As a companion to this blog post, I've also created a video that demonstrates some of the concepts discussed here. I encourage you to watch it alongside reading this article for a more comprehensive understanding.
Kent Beck's Insight
Here is a tweet by Kent Beck (a programming guru) that I really liked because I think it's onto something:
His conclusion: "I need to recalibrate." It is this recalibration that I'd like to discuss because I think it's relevant for a lot of professions, not just programmers.
Impact on Developers
First, I'd like to talk a little bit about what this means for developers, as I think it's a very clear case. I'll use some illustrations from my colleague Henrik Kniberg for this. If you look at developer time split before generative AI, they spent quite a bit of time understanding the problem, and then of course writing and debugging code was quite a big part.
Now, let's look at what it means to develop in the age of generative AI. This is just a very simple example. I don't write as much code anymore as I used to. Instead, it's now possible to just ask AI to write programs for us. For instance, we can ask Claude to create a simple snake game, and then add rainbow colors to it (I show this in the video). Claude will quickly generate the code for us. In the past, just thinking about the problem and then writing code took a lot of time. But now, this process has become much simpler. The code is done by the AI. In reality, if I was going to actually deploy this, I would have to copy-paste the code into different files and work a little bit with it.
New tools are emerging that allow us to use AI right in the middle of our development environment. There's one tool that I love called Cursor that lets you do this inside your IDE. Now, understanding the problem still takes as much time, but the other tasks have really shrunk.
You have saved a lot of time.
Hulten's Theorem
This shift in productivity reminds me of Hulten's theorem. It's from an economist who basically says that in the overall economy, if one sector improves and becomes better, then the cost of producing things in that sector will reduce. And even though it produces more, it will be a smaller part of the economy because the other parts of the economy will increase and take a larger share.
To get a better understanding of what this means, we can look at some previous examples:
- Agriculture: If you go back 5,000 years, what share of the economy do you think was agriculture? Probably more than half. Now, it's almost nothing, like a handful of percent in advanced economies.
- Manufacturing: In the industrial revolution, a lot of people worked in manufacturing. Now, that has also shrunk.
- Manual computation: In the early days of computation, when bureaucracy came and you had to calculate, there were a lot of people sitting with abacuses, moving balls on a frame to calculate things. That doesn't exist anymore. We just use computers for that.
There's also a more granular version of this, where instead of looking at the whole economy, you look at one company. Let's say before there's a productivity improvement, there are two activities A and B. Now activity A becomes much more efficient, then we have the same effect, but on the activity within the firm scale instead.
So, the effect looks something like this:
- Productivity increases in certain tasks
- Relative importance of highly productive tasks decreases
- Focus shifts to tasks where productivity gains are harder to achieve
What this looks like with Gen AI
So what does this mean for generative AI changing how we work? Some basic programming, writing, doing different kinds of repetitive tasks or simple data analysis becomes more and more something that AI can do. That means we shift what becomes valuable to what AI cannot easily reproduce: the personal skills, the human connection, creative problem-solving, critical thinking, using your judgment. Those things became much more important.
Future work, like human-AI collaboration, is going to be crucial because if you can get at this productivity benefit, then you will really be a winner in this new economy. It also means that you need to tweak your skills to figure out what are the parts of your activity set where AI can really help. And of course, this also means that some sectors will change in scope and size, like agriculture and manufacturing before.
A Path Forward with Process Re-engineering
One thing I really recommend people to do, and I've done this a lot with many companies, is to roll up the sleeves and go in and look at a role (or a value stream or a process) and do a kind of "process re-engineering", with a Gen AI perspective.
First ask yourself: What is this role now? What are the main areas? In each of these areas, what are the activities that take up the most time?
Then say, okay, but how could I use generative AI or AI agents to save time on these activities? You're just thinking: How might we do this with Gen AI? First ask "How can I save time?", then flip around and say, "Wait a minute. If this is now simple, how can I create more value, what can I do more of?"
As you do this, think "4-dimensionally", including the time perspective. Think about what we can do already now with ChatGPT, what we could do with some other additional tools, and what is possible with autonomous agents. The future is coming at us at a fast pace.
Here is simple illustration of how to do this:
One time I did this exercise with a company can serve as an illustration of how impactful it can be to do this. It was a one-day workshop. In the beginning of the day, one guys said about the questions that ChatGPT came up with: "yeah, but I already knew all these questions", indicating he was somewhat sceptical of the potential of Gen AI. At the end of the day, when we had gone through all their process and made prototype prompts for the different parts, including some calls to an internal Gen AI tool the had (a RAG tool), he exclaimed "but what are we then going to spend our time on?". It wasn't that dramatic, there was still plenty of stuff left for them to do, but the change in his mindset illustrates the value of rolling up ones's sleeves and diving into concretely how Gen AI can be helpful.
Conclusion
I've shared these thoughts because I've been thinking quite a bit about this topic. I've been working with many companies to understand what this means in practice. And I'm starting to see that this is something that will become really important as time goes by. Doing this type of exercise across different parts of a business will be an important part of any Gen AI transformation work.
If you have any thoughts or questions about this, or want to get help with implementing these ideas at your company, don't hesitate to reach out. And regardless, I encourage you to think about this yourself and try it out. Thank you for reading, and don't forget to check out the accompanying video for a more interactive demonstration of these concepts.
