For many, the release of ChatGPT in 2022 was transformative.
Difficult emails were drafted in seconds, complex documents that would have taken hundreds of man-hours to scan were checked in minutes.
But for some, whose livelihood depended on providing just the type of service that ChatGPT offered - providing copy, for example - 2022 was a difficult year.
Chief Product Officer at Jacquard.
We were in the latter group. Our company, Jacquard, was founded in 2015 - seven years before ChatGPT hit email inboxes and LinkedIn feeds around the world.
Jacquard’s basic offering was a platform that used natural language generation to write better email subject lines. Around 2015, we tried taking the technology to investors. They were skeptical. Intrigued, perhaps, but skeptical.
At that time, we were one of very few companies in the language generation space, and we felt it. It wasn’t all bad though - and we eventually got used to being one of the few players in the game.
Then ChatGPT launched.
Our offering, AI-powered language generation, was suddenly in the hands of millions. For free. The investor skepticism we'd spent years navigating disappeared overnight, and in its place: a gold rush.
Hundreds of startups emerged, many founded by people who'd never worked in marketing or natural language processing.
On paper, this could have been the moment it all went wrong for us. Instead, if anything, it only propelled us further forward.
Surviving AI
It wasn’t an easy few years, but looking back, I can now attribute Jacquard’s survival to four key factors: AI fatigue; proprietary data; user base evolution; and an outcome-focused approach.
First, AI fatigue. When ChatGPT launched, everyone rushed to build, and then market, a GPT-wrapper. If it was a gold rush - to take the metaphor a little further - then too much gold flooded the market, and the value depreciated fast. Fatigue was everywhere, and it set in quickly.
Stakeholders grew tired of sitting through pitches that promised something revolutionary but delivered mediocrity.
Instead of a one-size-fits-all solution, what we all saw was generic copy, off-brand tone, and outputs that, at best, needed significant human reworking before they were anywhere close to usable.
85% of brands now use ChatGPT, in some capacity, to produce their marketing content. You can find the exact same turn of phrase in the marketing materials from a multinational conglomerate, as in your local café.
Our offering - the ability to maintain a genuinely distinct brand voice across complex segmentations at scale - became more valuable precisely because the market had made the problem worse.
Which leads to proprietary data. We have 60 billion data points built up over a decade of real enterprise campaigns. Far from being just another wrapper, we had ten years of learnings that no newly founded startup could acquire overnight. We trained a prediction engine on that data - and could back up our promises to deliver increases in engagement.
Of course, not every business starts from this position. But proprietary data, whatever form it takes for your industry, is worth finding. It’s likely more valuable than you think.
An evolving userbase
Our user base evolved alongside all of this. The client conversation shifted. Early on - pre-GPT - every sales meeting began with an education: what is natural language generation? Why trust an algorithm with brand copy? How does machine learning actually improve campaign performance? Post-GPT, those questions disappeared entirely.
They were replaced by a harder one: what makes your AI different from everything else? And so the bar moved. Clients came to us already exhausted by generic tools, already aware that AI-generated copy had a sameness problem. We no longer had to explain what AI was - now, all we had to do was prove that ours was worth the switch.
Last, and perhaps most important, our platform was never built solely around AI. AI was simply the most effective route to our solution. The businesses that folded had built their entire proposition on the novelty of AI. Once it wasn't, they had nothing.
We had been working on the same fundamental problem since 2015, and the arrival of large language models, ultimately, didn't change what that problem was. It just changed the tools available to solve it.
What we know now, having been through the full cycle - from obscurity to gold rush to consolidation - is that the brands who came out ahead weren't the ones who adopted AI fastest.
They were the ones who were clearest about what they were trying to say, and had the infrastructure to say it consistently. That is a harder problem than it looks. It is also the one we have spent ten years solving.
We've reviewed, rated, and ranked the best small business software.
This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit