‘Probe It, Test It, Try to Stretch the Limits’; an AI Practitioner Shares Secret Sauce

There was a terrific article written a couple weeks ago by Sophie Culpepper on NiemanLab headlined, “Can AI Help Local Newsrooms Streamline Their Newsletters? ARLnow Tests the Waters.” She introduced me and many others to two journalists doing incredible stuff with AI in getting out their publications. (I’ve been in touch with both, and they will join us on a future webinar/Editorial Council meeting.)

Scott Brodbeck is the founder of Virginia-based media company Local News Now. For his ARLnow site (covering Arlington, Va.), he wanted to develop a morning email newsletter with more voice than his afternoon one. So, Culpepper wrote, “he began experimenting with a completely automated weekday morning newsletter comprising an AI-written introduction and AI summaries of human-written stories. Using tools like Zapier, Airtable, and RSS, ARLnow can create and send the newsletter without any human intervention.” Now he’s working on a two-minute, AI-generated audio news summary and tells me he’ll have more for us in a couple months.

Joe Amditis is assistant director for products and events at the Center for Cooperative Media at Montclair State University in New Jersey. He “recently released a free handbook detailing applications and considerations for AI use by local news publishers.” I was fortunate to get to talk to Amditis last Friday about what he’s doing. Interestingly, Zapier and Airtable came up often in his conversation as well. In fact at one point, he stopped and said, “Okay, well you need to drop everything you’re doing after this call and just learn Zapier.”

In addition to featuring Brodbeck and Amditis at a later date, our AM&P Network will be exploring AI in a Main Stage session at our upcoming AMPLIFY 2023 Content & Marketing Summit, June 27-28 here in Washington, D.C. The session, Practical Things You Can Do Right Now With AI (and How to Use It Ethically) will take place on the 28th. It will give you an excellent foundation for your AI learning curve. Check out the full agenda here and registration information here. It’s an event not to be missed!

Here’s some of the enlightening conversation I had with Amditis:

RL: How have you been using AI mostly?
Joe Amditis: So we have like a million newsletters, and it drives me crazy. But part of the reason why [AI] is so useful is because we have so many, and it takes a lot of time to compile them. And personally, I found that these tools are most effective and most useful, not when you’re just going into them blank and asking it to make something for you from scratch. It’s awful. Honestly, it’s high school writing level, at best usually, unless you get really specific with it. But what it can do, what it’s really good at is taking already existing content, and then synthesizing it or repackaging it or repurposing it into a different format—or extracting takeaways and doing summaries and things. So acting on an existing set of information, as opposed to asking it to sort of generate from scratch is incredibly useful.

Where’s the majority of your usage?
I’ve tested this thousands of times, quite literally, at this point. The majority of our usage is—I would imagine we’re very similar to [your audience]—administrative tasks or copywriting and copy editing and organizing text on your screen, and to summarize and generate takeaways. We can make things more concise, repackage them for different platforms or mediums whether for a newsletter, or a LinkedIn post or something like that. We’ll take our existing stuff that we’ve written and produced, and then repackage that.

Can you explain further how you do this?
When I have a new article that we’ve written, let’s say I want to do a series of actions. I might want to make this easier and more concise, like a summary for our newsletter or for a partner newsletter that goes out. Let’s say for our collaborative journalism newsletter we have a news item and a headline and summary of the main points and then we have a link. What I’ve done is gone to Zapier and to Airtable, and I’ve created a base for automation, specifically for that. And I’ve got this setup here [to] use the Airtable Web Clipper—it’s a little browser extension that allows you to pull anything from the Internet into an Airtable base. Now I have the whole text of the article and then I can select the tone or the vibe of the text that I want Chat to give to me. I want to use maybe “educational, thoughtful and optimistic” let’s just say. I want to select a purpose. We have our different newsletters or promotional channels so I’ll say, general promotional channel; that’ll end up in the promo and will channel in Slack ultimately. And then here it automatically pulls in the URL to that post. So I hit Add Record.

Okay, I think I’m following.
Zapier sees that record in Airtable and then it sends that information to ChatGPT through their integration with the prompt. So I’d say summarize the following text; the summary will promote the event article and information contained in the text. And I wanted to create a list of three possible headlines. And again, the tone of the headline should be thoughtful, educational, optimistic. And it can draw on that as additional context for the audience that it’s going to be sent to. Then I do the same thing for a series of tweets, five possible social media posts. Again, here your job is to create social media promo copy,

And the last thing is why people should care?
They care about community engagement, trust in journalism, marginalized voices. So this is now like cannon fodder for our newsletter writer and compiler to go through, and not only have some starting material to get going, so they don’t just start with like a blank MailChimp campaign. It’ll be heavily edited. In fact, most of the language that the bot spits out, probably won’t make it in its original form. Although sometimes it’s just like, “this is actually a good summary.” And then our newsletter writer can take that into MailChimp. And that why should people care about part gives us a sense of the angle or the approach that we might take to sharing and promoting this, because a lot of times, I’ve found personally and with our colleagues that your brain just locks up when you go to tweet or share something out. And you’re like, “you should care about it, because it’s new, and I made it, and I’m proud of it, right?” And instead, this gives us an opportunity for a little bit of help contextualizing.

Are those apps all separate or part of a bigger thing?
They’re all separate apps; they existed before opening AI and everything. I’ve been using Airtable for about six, seven years now. You don’t need Airtable for this. In fact, you can use Google Sheets.

How did getting up to speed on all this work for you and your department?
I still can’t get my boss and my co-workers other than maybe one (on board)—and we only have four co-workers who are in the office at any given point. Only Cassandra, our newsletter writer, has really even started to lean into this a little bit, and it’s because after a while, she’s like, “Wow, it’s taking me three days to do this bi-weekly newsletter.” And [I say,] “Wouldn’t it be great if you had a tool that could cut off some of that time? Then I build the automation. And now it becomes almost irresistible because she can’t not look at the summaries; they’re in the Slack channel. It’s a very low-hanging fruit for her.

It’s not easy, I imagine, after we’ve all been doing this a certain way.
There’s this feeling with a lot of people when it comes to using their computer, especially if they’re not already confident, or they’re very nervous about computers, that they feel like if they press the wrong button, or they click on the wrong thing, it’s going to delete their entire operating system… And there’s this tension that people are still a little hesitant about. That’s going back to just computers in general, let alone this new, scary slash exciting AI stuff that they feel that same pressure to not break everything. And the reality is, all you have to do is talk to it, back and forth and see what it can do. Probe it, test it, try to fool it, try to stretch the limits of what it’s capable of doing. And you’ll start to see the outlines of what the actual uses of the tools are—if you go through that experimentation process.

More can be found on The Center for Cooperative Media here. Amditis said to let them know if his handbook and/or advice are useful for you. “You could also tell funders to give us money, but that’s a whole other thing.”

Comments are closed.