On the Matter of AI in PR

I recently penned a piece on NotebookLM, an AI platform now in beta at Google Labs. Toward the end, I wrote, “I’m pretty sure I’m just scratching the surface of what this cool new AI tool is capable of doing.” The PR world is ablaze with those claiming to use AI and machine learning technology in their daily workflows. I sense that most are not. (There’s considerable hyperbole in our space.)

Some AI-infused resources I regularly use include Muck Rack, Grammarly, Rytr, and Otter.ai. But for this post, I thought I’d talk about a few of the free GenAI chatbots.

We all know about OpenAI’s ChatGPT 3.5 and Google Gemini (formerly Bard), but my default for surprisingly informed conversations is Pi.ai, “the first emotionally intelligent AI.” My wife and I have held congenial conversations with Pi on a range of subjects, including which mortar to use to mount ceramic tiles, the caloric count of lobster meat, and when to harvest the garlic we planted in the fall. No keystrokes needed. Pi’s easy to talk to and prompts further discussion on the subject at hand.

As for NotebookLM, I just used it again for an upcoming editorial board meeting we arranged for a client CEO with an influential national news organization. To prepare, I inputted onto the platform all the editorials and opinion pieces (<20) the publication ran this year on the industry in which my client resides. I then posed to it the following questions: what is the publication’s position on this issue? What are the ten likely questions a CEO from this industry will be asked? Almost instantly, I had what I needed to greatly enhance the meeting prep document.

I have used both ChatGPT and Google Gemini for business queries including recently to help identify a high-profile individual to serve as a co-chair for a big citywide extravaganza. Both yielded excellent choices — except for when they didn’t. Barbara Walters and Sally Ride are just not available.

A client recently asked me to put together some background information on an influential reporter poised to meet her CEO. In addition to the usual fare, e.g., name, title, bio, photo, and past pertinent articles, I found in my Dropbox an audio file of the CEO’s last meeting with the reporter.

Dropbox offers audio-to-text transcription, as does my premium service Otter.ai. I ran the audio through both. The results required a manual listen alongside the text to identify who was speaking. Otter.ai fills in the names after the first few speakers are ID’d. Dropbox doesn’t, while its formatting — one contiguous text with multiple spaces between words — is too time-consuming to correct.

Enter the third LLM chatbot, and my new favorite: Claude from Anthropic, which just released its next generation. FWIW: Amazon just invested $2.75 billion in Anthropic, even as it seeks to build its own chatbot — Olympus. Claude 3’s benchmarks apparently outperform those of ChatGPT4 and Google Gemini.

Back to my use case: I took the unwieldy text from the audio of the conversation between my client CEO and the reporter and cut and pasted it into Claude3’s text box. I then asked, “Can you organize and summarize the transcript of this interview?” In less than five seconds, Claude produced a document in sections with headers and bulleted points my client made under each.

I followed up with a second question: “Thank you. Can you now summarize the reporter’s questions?” Again, in under five seconds, Claude produced a list of the ten questions this reporter posed that day two years ago. It was followed by a paragraph summarizing the reporter’s interests: “In general, the reporter was probing for details on…”

Again, the key to AI adoption by our (or any) industry lies in the use cases the technology can create and the smart, time-saving productivity that results. Claude certainly delivered on the promise.