How AI changes what we do, how we do it and who we do it with

Someone recently asked me how long it would be before we used AI-generated code in a production environment. The answer is that we already have, probably since the early days of ChatGPT. It will have moved through our rigorous change control and quality assurance processes and passed (or failed) like any other piece of code. The more significant impact I’m already finding is not just how individual writers, designers, or developers create deliverables and outputs. It is in how we scope, develop and deliver as a team.

One way I use AI is by doing rapid prototyping to explore a theory. For example, I ran a quick analysis to investigate whether there's any value in using machine learning to detect seasonality in transaction data with the aim of improving forecasting. The work I did was, ultimately, thrown away. However, the rapid analysis made it clear whether it was a worthwhile task for someone with better data science skills than me.

What proved to be a real lightbulb moment, however, was working on a website application to help an external audience visualise, understand and explore the insights from a research project. As recently as last year, we would have chosen an interesting but a relatively simple way to illustrate the findings – charts, tables or animations, maybe. A team would have set off to create and publish them. It might have taken a couple of weeks from start to finish.

This time, we were able to go much further, make it a dynamic and interactive way of bringing the research findings to life. Not only that, but a small team of two marketers and a developer sat down and delivered it in an afternoon!

The potential speed of delivery is exciting, but it's just the tip of the iceberg. We were able to go, maybe, 20x quicker. The team iterated in a hugely satisfying way, but it required much less in terms of resources and communication. The feedback cycles were dramatically shorter and more productive because we could prototype what we were discussing at almost the same time as discussing it. Not only that, but we also tried a range of different approaches and delivered a much better solution.

There's so much to unpack from just one afternoon's work. What was created would be just a prototype for some projects, ready for further development. In this case, it was entirely suitable for external use. Click here to explore the ‘workplace message assistant’.

What does this mean? It radically changes the cost profile of delivering and encourages people to work together. This new software development toolset presents us with an exciting proposition: Should we choose to do a task 'X times' faster, invest the time to do it 'X times' better, or some combination of both?

This experience has confirmed that these tools are both valuable and powerful, and it was exciting to see everyone thinking about what else they could do. It created a really strong dynamic to explore how we could solve the problem we had. As someone once said, the more we ‘mess’ around, the more we find out.

I think another major outcome is that the speed at which results can be produced encourages us to work synchronously. Whether in the same physical place or not, working on the same thing with other people simultaneously is hugely powerful. Previously, we would imagine something, then need to wait for others to try to create it and return with the results for review. That tends to push teams towards asynchronous working with lengthy hand-offs between tasks and minimal time collaborating and innovating. Not anymore.

The impact on how we create and deliver is dramatic. It allows us to put more emphasis on the thinking work. We can spend more time trying different options. As we embed AI tools in our developers' toolkit, we enhance human capabilities and unleash our ability to observe problems and challenges and ask ‘what if’. If even the most difficult parts of the process are 10x faster, it unlocks time to explore. And that's a fascinating prospect.