5 ways Chat GPT-4 (and AI in general) can help PR agencies

Category: PR Insights

By: Rich Leigh

In 15 years of PR, I’ve seen my fair share of Next Big Things. That, almost always, gives way to the Next Big Thing after that. 

By now, drones should be delivering our 3D printed condensed calorie capsules, while we wave to the colleagues we’ve not once even met outside of our office in the metaverse.

PR has always been quick to pick up new technology if it means selling an idea to a client to capitalise on initial media interest in that technology. It means being quick to use it in a token way, while the press might care, and then dropping it when that wanes, usually as a result of the tech not being able to quite deliver on its promise.

This isn’t technology orientated, and it’s a story I’ve told a few times now, but I remember delivering a talk about PR for SEO, or linkbuilding, back in 2013, having been at that point earning editorial links for clients for 5 or so years. 

‘I’ll never be geeky enough to care’, said one person to me afterwards. It was good-natured, but telling. 

Well, SEO agencies, many having run out of links to buy (HA), bulldozed their way into what could and should have been PR client wins, and the industry has been left, arguably, in a worse state as a result (that’s a blog for another day). I’d assert that the ‘not geeky enough to care’ attitude from more traditional PR practitioners made it all the more easy to do. PR should own link-building across news publishers for many reasons… anyway, I digress.

I’ve recently seen people say that they’ve muted AI keywords on social. That it’s ‘not interesting’, and that it’s being talked about in a way that’s ‘too much’.

That’s ‘I’ll never be geeky enough to care’, 2.0, to me.

Everything related to Chat GPT, Dall-e, Jasper, Midjourney and the like feels very different to other Next Big Things, and I’m one of the people asking the PR industry to pay attention. 

Not because it’s shiny, but because I believe this is transformative. We rightly baulked at the idea of the Zuckerverse, and it appears even he is taking the hint, his ambition and investment instead being redirected towards AI. One socially awkward man, seeking solace in an unreality, rich as he might be, cannot define the boundaries of human interaction.

Last week, my friend and business partner Alex Wilson wrote about how he believed AI might impact PR.

I’ll get a bit more tactical and agency-focused.

Here are five ways I believe the AI happenings of this last week (and indeed, the last six months, if you count the public release of GPT-3.5 and Dall-E, which I had a good and practical look at here last September) can help PR, and public relations agencies in particular.

1. The Good First Draft

Some first attempts at content are ready to be viewed by client eyes. Very few are. Many need much more work than that.

In all of my playing around with Jasper and Chat GPT, and of course it depends upon the prompt, I’ve not seen anything less than a Good First Draft. Something to build upon. 

This really isn’t nothing.

Being incredibly optimistic, a good first draft of a press release from somebody in their first few years of PR can take around an hour. That will then need reviewing and rehashing. Back and forths that easily add up in terms of time and cost. The same for blog content, scripts, social media content and all other written content a communications professional might expect to produce. Sometimes, getting started is the hardest aspect.

GPT-4 can turn a prompt of a few sentences, with guidance about style and tone, into a legible press release within a moment.

Microsoft’s CoPilot launch video from last week even featured a press release draft, and how quickly that announcement could be turned into a PowerPoint presentation. With the critical mass of users Microsoft has across its suite of Office products, it’s easy to see how integrated AI wins in a way that the metaverse and other VR/AR attempts haven’t yet.

Reactive quotes can be penned within seconds – and we all know how small the window is to write, approve, get approved and pitch when it comes to newsjacking. The sooner you have a couple of paragraphs in front of you, the sooner you can add, amend, get client approval where needed and fire it out to relevant media. 

2. Tool-builds

I learn by picking things up and tinkering. Within a few hours of GPT-4 being publicly released (if you’re willing to pay $20 per month) this week, I’d created this PR stunt and campaign idea generator, inspired by a thread by a man who’d prompted GPT-4 to code a (genuinely fun) version of Nokia’s Snake game. I’ve next-to-no experience of coding.

Now, it’s not about to replace your creative director, and it needs some design love. But, every aspect, from the coding to the ideas, was created by GPT-4. I simply had to a) have the idea and understand enough to consider the sectors we might want ideas in as an industry, and b) paste the code into Replit.

I built it to create ‘safe’ and ‘controversial’ ideas. Besides one I saw it come out with – it involved creating politically-themed sex toys and brought Margaret Thatcher into it – it’s not shy in coming up with… interesting and potentially morally-questionable tactics!

We’ve spent a lot of client cash and agency time building and promoting onsite tools, calculators and games for clients. Many of them began as ideas to force the link from media outlet to our client’s site, with some going on to be fantastic anchor content and traffic magnets for those clients (like this childcare costs tool, and this for WhoCanFixMyCar, arming motorists with accurate likely-costs for vehicle fixes, all based on data they already had at their disposal).

At the very least, we can arm developers and designers with, again, first draft tools to visualise what we’re hoping to achieve. Beyond that, with the right knowledge and prompting, one can create and deliver onsite tools for clients, from idea to execution. Quickly and inexpensively. Whether this is a positive thing long-term is for people far brighter than me to consider, and I do have an in-built preference for paying experts to do a job well, but as these AI tools and our interfacing with them improves, so does the quality of the output. One only has to consider how Canva has made mid-quality design easy, accessible and affordable.

3. Creative ideas

Following on from that thought, I’ve been quite dismissive of AI’s ability to turn out workable creative ideas. ‘Creativity is what clients will pay a premium for’, I’ve thought.

And, I still believe that.

But have a play with that ideas tool. Three quarters of those ideas would be thrown out in a brainstorm, and either binned or built upon. You’ve almost certainly heard similar riffs on many of those ideas.

What’s better – to pay five people to sit in a room (or worse, on Zoom – I bloody hate online calls for this) for an hour, spending three quarters of that time thinking about and getting the obvious ideas out, or those same people first prompting AI, and coming in with a list that can be easily dwindled down to winning thoughts?

In Alex’s post, he says:

With the right prompting, we’ve been surprised at just how good AI has been at ideation. However, it’s inconsistent, and humans are still needed to choose and refine the right ideas.

In the “Creativity” module on our course, we talk about “divergent thinking” and “convergent thinking”. Essentially, AI is surprisingly good at the “divergent” bit, but you still need a human for the “convergent” part of the process.

So don’t dismiss AI-powered creativity, and think about how you can build it into your creative processes (while not relying on it totally).

I’m not saying you’re getting a Cannes-winning idea from AI (yet – though some prompting asking it to do exactly that did get to quite a nice idea about repurposing ‘old, gas-guzzling vehicles into environmentally friendly solutions or creative public installations’, which is a nice jump-off point for a company that sells electric vehicles).

With the right prompting, it can become just another tool in your creative arsenal.

4. Profitability

I’m going to say it.

It’s been a rough few years for PR and marketing. If we can point employees in the direction of work that makes us more money, and, between us, automate the aspects that don’t, I’m absolutely going to do that.

I’ve seen two plays this week alone at introducing AI into the journalist discovery and pitching process: PRprophet.ai, and Propel’s Amiga.

I want to see demos of each, and am hesitant to entrust AI with aspects of PR I’ve only ever, of course, seen done well by human beings, but if I can build an automation stack that adds up to a more profitable agency, and there’s no loss of quality or results for clients, I’d have to be a fool not to consider it.

5. Content finessing

Whether it’s a tweet that’s a few characters over, a press release you’d like written as onsite blog content for a client or even blog content you’d like written into a LinkedIn post to share the link, I’ve found ChatGPT useful.

Spell out what you need it to do, and it’ll do it.

This LinkedIn post promoting our next High Flyers PR Course intake of students was written entirely by ChatGPT, for instance.

I then asked it to shorten it to fit into a tweet.  Here’s what it did with that:


We’re now at the point where we can move beyond mere passing interest, and start to look at how we can implement AI in our client work.

I’m most interested in how and where we do that, and will continue to tinker while we all make sense of it.