AI Product Development: The Foundational Framework

Main Talking Points

• Building software and services that are useful and people are willing to pay for is hard
• Building AI products powered by LLMs is exponentially more difficult because output from LLMs is nondeterministic and has exponentially more places for something to go very wrong as the tasks we want them to do become more complicated
• Building AI products is a new software development practice and we are in the earliest of days of figuring out what works and what does not
• If your team was excellent at shipping products and features using classic deterministic software logic, it does not mean that they will succeed when building AI products that are more powerful with nondeterministic logic branches and tasks
• Teams that were struggling in classic development with find themselves unable to deliver at all
• Anyone who tells you they have all the answers is not being honest. This is new for the entire industry, and we are learning together on what works and what does not.
• Those who will be succesful are those are are willing to experiment, move fast, admit when things are not working, and find solutions
• Even with all of this, there is a clear Foundation of proven tactics that should be the base process and cycle that your team uses to achieve delivery goals for products that work as expected and add value for you customers

code block

code

var a = 1;