In this post we discuss the shift in traditional product development when working with the uncertainty and capability of an LLM There are well-establi

AI Product Development Lifecycle: tackling uncertainty

submited by
Style Pass
2024-11-27 17:00:06

In this post we discuss the shift in traditional product development when working with the uncertainty and capability of an LLM

There are well-established and effective frameworks that companies adopt when building and shipping software. Whether it’s an early-stage startup, scaling startup or a larger enterprise, there will be processes in place (or not!) to ensure cohesion and efficiency. While these processes will change company-to-company depending on context, the fundamental goal should remain: ship high-quality software that impacts products and customers.

Discovery: Deep dive into the problem. Start to design and spec out a solution. Often here a PRD will be finalised as well as product design and technical specifications / design.

In this traditional product development cycle, teams have well-established pipelines and processes in place. Code is written, tested though unit and integration tests and deployed to non-prod and prod environments through tried and tested CI/CD pipelines that have now become industry standard. The point here is that teams have significant control over the context, and with that comes an element of certainty and confidence - you can define the inputs and business logic, and to a large extent control or predict the outputs with a high degree of certainty. Even from a design perspective, designers have confidence in the space and content they are designing with. When bugs occur, they can be reproduced reliably (most of the time!), hot-fixes applied and shipped. This level of predictability has become the norm and at the same time fundamental to how we’ve approached product development, enabling teams to ship with confidence through the use of having controlled datasets and environments.

Leave a Comment