The Work-GPT Paradox, AI and Job Crafting
Not that many of us are actually using AI at work, but we could make our jobs better if we did
We’re almost 2 years into the generative AI revolution, and the hype has held steady but there is a bit of a paradox: Despite the massive expectations of leaders and investors for massive, AI-led transformation, not that many people actually use generative AI. Recent research from Pew reported that only 23% of adults in the U.S. have ever used ChatGPT. Adoption rates do diverge sharply by demographic, with younger and more educated people more likely to have tried a generative AI tool. Still, taken altogether, just 20% of employed adults in the U.S. have used ChatGPT for tasks at work.
This is a starkly different picture of AI's impact in the workplace than all the buzz would lead you to believe, but there could be many potential explanations for this. It could be that large language models aren't as useful as we initially thought. While there certainly could be an element of over-hype, I find this explanation doubtful, as numerous experimental studies point to LLMs providing time savings and improvements in work quality across a variety of tasks (example).
Alternatively, it could be that many people prefer other AI tools such as Bing, or whatever Google's tool is called this week. This could be part of it, but data suggests that Americans are even less likely to have heard of these alternatives.
Instead, I would favor a couple of simple psychological explanations: Firstly, concerns about job loss due to automation could be creating trepidation. If I use this tool for work tasks, will I discover that I'm no longer needed? Concerns about negative impacts of AI in general, whether due to job-loss, or a lack of trust in the outputs, could be negatively impacting workers' overall perception of AI tools, leading to lower intentions to try it.
Second and more practically, many could be interested in using AI at work, but don’t understand exactly how it could help them. Knowledge about which tasks large language models are good vs. bad at, for example, is knowledge that AI nerds may take for granted but could really help the busy non-expert get the most out of available tools. Below, I will share a quick guide to help with this, but first I want to take a quick detour and talk about why more people should be using AI more and how it’s in their best interest.
Job crafting: We all have an opportunity to make our jobs 5-10% better
Dr. Amy Wrzesniewski, a professor of Organizational Behavior at Yale, was a graduate student when she did research about cleaning staff workers at a large hospital. The findings surprised her: even though job descriptions were standard across workers, the jobs people reported doing looked quite different. Wrzesniewski found that the most satisfied workers were crafting their jobs by taking on additional tasks that weren't part of the job description. For example, one cleaner reported helping elderly family members find their way through the complicated corridors of the hospital. Another worker made creating relationships with patients and providing them emotional support part of their job.
This coloring-outside-the-lines is called job crafting and it can come in many forms, including adding tasks that you see as being meaningful for your development, or taking the initiative to forge working relationships with people that you weren't necessarily asked to. A sizable academic literature has documented the benefits of job crafting - people who job-craft are happier with their jobs and they have better job performance ratings.
Dr. Wrzesniewski offers a couple of thought-provoking questions that might be interesting for you to think about:
What is the aspect(s) of your job that you find the most meaningful?
What could you do to make more of that aspect possible in your work?
Unfortunately, the second question highlights the drawback of job crafting - it's time consuming. It can be difficult for people who have limited bandwidth at work to begin with, or for working parents who don't have a lot of extra time outside of work. I think it would be easy for most of us to imagine how our jobs could be more meaningful and impactful, if only we had the time to pursue our true interests! That’s where AI comes back in to help us create some space.
A quick 3-step guide to identifying opportunities for micro-automation in your job:
In their book Reinventing Jobs, Ravin Jesuthasan and John Boudreau argue that our stereotypes about AI and automation don't hold up to scrutiny. They use the job of bank teller as an example, showing that while the ATM did automate some tasks, many other tasks of the bank teller evolved, rather than the job being displaced. Today, the job looks very different - bank tellers perform more complex tasks like helping businesses obtain loans, or selling sophisticated financial instruments to consumers.
Jesuthasan and Boudreau also offer some really useful insights on automation that I think we can apply on a smaller level to find our own opportunities individually. The goal here is micro-automation, to find a few tasks here and there that we can delegate to AI, so we can spend that time on more exciting and meaningful work.
A first step is to "deconstruct" your job into a list of tasks you are responsible for. These should be individual, discrete activities that taken together are representative of the work you do. The second step is to determine which of these tasks you could delegate to AI. See below for an example using ChatGPT as our AI tool and "data analyst" as our job.
We can see that there are several opportunities! How did we determine these? Well, keep in mind that ChatGPT is really great at generating new content very quickly, but may "hallucinate" or provide information that is not attuned to your context, or is even false. That means that ChatGPT is especially helpful for tasks like:
Brainstorming and idea generation
Writing first drafts of communications
Writing first drafts of code
Acting as a research assistant (check out Perplexity AI)
Debugging/troubleshooting errors
If ChatGPT can save us even 5% of our capacity, that's precious time we can use for job crafting, which brings us to the last step (really two steps) which is to take action: form a habit of regularly using AI to accelerate the tasks you've identified. Whatever your answers to Dr. Wrzesniewski’s two questions above, now is your opportunity to create more learning, meaning, and fulfillment in your work.