Brief
Today, AI (Artificial Intelligence) is mainstream but still facing hurdles preventing companies from fully benefiting from the huge and increasing amount of data they generate.
Executives have clearly realized there are some really cool use cases and apps using AI.
They have heard about machine learning, they want to use it. Every company has data, yet they look at it and ask themselves how do we create value from that?
Luckily “ML Canvas” exists and helps a great deal working thru the hurdles (ML is Machine Learning, the less hyped terminology for AI, since machines are not that smart).
“ML Canvas” is a great workshop that will foster team common understanding and cohesive creation of the overall AI solution your company needs to deploy for building a competitive edge.
My recent “ML Canvas” workshop was very well received with a NetPromoterScore of NPS=+60 (Excellent) and is helping drive my customer AI progress forward. A great first step and anchor for their AI roadmap.
Details
ML Canvas introduction
First off, giving back to Caesar what belongs to Caesar, “ML Canvas” method was created by Louis Dorard, who has authored a book on the subject, that can be downloaded from his web site https://www.louisdorard.com/machine-learning-canvas. I highly recommend Louis Dorard’s books as they are very pragmatic with solutions clearly coming from working experience.
AI challenge
It’s too bad to realize most organization still fail to deploy machine learning. Some of the reasons behind this situation are the lack of a clear question data scientists need to answer. Also results produced are often not used by decision makers. The AI projects have as well been lacking domain expert inputs.
ML Canvas solution
The originality of the “ML Canvas” method is to bring data scientist, domain experts and executives together in one room during a workshop that facilitates the creation of a one pager.
This takes a team effort to cover all the aspects of an AI system as there is no one profile who can master them all. At the end of the session, the “ML canvas” describe all the learning bits of the “AI system” in this 1 pager: What data does it learn from? How does it use predictions? How do we know it “works”?
“ML Canvas” first step is key to elaborate the value proposition, who your customers are, and how do you translate this value into an ML task. This probably the biggest step and this way everyone on the team is align on the same goals.
It is also crucial how “ML Canvas” make the domain expert collaborate with data scientists and put all the high-level considerations in 1 place. Thus, domain experts are involved in using the outputs of the system, then, even better, partially or completely, with oversight, automating the decisions the system can make.
Finally, measuring how you achieve the value you propose is another key aspect “ML Canvas” helps formalize too. Your data, customer needs are living thinking. You need to have this feedback loop, a regular process where you review your system performance, adjust criteria slightly and improve its response to this ever-evolving context. Decision makers need be involved about the decision metrics that will drive how your data scientists optimize your models and how your business perform.
“ML Canvas” workshop feedback
Organizing an “ML Canvas” workshop was quite some preparation work, even for an experienced ML product guy like myself. Nonetheless I wanted to check if all this effort was as well received by my audience. It was with great pleasure I saw my survey’s results: the team involved in the “ML Canvas” workshop gave it an NPS=+60 ! (Given the NPS range of -100 to +100, a “positive” score or NPS above 0 is considered “good”, +50 is “Excellent,” and above 70 is considered “world class.”)
Comments