Why insight and analytics professionals aren’t doomed by AI

For years, many knowledge economy workers have felt a sense of unease about the prospect of artificial intelligence taking their jobs. But the threat was vague and non-specific, the realm of dystopian science fiction and not any kind of real and present danger.

Then, in November last year, OpenAI made its AI chatbot, ChatGPT widely available. GPT is short for Generative Pre-Trained Transformer, a Large Language Model (LLM) trained on the contents of the internet up to late 2021. Suddenly, the power of generative AI was in everyone’s hands, a tool seemingly capable of creating an infinite variety of content, form, and structure at the click of a mouse.

Microsoft incorporated ChatGPT into its Bing search engine and Office suite of productivity tools (Word, Excel, PowerPoint). And before Google could respond by rushing out its apparently ‘sentient’ chatbot, Bard, in March 2023 OpenAI launched its vastly-improved GPT-4. ChatGPT is the fastest-adopted software platform in history, reaching a million subscribers in just five days. There’s a good series of three blogs from the consultancy Ekimetrics over here, covering what generative AI is and why it matters, its strengths, limitations, and the opportunities it provides.

Use case after use case

Suddenly, knowledge economy workers felt a very profound shiver down their spines. Need to develop a training course? Ask ChatGPT. Write a blog on trends in commercial property? ChatGPT can do it for you. The next six months’ tweets / LinkedIn / Facebook posts on top tips for digital marketing? It’ll be better, faster, and better constructed by ChatGPT. Need to write a discussion guide for research about beer drinking in Latin America among 20-30 year-olds? You guessed it.

And it’s not just words. Need new images and illustrations? Try MidJourney. Design? stockimg.ai. Speech? Elevenlabs. Questions into data? usechannel.com. Automated translation of videos? papercup.com. Text summaries of YouTube videos? Glasp. Decision making? rationale.jina.ai. This Twitter thread might make you feel very queasy indeed.

Beware falling off the hype curve

Microsoft founder Bill Gates – not known for hyperbole – called AI “the most important tech advance in decades”, as revolutionary as mobile phones and the internet. Commentators were falling over themselves to predict the demise of almost every job in the service industry, with “the new printing press” becoming one of the most careworn-phrases in articles on AI. And investment bank Goldman Sachs predicted AI could replace 300 million jobs (but also boost global productivity by 7% annually).

Clearly, ChatGPT and thousands of other LLMs are causing a huge amount of disruption, uncertainty, and swirl. It is also – already – actively removing huge amounts of repetitive drudgery from knowledge economy jobs, as well as opening up new opportunities. This “Business Leader’s Guide to Using AI and ChatGPT” from the whip-smart team led by Alexis Kingsbury at Air Manual is an excellent example of straightforward, smart advice about how businesses can and should embrace the new technology. In a blog earlier this year, I dubbed ChatGPT “the perfect apprentice for the data storyteller” – but only that.

For the research and analytics community – data storytellers who blend the sometimes ‘fire and ice’ world of narrative and numbers, stories and statistics – AI can feel particularly threatening. This is in part because experts in these fields often work hand-in-glove with data scientists who are in the vanguard of developing and working with AI. And it’s in part because so many of the early use cases of generative AI have focused on simplifying data, coding, analytics, and research.

Where AI falls short for creative, insightful thinking

But I think there are several reasons why the reports of the death of human researchers and analysts have – to quote Mark Twain – been greatly exaggerated in the face of the arrival and rapid spread of generative AI.

  1. Although the current generation of AI tools appear to be hugely powerful, they work by boiling down the vast amounts of information they’re trained on to create a best-guess synthesis of that training data. As the Air Manual guide points out, “ChatGPT doesn’t replace creativity”.
  2. Creative thinking – including effective analysis of data and moving from data to insight – requires a combination of convergent thinking (making choices, at which ChatGPT excels) and divergent thinking (creating options, at which it’s pretty hopeless).
  3. As NYU professor Colm O’Shea wrote in the Times Higher in February: “Although not identical to creativity per se, divergent thinking is an important precursor to creative work … It proceeds via mechanisms such as deep pattern recognition and analogy (verbal, visual, mathematical) that software such as ChatGPT, which gleans a ‘gist’ from dizzyingly large datasets, is not good at.”
  4. Many users have reported ChatGPT and other AI tools “hallucinating” – making up references and sources – and while this has improved in GPT-4 compared with GPT-3, there is still a huge reliability issue that means all outputs need to be verified.
  5. ChatGPT is also hopeless at humour – at writing jokes – because of its divergent thinking shortfall. Try asking it to create a comic routine using a switcheroo – confounding expectations – and it flounders. It also has little idea how to create puns (I spent hours trying to get it to make up funny, sports-related cocktails, to no avail).

For these reasons and more, insights and analytics professionals should rest relatively easily in their beds. ChatGPT and the new generation of AI tools will help us all do our jobs quicker, faster, and better. But there are some essentially human forms of cognition that are highly prized in this sector, and – quick learner as it is – generative AI is still many years from these. As you were!