Trending Articles
Trending Topics
Trending
Businesses are rushing to use AI – but it’s not always clear for what purpose.
OpenAI has sought to answer that. Earlier this month, the AI firm released a study explaining how people are using the commercial version of its AI bot, ChatGPT, finding that 70% are using it for personal reasons and dominated by requests for information and writing assistance.
However, that only looked at consumer versions of the chatbot, rather than Enterprise users. A new report by OpenAI, ChatGPT usage and adoption patterns at work, looks to reveal similar information about how the AI bot has been picked up in the world of business. And this is no survey. OpenAI based the report by analysing actual user data – all anonymised and in aggregate, of course.
The first point OpenAI would like to make is that AI is being widely used for work – not just memes or virtual girlfriends. The report pointed to separate figures from Pew that showed a quarter of US workers said they’ve used ChatGPT at work, a figure that rises to 45% for those with graduate degrees.
That fast rollout – ChatGPT landed just three years ago, believe it or not – is down to people’s ability to sneak it into their workplace, an idea known as shadow IT or shadow AI, though OpenAI didn’t phrase it quite that way.
“Enterprise tech has always followed a familiar pattern: big upfront costs, long rollouts, and slow adoption before the payoff,” the report said. “ChatGPT broke that mould when people ported it from their personal lives into their jobs. They didn’t need months of training or complicated onboarding; they just started using it to get meaningful work done.”
According to the report, 27% of American weekly active users (WAU) on business subscriptions work in IT, followed by 17% in professional, scientific and technical services, and 10% in manufacturing.
There’s a wide range of other industries represented though, from retail to real estate (both at 3%).
“Retail, construction, transportation, wholesale trade, and agriculture all show significantly lower adoption,” the report noted. “In most cases, this tracks with their smaller share of knowledge workers, where the need for AI tools is less immediate.”
Across all sectors, for the first 90 days of a ChatGPT enterprise deployment, writing and text generation remains the most popular use case – echoing the consumer version. That’s followed by gathering factual and how-to information and computer programming.
Naturally, that depends on the role in question though. For IT workers, 30% of use was for programming, with 24% looking for how-to/factual information, and 20% for writing. For comms workers, 50% of their queries were about writing or text generation, followed by 19% for factual information and 2% for coding.
OpenAI noted that most people were using ChatGPT in fairly basic ways. “Advanced features remain underused, even where they could deliver broad impact,” the report noted. “Barriers may include discoverability, awareness of use cases, or the setup required to use them.”
Perhaps they could use ChatGPT to generate a how-to guide on using such features?
It’s worth noting that this is how OpenAI’s chatbot is being used – not AI more widely.
The small number of people using ChatGPT to code may be indicative of the fact that serious software engineers would opt for a dedicated product – perhaps GitHub’s Copilot or Windsurf or even OpenAI’s Codex – rather than a more generalist bot.
That said, it’s interesting to note that people are more likely to use it for basic tasks rather than more advanced ones. OpenAI argues that will eventually change: “As enablement programs expand and product improvements lower the barrier to entry, adoption will likely shift from core daily tasks toward deeper reasoning and collaborative workflows that reshape decision-making across the enterprise.”
The suggestion from the data – and from the need for the report itself – is that many are trying to figure out the same question when it comes to ChatGPT at work. What’s it for?
Right now, the answer seems to be drafting text and searching for information. Check back in a year, and we’ll see if OpenAI’s prediction that we’ll get better at using its bot holds true.
