Previously, I argued that generative AI is a boon for the mediocre worker, or more generally that people with a lower skill level could now perform tasks on par with higher-skilled workers when assisted by AI. Things move fast in the world of AI and since that article, a raft of empirical evidence has come along to confirm the theory. 

Frey and Osbourne who made headlines a decade ago with their prediction that 47% of jobs were at risk of automation, released an essay in September 2023 suggesting that AI will also make many jobs easier to do for people with lower skills” 

A study from Harvard Business School went further than this by testing a real intervention at the consulting firm Boston Consulting Group. In it, 18 different tasks were selected to be realistic samples of the kinds of work done at an elite consulting company. “[C]onsultants using ChatGPT-4 outperformed those who did not, by a lot… AI finished12.2% more tasks on average, completed tasks 25.1% more quickly, and produced 40% higher quality results than those without.” The study also uncovered another interesting effect that was increasingly apparent in other studies of AI – that  it worked as a skill leveller. “The consultants who scored the worst when we assessed them at the start of the experiment had the biggest jump in their performance.” 

This is great for businesses struggling to fill skills gaps because higher-skilled workers are the ones in shortest supply, while lower-skilled workers are much more abundant. There has not traditionally been a short cut to gaining higher skills. One had to put the hours in, be it years of experience on the job or in the classroom. Malcolm Gladwell popularised the notion that it typically takes 10,000 hours of practice to master a skill. The benefits of leveraging AI to complete higher order tasks are therefore manifold and include: 

  • time saved in mastering skill 
  • time saved in completing task 
  • increase in output quality 
  • lower cost of completing task. 

What’s stopping the diffusion of AI? 

If the benefits are so great, what’s stopping us? 

Diffusion of the latest technology into UK businesses is a perennial challenge for the UK economy and SMEs in particular. There have been various programmes aimed at achieving this. The biggest and best firms tend to adopt the newest technology but there is a long tail of underperformance. It is a key part of boosting productivity and with it living standards, but getting businesses to adopt new technology is not always straightforward. 

Richard Susskind* – author of Tomorrow’s Lawyers, tells of how in 1996 he suggested email would become the main method of communication between lawyers and their clients. Many saw this as laughable. Critics asked how such an insecure method of communication could ever work. They believed that he fundamentally misunderstood the lawyer-client relationship. Detractors were of course humbled in time. A similar thing is happening with AI today. While some people are busy finding faults, other are busy finding new use cases for their work. Businesses' responses to generative AI (notably ChatGPT) can be likened to the five stages of grief. 

  1. Denial – This is just a gimmick. Sure, it can write a Shakespearean sonnet about cheesecake but that’s not going to change the world. 
  2. Anger – It makes stuff up! And what about data protection! 
  3. Bargaining – OK this is clearly a thing. We had better come up with some policies to try and control its use. 
  4. Depression – I’m discovering new uses for this every day. What tasks will be left for me? Is the jobs apocalypse coming?  
  5. Acceptance – I’ll be mad if someone didn’t use AI to complete this task. We’re not made of money and our competitors are all over this thing.  

Most people lie somewhere in the middle, perhaps at the bargaining stage.  

Grounds for optimism 

There are reasons to believe that adoption of AI will be smoother than other new technology. Major barriers to the adoption of new technology include price, uncertain return on investment (ROI), and the need to upskill. All these barriers are low with generative AI. 

Low entry price 

It’s cheap. Dirt cheap. ChatGPT3 is free, and the premium version is £18 a month. The average hourly pay in the UK was £18.71 in 2022** (and this doesn’t include on costs of employment). Investment in technology doesn’t always happen because of uncertain ROI, but this investment will pay for itself if just an hour can be saved every month. For higher paid workers it’s even less time still. The ubiquity of this technology makes it democratic, which also makes it more difficult for employers to control. This also makes it highly disruptive. SMEs and individuals can access this technology which can pose a challenge to larger incumbent market leaders.  

Passive adoption anticipated 

AI is already being worked into many software applications we use every day (web browsers, Microsoft Office Suite, Gmail, etc) and will only become more pervasive. The process of adoption then becomes completely passive. Much like how we use emails without a second thought. 

Technical know-how 

Speaking from experience, learning to use new technology can be painful. The genius of generative AI is that it can interpret instructions in common English. In my own work, I have used ChatGPT’s code interpreter function to create outputs that would have taken months if not years to learn how to create using Python (coding language). It’s not unusual for someone in my field to lose half an hour trying to change something minor like converting the numbers on an axis from decimals to percentages. With ChatGPT, I simply ask the programme to do this without worrying about the code that sits behind it. Although it is true that over time you can get better outputs as you refine your prompts, the barrier to getting started is very low.  

Necessity the driving force 

Many businesses are worried about the risks of AI, but the biggest risk is likely to be a failure to adopt it. If the carrot of low price and easy use don’t work, then the stick of competition will soon force firms to adapt and adopt. Watch a content generator write a blog with and without ChatGPT side by side for example. The manual version takes 227 minutes compared to 20 minutes with ChatGPT. A 20-minute saving would have been impressive but to save 3 hours and 27 minutes is a full morning’s work. This person is bidding for work remotely using a platform. Anyone not using this tool will soon find themselves priced out of a job.  

Perhaps obviously, most people who have witnessed ChatGPT complete a task in minutes that would have taken days are convinced of its value and could not go back to the old way of doing things. Providing a safe space for this experimentation will lead to more lightbulb moments.  

General purpose 

I was surprised to learn that ChatGPT stands for Chat Generative Pre-Trained Transformer. I had assumed it meant General Purpose Technology. OpenAI – ChatGPT’s creators - were clearly being creative as this is no coincidence. A general-purpose technology is one like electricity or the internet. One which has applications for all sorts of industries. I use ChatGPT most days for tasks that are very specific to my job, such as cleaning up a spreadsheet to make it usable for analysis or troubleshooting problems with code. All these tasks were achievable before but were very time consuming (not to mention tedious). If I can find so many applications in my own role then there must be many more opportunities in other roles. With applications being so specific, experimentation and peer learning will be key.  

Sometimes businesses need a push to adopt new practices. Homeworking is a clear example. It took a pandemic for that revolution to occur, but the technology that facilitated it had existed for years beforehand. Status quo bias is a powerful thing. AI, however, is well placed to diffuse and to diffuse quickly into the economy. 

 

* Susskind also co-authored The Future of the Professions with his son Daniel Susskind. This book was prescient in that it considered the impact of AI on professional jobs at a time when commentary was focused more on manual tasks and automation. 

** Annual Survey of Hours and Earnings Table 1.5a, mean hourly earnings of all employees. 

About the author

Jon Boys, Labour Market Economist, CIPD

Jon joined the CIPD in January 2019 as an Economist. He is an experienced labour market analyst with expertise in pay and conditions, education and skills, and productivity.

Jon primarily uses quantitative techniques to uncover insights in labour market data, both publicly available and generated through in house surveying. Jon regularly contributes commentary and analysis of economic issues on the world of work to online, print and TV media. Recent work includes the creation of an international ranking of work quality, analysis of firm level gender pay gap reporting data, and an ongoing programme of work looking at the changing age profile of the UK workforce. 

More on this topic

Podcast

Ireland Podcasts

Listen to episodes from our CIPD Ireland Podcast Series on a range of topical workplace, HR and L&D issues

Listen now
Tools
AI in the workplace quiz

Test your knowledge of AI in the workplace with our quiz

Bitesize research
Factors influencing ChatGPT adoption in the workplace

What are the barriers to the adoption of generative AI tools at work and how can they be overcome?

For Members
Factsheet
Technology, AI and the future of work

How artificial intelligence (AI), robots and automation are shaping the world of work, the ethical considerations and the role of people professionals.

More thought leadership

Thought leadership
Critical role of HR business partnering in HR operating models

How can people teams balance line managers’ need for operational people management support while growing their team’s strategic influence through the HRBP role?

Thought leadership
Analysis | Quantifying the impact of generative AI on HR

We examine and outline recent research investigating the impact of generative AI tools on the HR profession

Thought leadership
What makes an effective HR operating model transformation?

We look at how investing in digital technologies, HR skills and culture drive success in restructured people functions

Thought leadership
How are organisations transforming their HR operating models?

We look at the main focus areas and share practical examples from organisations who are optimising their HR operating model