Running the first AI pilot project

KM AI course. June 15th. Session 11. Today we meet Nicole Talbot from Levatas and learn about running the first AI pilot project in an organization.

What is so important and special about the pilot? probably everything...

Expectations? sky-high. costs? They are always high, and in the case of AI.. no doubt.

And the risks? The risks are higher than in any typical AI project. Mutual learning is a process and that journey only starts with the pilot.

In one sentence: No place for mistakes.

Talbot shared with us many do's and don'ts. Many of them are logical and some relevant also to non-AI projects. For example- we have to set expectations and see they are not too high. The success criteria should be designed in a manner that indeed is measurable and relevant. Think carefully about the time and cost of the specific use case. Take into consideration that the project isn't running in a vacuum and interfaces should be set. and use human feedback to improve the model and its performance.

However, I wish to return to the starting point- Choosing the use case. 

In choosing where to start, I gained some new understandings:

We always speak about focusing on the organization's core activities and main challenges, when choosing where to start. Talbot spoke about three additional factors that should be considered:

1.  Good and sufficient available data;

2. Easy to achieve management buy-in to this specific use case; and

3. Choose a challenge that even if we solve part of the challenge, it be valuable.

WOW! what learning.

And every time I re-read these sentences, I think about another situation/project/context that these learnings are relevant to as well. Thank you.

 This post was initially published in LinkedIn