Part One: How I made myself redundant with AI
Welcome to the first PromptlyPR blog post, and the first in a four-part founder series by Victoria Conroy, Founder and CEO of PromptlyPR
In late 2022, like millions of awestruck people, I opened ChatGPT for the first time, and got frightened straight away. Those first fumbling, flummoxed questions of mine, tentatively typed into the chat box, were probably similar to most other first-time users:
What or who are you?
What is it you actually do?
Are you Skynet? (For the ‘Terminator’ fans out there…)
I’m no stranger to AI. Back in the 1990s, it formed part of my computer programming education but AI back then was a mere dot compared to the behemoth it’s grown into today. At first glance, ChatGPT was like a chatbot. Like the one my bank uses to gently usher me to the right department when I’m querying a duplicated transaction notification. But I also saw something that hit me like a lightning bolt.
As a financial journalist and editor turned B2B PR strategist, I'd spent more than two decades writing blogs, thought leadership articles, media comments and award entries for some of the biggest names in banking, payments and fintech, ghost-writing reports and speeches for CEOs, and editing various pieces of content from other writers.
And within a few hours of experimenting with ChatGPT back in November 2022, I realised something that was both exciting and deeply uncomfortable.
I realised I might be looking at the technology that could eventually automate a significant part of my own career. And would there come a point when my entire job would get automated away? Because if this technology was already capable of producing something vaguely resembling a blog post after just a few seconds, what would it be capable of in two years' time? Or five?
AI and my short-lived existential crisis
At the time, I was working in a not just fast but frenetically-paced B2B tech PR agency. My days revolved around creating content for banks, fintechs, payment providers and tech start-ups and scale-ups. Blogs, thought leadership, media commentary, award entries, whitepapers. CEO ghostwriting, industry reports. If it involved explaining a complicated technology to an audience of senior decision-makers, there was a good chance it landed on my desk.
I'd also spent years before that as a fintech journalist and editor, learning the discipline of verifying every data point and quote, questioning tone and narratives, and pushing back on outlandish claims. Ask any journalist - those habits become instinct after enough years. And the instinct to shine a journalistic harsh spotlight on people’s claims is difficult to switch off, even after I pivoted to the PR world where softer lighting is favoured.
So when ChatGPT and others like Perplexity and Claude appeared, I explored these new-fangled AI tools to see if we could incorporate them into our workflows. I wasn't looking at it through the eyes of somebody searching for a shortcut. I was looking at it with an editor’s eagle eye for detail, and sniffing it out with a PR director’s nose for a new angle. And after a few months of tinkering with these new AI tools, my initial trepidation shifted into bemused scepticism.
Amid the blaze of publicity that ChatGPT created in the first few months after launch, there were headlines, forum discussions and the inevitable LinkedIn influencer posts, all saying pretty much the same things:
“Content creation will never be the same.”
“AI will replace writers.”
I nodded in agreement at the first assertion, but scoffed heartily at the second one for lots of reasons.
Accelerated content does not equal accurate content
ChatGPT, Perplexity, Claude and other tools most definitely were changing content creation. What many people wrongly conflated was that accelerating content creation speed would also accelerate content quality, to the point where “AI will replace writers”.
Wrong. Laughably wrong. AI could write, for sure, thanks to ingesting the hard work of the human writers it was learning from. But after my first few content pilot tests using various AI tools, it was immediately obvious that they just didn’t know how to write well enough to be taken seriously in a niche B2B context like tech PR. I asked them to do a range of tasks with increasingly expansive prompts:
Write a blog…
Review this blog and include some recent statistics on [topic]…
Suggest a media pitch for this tech publication…
Shorten this newsjacking quote by 50 words to fit [publication] word count limit…
LLMs were successful in some tasks, and downright shocking in others. LLMs, in their puppy-like eagerness to please users by answering user queries at speed, began churning out poorly written AI slop gleaned from poorly written human slop already on the web.
When it came to long-form content, like blogs, or press releases and award entries, the output was outright diabolical. Far from saving time, LLMs were actually forcing me to waste even more time reviewing, rectifying, and in most cases completely ignoring their outputs.
There was no way I was going to let them loose on real client briefs.
The lack of writing quality was bad enough, but LLMs spewing out hallucinations, phantom spokespeople and fake stats was even more problematic.
In my industry - B2B content and PR - there is no room for error. Over my career, I’ve spent hundreds, probably thousands of hours, manually scouring bank annual reports, statistical bulletins, shareholder presentations and product tech specs for nuggets of data and trend narratives. I’ve worked with ambitious companies seeking funding and needing to position themselves for the next growth stage. When millions or billions of dollars are at stake, the pressure to tell their story convincingly is immense.
AI writing, human judgement and the gaping credibility gap
When your B2B content is flagged by media editors for containing LLM fake citations or sources, their credulity won’t just be strained - your credibility will go down the toilet. If your trade media op-ed is citing fake statistics presented as fact, forget about being published again. You’ll probably be blacklisted and banished forever - and possibly expose yourself to legal action. And if your investor presentation is riddled with AI hallucinations, you’ll be laughed out of the pitch room and you can forget about securing any funding forevermore.
AI writing had a massive content credibility gap. Because algorithms aren’t communicators. Professional human B2B and PR communicators, journalists and editors don’t just fill space with words that relate to each other and tell a bigger story. They exercise judgement, discernment and strategy, and understand the difference between thought leadership and self-promotion. They know when a stat is suspicious and when a claim overreaches. They know how to write something that journalists and target audiences actually want to read. Those are editorial and PR skills, not just writing skills.
So when PR agencies, comms teams and marketing people began to outsource their writing to LLMs from 2022 onwards, nobody could hold back the tsunami of AI slop that was soon to swamp the B2B media landscape. But to paraphrase Baldrick from ‘Blackadder’, I had an inkling of a cunning plan…
Coming next...
Part Two: How I accidentally replicated myself inside ChatGPT
I'll explain how life inside journalism and PR agencies, repetitive content work and broken production processes led me to reverse-engineer my own editorial and PR workflows. Plus, how my modular content methodology became the inspiration for PromptlyPR.