“We’re not very good at telling our own stories,” Joe Amditis, assistant director for products and events at the Center for Cooperative Media at Montclair State (NJ) University, told me last week. That came up when we spoke about the importance of audience trust. “We need to be “illuminating the process behind [our] journalism, not just in the use of tools, but more generally like what goes on behind the scenes or behind the presses.”
Here, Amditis shares more of his work using AI.
“If you’ve ever set up an auto pay for a bill, then you can set up an automation for your newsletter for your content process, or whatever your journalism process,” Amditis said. “I would advise people to start small, choose one task that you do over and over that you would love to [automate]. The steps can remain the same, but the content may change a little bit.”
Amditis has put out the “Beginner’s prompt handbook: ChatGPT for local news publishers.” The guide takes users through creating the best prompts; talks technology terms; tells how to clean up transcripts and create outlines to “red-team” your story ideas. It also advises how to use AI as institutional memory for your newsroom.
He also works with AI every day in his newsletters, social platforms and events. I posted a Q&A with Amditis a few weeks ago. But everything changes so fast in our AI world that I wanted to catch up with him again before we escape into summer.
Ronn Levine: You talked last time how your process has changed. Can you expand on that?
Joe Amditis: Our daily News Roundup is the one where we started [with AI], because we used to spend like 3 hours a day doing this newsletter. And a lot of it turned out to be stuff that people can get elsewhere. Now it’s just 5 headings—politics, environment, housing and development, education and health, and ‘the bucket’—anything that doesn’t fit in those categories. And so, as I scroll or scrape through the Internet in the morning to put this together, I find stories. I open them up in a new tab, and then, once I’m done collecting, throw it into the Airtable web clipper. The automation packages it and spits it out into a Google sheet fully formatted with HTML for bulleted lists and links and attributions. So in that end that’s all automation. For the internal stuff we use AI to summarize and create possible angles or give us a list of possible promotional tracks for the stories that we write or the content that we produce.
RL: Where would you advise that people start, maybe some time-saving elements you’ve found?
JA: To avoid communication redundancy, I would say. Not everybody works on the same apps, uses Slack and has the timeframes everyone else has. That can be frustrating sometimes. Instead of having to manually open up a new window and do all this and blah blah blah, you could automate that very easily and save yourself some time and a little bit of sanity—and then just build from there. Use that as a jumping off point or a starter project to try to free yourself up to do more substantial and consequential tasks.
RL: Do you have an example?
JA: Meeting notes for one. You can take a transcript, or maybe even a series of minutes that someone’s typed up and reformat them, putting them into bulleted lists—adding Markdown formatting. (Markdown is an incredibly useful computer language.) If you have a set of instructions on how you’d like to format something, you can set up those instructions through Zapier. The bot will take the text that has been submitted—whether through a Google form, or even a Slack message with a certain tag or in a certain channel. It’ll make decisions on how to label things, and once you iterate and tweak that process and give it a good run through to make sure that it’s consistent. Now you have an auto format that can take any content within reason with the instructions you provide and output it in whatever format or organization you want.
RL: There’s a tendency for content folks to overlook social media—we have so many other things that we’re doing. How can AI help there?
JA: Have [ChatGPT] suggest alternate headlines and tweets. Ask it to suggest a list of five possible social posts—things like that. I really respond well to and work well with lists. Maybe it’s one-paragraph summaries. Your approach to this stuff will determine ultimately what the most useful applications for these tools are for you and your team. Just dig in and start playing around with it. I can submit something into Airtable via the form, the bot takes that and creates a one-paragraph summary of the content—a bulleted list of why people might care about this. I find that really useful.
RL: You also mentioned transparency the first time we talked.
JA: We disclose how we do our journalism. Journalists could probably, in many cases, do a better job of illuminating the process behind their journalism, not just in their use of tools, but more generally like what goes on behind the scenes or behind the presses. We always need to do a better job of opening up the doors and showing people how we arrive at the journalism that we produce, because that’s what ultimately builds, maintains and sustains the communities’ trust in the work we do. [Check out @emilydaviesreports on Instagram for a great example of this.] If we don’t have that, then no tool, no quick fix, no artificial intelligence, and no chat bot will help us.
RL: Thanks again Joe. Will talk to you in the fall.