We hear a lot about how AI will automate 'routine, boring tasks' and take away the 'pain and mundane' from our work. But in the 18 months since GPT-4 came out, I've spent an annoying amount of time playing scheduling ping pong over email and doing other 'boring' tasks I thought would be automated by now. I'm sure you have too.
But something else has happened. I've found that GPT-4 and Claude 3 are surprisingly good at the 'higher order', cognitive aspects of my job. These tools are helpful for brainstorming ideas, developing strategies, or helping me write a persuasive executive summary.
This got me thinking – what would it mean if AI doesn't automate away the boring tasks? What if all the benefits are actually realised when we’re working alongside AI? That would be a pretty radical re-conception of GenAI and work.
In this article, we'll explore how GenAI excels at more complex, human tasks, how it can augment high skill roles – and how we should respond.
What AI can do
There is a growing body of research showing AI tools are quite good at 'cognitive skills'. Ethan Mollick summarises some surprising findings in a recent blog post:
Persuasion: AI is 87% more likely to persuade you to its viewpoint via debate than a human would.
Ideation: GPT-4 generates startup ideas that outside judges rate higher than those created by graduates from elite business schools.
EQ: GPT-4 out-performs human advice-givers and helps people work through tough emotional situations more effectively, empathetically and creatively than 85% of humans.
Empathy and judgement: Actors playing patients in a recent study rated AI as higher in empathy and better in judgement than human doctors – in fact, AI beat primary care doctors on 24 of 26 characteristics.
Honestly, these findings astonish me. It’s hard to comprehend that we are seeing performance like this across tasks that are complicated, cognitive - and sometimes, compassionate.
Don’t get me wrong. AI isn’t superhuman. LLMs make errors, may be rife with bias, and aren’t useful in some situations.
But the reality is we are at the point where technology is as good at some humans at real world tasks This has never been the case before. And that could have implications for our jobs.
AI overlap at the job level
Our jobs are effectively bundles of tasks - and these bundles have varying levels of exposure to automation from AI tools.
A growing number of research teams are trying to quantify exactly how much overlap there is between what AI can do and what humans can do. And it turns out that almost all knowledge jobs overlap in some way.
For example, a 2023 IMF Report looked at some of this research and concluded that:
“High-skill occupations, which were previously considered immune to automation because of their complexity and reliance on deep expertise now face potential disruption.”
“Jobs that require nuanced judgment, creative problem-solving, or intricate data interpretation—traditionally the domain of highly educated professionals—may now be augmented or even replaced by advanced AI algorithms...”
For example, the report considers the impact of AI on judges, noting that "because of advances in textual analysis, judges are highly exposed to AI – but they are also highly shielded from displacement" because of societal expectations and professional regulation. As a result, AI will "likely complement judges, increasing the productivity rather than replacing them."
What can we do to maximise the benefits of this powerful, potentially complementary technology?
How we should respond
We have powerful AI tools that could also help us work better and make higher quality things. Here are three ideas for how to bring this to life:
1. Amplification
First, we should look to where AI can amplify our existing knowledge and expertise.
For example, if you’ve practiced for 15 years as an Intellectual Property lawyer for FMCG companies in Australia, AI may be able to amplify these skills into other industries (e.g. Real Estate or Fashion) or other jurisdictions. It could, for example, provide tailored summaries of key issues in new industries, or tailor previous advices to new jurisdictions.
In this way, AI is a knowledge multiplier; a tool that compounds your expertise and effort.
2. Complementarity
In addition to amplifying our strengths, we might think about where AI can complement our weaknesses.
Consider creativity. I don’t consider myself to be a particularly creative person. I struggle to generate cool ideas on the spot - whether that’s coming up with a catchy business name or blog post title (can you tell?). But I do think I’m capable of identifying a quality idea or catchy phrase when I see one (don’t we all…) Generative AI is a great complement here. GPT-4o and Claude 3 are wonderful at producing a high volume of ideas, from which I can select using my own judgement, and then refine or build on them.
And so AI can both complement and amplify our strengths. What is the best way to make this a reality?
3. Attitudes
There are at least 3 important attitudes that we can develop to help us work with AI in a way that amplifies and complements our strengths:
Continuous Experimentation: Ethan Mollick suggests ‘bringing AI to every meeting.’ These tools are evolving so rapidly that we are all on the bleeding edge. You are likely the one of the first people to ever use GPT4-o level tools in your exact job for your exact tasks. So we must continuously test what works and what doesn’t, and build our own mental maps of how to work with these tools in our specific settings.
Authorship Retention: We must own our outputs. I think about it like this: “I used AI to write this article, but none of the sentences in this article were written by AI.” We might use AI to generate ideas, suggest phrasings, or provide research - but ultimately, the final product is our own judgement and our own voice.
Thought Partnering: Building on the above, we have to approach AI as a thought partner: a mental prosthetic or intellectual [10 options for this word] that pushes our thinking further. I’ve found that ongoing, conversational dialogues with AI help me flesh out ideas for my articles and work, identify weaknesses, and (hopefully) arrive at more robust conclusions. This is a shift in thinking - from AI as a tool to which we offload tasks, to a tool that is intertwined in our creative and thinking processes.
The way forward
Previous waves of technology promised to automate repetitive, boring, and dangerous work. We thought Generative AI would be like that too - but it has turned out to be something different all together.
What we have are a series of tools that are very good at cognitive work and higher order thinking. Instead of taking our jobs, they are likely to complement and amplify our existing skillsets.
But that won’t happen without adjusting our approach - we need to be continuously experimenting, retaining authorship, and treating AI tools as thought partners, in order to reap the rewards.
And all signs point to those rewards being pretty significant.
