Posted
Success is best measured by the real impact AI products have, solving user problems, and improving experiences. It is not how well we hit arbitrary deadlines with fixed-scope features. You could argue that this is beneficial for the delivery of any type of product, but in the case of AI, this is crucial. Predicting effort and impact can be especially tough. Locking in scope upfront also comes with huge risks. I’m not suggesting that you don’t need to plan but simply a reminder to put the right level of effort in this space, so that adapting to change does not become a wasteful option.
Looking back at our quarterly planning outputs, we’ve had to change direction more often than not. Sometimes it’s due to the usual challenges, e.g. dependencies to tackle between teams in a large organisation. But many pivots related to the AI world, driven by new insights, shifting stakeholder risk appetite, and rapid advancements in technology that opened up entirely new possibilities. However, our north star and strategy have held a steady course, and by allowing for roadmap flexibility has helped us to get value out the door (even when we needed to change which one to open).
Experimentation and learning are central to how we work, keeping up with the evolving landscape. That’s why we encourage reading, sharing, and discussing new studies and research. With a fair few PhDs on the team, staying up to date comes naturally to the team, but we also have internal showcases, specific Slack channels for research sharing and many huddles on topics that spike the team's interest.
We also aim to get products into pilot stages quickly so we can learn from real-world use and refine them based on feedback. It’s not about chasing perfection, but about making steady improvements that bring real value.
Posted