by many of the major internet platforms, from Google and YouTube to Meta, TikTok, and beyond. (So now is a good time to let you know I made the cartoon image above in ChatGPT!)

If you're using AI to engage with the public–a customer service chatbot, for example–disclose it. Going back to the quote that began this issue, I wonder if Etsy customers knew they were talking to an AI chatbot.

4. Data protection and privacy

If you're developing a custom GPT for specialized tasks in your business or industry, put strong measures in place to protect sensitive customer and/or company data.

If you're intentionally collecting consumer data for the purpose of feeding a custom GPT, get permission and be clear about how you'll use the data.

5. Accountability and education

We're at a point where you (or your marketing team) need some kind of systemized approach to A) monitoring all of the above with regards to how your company is using AI, and B) staying abreast of the latest developments, new tools, and especially the potential risks and legal minefields that come with artificial intelligence.

What have I missed? How are you ensuring that you use AI responsibly?

Wrapping it up: I don't think AI has yet lived up to the hype and promise that came along with the launch of ChatGPT and the development of competing platforms. (It may never live up to the hype; see Ed Zitron's article below.) Right now, AI is best thought of like the baby or puppy you have to keep a close eye on at all times. That's why it's smart to have guardrails in place. Hopefully, the five above can help you use AI responsibly in your marketing work.

Go Deeper

For more about using AI, check out the following:

• ​Pop Culture from Ed Zitron, who says the AI bubble is deflating
• ​13 Principles for Using AI Responsibly from Harvard Business Review, which goes beyond marketing to examine the bigger picture of AI at work
• ​AI ethics: How marketers should embrace innovation responsibly from Sprout Social

​

From the Narrative Bent Slack

"I've got this one mailing list for real estate where the founder uses heart-stopping subject lines like:

• 'Oh no!'
• 'It ends today'
• 'Was this a mistake?'
• 'Are you okay?'
• 'Help'

And I don't think it's a valuable strategy, but damned if I don't look every time."

-Jenna Inouye, Narrative Bent editor

-30-

You made it! This is the end of our first issue of The Narrative. You'll know you've reached the end when you see that -30- ... that's how journalists historically indicated the end of a story in pre-computer days.

Since this is The Narrative, let me share the story behind the use of -30-. It's interesting–mainly because no one is quite sure when or why it began. One theory is that it dates back to the days of telegraphs, and -30- is how you signaled "the end" in Morse code. Another theory surrounds a telegraph operator who died on the job and hit the "3" and "0" keys when he collapsed. It was his "end." Morbid, huh?

Anyhoo...thanks so much for reading The Narrative. The opinions shared above are my thoughts, but our team is full of sharp marketers with their own takes. I expect you'll hear from them in future issues. We're a thoughtful bunch who love diving into all sorts of marketing debates–it keeps things lively and pushes us to stay on top of our game.

I'd love to hear what you think so far. Reply to this email, or just use the links below to tell me if you liked it or not. If you have ideas or requests, maybe some topics you'd like us to discuss or questions we can answer in an upcoming issue, we're all ears. The Narrative has our name on it, but it's your newsletter. Don't hesitate to tell us how we can make it better and more relevant to you.

Our plan is to publish every fortnight, so we'll see you again in a couple weeks.

Until next time,

Matt McGee

‍