Starting an AI business without AI

Brian Hankey has access to a large dataset from his loan business, and was investigating how to use AI to draw conclusions which would be monetizable to other lenders.

At Source Summit AI, Krassimir (Casey) Paskalev suggestion focused on how this can be done quickly and efficiently:

Technically, the fastest cheapest easiest way [to extract conclusions from data] is that a data scientist has to sign an NDA and run with it for a day or a week. That will be cheap, that scientist will have lots of interesting insights and they can capture 90% of the business value that way.

Lyuben Belov, partner at Launchub Ventures, pointed out that Brian has to make sure the insights he wants to get from this data will be worth it before he makes a serious investment:

Get third party data, increase the data set, test the waters with potential clients, don’t build whatever without knowing there’s a need for it.

When you see a commercial opportunity for AI, it’s worth separating the value of the output itself from the value in automating the analysis. Especially at the early stages where the commercial value is unproven, human-produced analyses can cut down your investment and provide a stepping stone towards automation.