Each week we'll gather headlines and tips to keep you current with how generative AI affects PR and the world at large. If you have ideas on how to improve the newsletter, let us know!
What You Should Know
Chatbots vs. Agents Explained
When Mark Zuckerberg decides he needs a “CEO agent,” people pay attention. It is the biggest indicator of the latest AI trend, one that hasn’t been very clearly defined. Where exactly do chatbots end and agents begin? Last year, Gartner warned that more than 40% of agentic AI projects would be canceled by 2027, in part because of the complexity of AI agents and what it called “agent washing,” the rebranding of existing AI capabilities as agentic.
To be clear, there are some innovative use cases that truly deserve the “AI agent” label. The easiest way to think about it is this: A chatbot helps you think and perform individual tasks. An agent executes entire processes without the handholding.
If you need to brainstorm, tighten a quote, draft a pitch, summarize a briefing, rework messaging, or scrutinize an idea, that’s chatbot territory. That is still the bread and butter for most communications teams. You’re using AI as a thought partner or writing assistant. It’s fast, useful, and usually enough. An agent is more useful when the job involves multiple steps, tools, sources, and some action.
An engineer in Europe created an agent to call more than 3,000 pubs across 32 countries to find the price of a pint of Guinness on St. Patrick’s Day weekend. It documented the answers and created an inde— sorry, a Guinndex of pubs and prices.
There are some practical use cases for communicators, too, but identifying them requires some thoughtful self-reflection on what you’d like AI to help you with and how to break down those workflows.
Take media-list building, for instance. A chatbot can suggest relevant reporters, but an agent could take a press release draft, identify the 20 best-fit journalists based on beat coverage, recency, and publication tier, check those names against your team’s relationship notes, then tee up personalized pitch drafts for you to review.
Same with reporter monitoring. A chatbot can give you ideas on which reporters to watch. An agent could continuously track what those journalists are covering, posting, and engaging with, then flag when a reporter’s focus starts shifting toward an area relevant to your client. It could also help generate meeting agendas by updating previous iterations based on a coverage tracker or folder contents.
Agents can be genuinely useful, but their value usually shows up when the work is structured, repeatable, and tied to action. Chatbots still make more sense for the everyday thinking and writing tasks that fill most communications roles. As more stories push the “agent” bucket, communicators will need to get better at separating what sounds impressive from what is actually helpful. Start with the task, then pick the tool that fits it.
Elsewhere …
- America’s Next Class War: AI Fluency
- Anthropic Economic Index Report: Learning Curves
- Google Unleashes Gemini AI Agents on the Dark Web
- Can AI Predict the Next Breakout NHL Star?
- OpenAI Is Shutting Down Sora, Its AI Video Generator
Tips and Tricks
⏲️ Set it and forget it
What’s happening: One way to dip your toe into agentic waters is to set up tasks in ChatGPT. Tasks run a schedule, even when you’re offline, and send results via notifications or email. OpenAI also says you can keep up to 10 active tasks at a time, but tasks do not support voice chats, file uploads, or GPTs.
What you’d use them for: The best use cases are small, repeatable jobs you already know should happen on a cadence but often don’t. Think daily scans for client, competitor, or industry news; weekly check-ins on what a handful of target reporters are writing about; Monday-morning prompts to surface fresh ideas, commentary hooks, or stories worth pitching around. The sweet spot is lightweight monitoring and prep work, not anything you’d want to share without a human checking the output.
How to set them up: Start in a normal chat window and tell ChatGPT what you’d like it to do. Instead of “monitor AI news,” try something closer to “Every weekday at 8 a.m., summarize any news about [client], [competitor], and [topic].” The more specific the prompt, the better the results.
To manage your tasks, go to Settings, Schedules, and click “Manage.” There, you can refine the instructions and timing.
Quote of the Week
“It is particularly painful that I made precisely the mistake I have repeatedly warned colleagues about: these language models are so good that they produce irresistible quotes you are tempted to use as an author. Of course, I should have verified them. The necessary ‘human oversight’, which I consistently advocate, fell short.”
— Peter Vandermeersch, Senior Journalist at Mediahuis, in a Substack post after he was suspended for using AI-generated quotes
How Was This Newsletter?
