We talk about challenges in moving AI solutions from testing into production or commercialising AI solution for all users internall / externally. We haven't really focused a lot of attention on the other extreme: the data collection and developing the proof of concept.  

Where do you start ?  Ask yourself, what is the outcome you want to achieve? What is the business problem you are trying to solve? Not understanding this is the #1 blocker to success.  

Prerequisities: There are a lot of moving parts just to get started, some examples, but not limited to:  1. what data is needed & where are you going to get it 2. availabilty of data; do you have access to this data in a repeatable way 3. what about data privacy, white room (data can not leave the site) 3. how will the data be organized 4. who is going to own the data goverance 5. do the data scienists have the right technologies to do their job 6. is the organization ready for this to start  Additional topics discussed:  - How to priortize where to start - 80% of the work by data scientist is data rangling - Example use cases in the insurance sector - Example use cases in asset management companies - Proof of concept validates if the outcomes we want to or expect to see, can really happen - Establishing the architecture in a way we can scale if the POC is sucessful - Technologies data scientist should have access to  - Monitoring data drift effectively  - Different languages across the roles. Takes a lot of change management to get this right. 

Ultlimately is the organization 'data ready'?