Thread by @_ganeshp

What did I learn from organizations getting started in #AI?

They start by not building, but by asking questions. (a thread..) (1/10)

What problems exist that needs to be solved?
Why can't this problem be solved by traditional software, Why do we need #MachineLearning ?(2/10)

The answer belongs to one in 4 categories:
Category 1) Need to process a lot of data and is not possible for traditional data analytics/BI methods to scale. Because you have to make many micro-judgements on pieces of information and you want to speed up decision making (3/10)

Example for category 1: You have new data from telemetry from your machines in the field and tech-support teams to troubleshoot manually is possible, but not efficient - lots of data, augment human decision making with machine learning/advanced analytics (4/10)

Category 2) Problems are predictive in nature - you are extrapolating into the future from what you have seen in the past - #Analytics is connecting the dots looking backwards, ML is learning from that dot to project forward - example: Forecasting kind of problems (5/10)

Category 3) Complexity of the problems are high - multi-variate - it's a the growing up moment with new data, when you realize that the same problems you thought you could solve, became harder due to new variables/ interdependencies eg: customer targeting, churn management (6/10)

Category 4) You have a core competency/SME and usually a lot of data (of customers, some behavior, transactions etc.) that you want to monetize - Data as business - you are building a new #dataproduct to expand into adjacencies, serve customers better (7/10)

Then there are the misc. categories - playing with ML, hired a data scientist, need to keep them productive; my CEO told me to get started with AI, and all of them could be triggers but try to answer the core question and you'll see it always come back to these. (8/10)

And post this, is when you'll start exploring ways to make it real - Do we have the right data, do we understand it, is the problem worth spending the time experimenting (as ML is experimental science), will the benefit will be worth it etc, we'll explore in another thread (9/10)

I'd like to hear what you see, what have you found working with organizations who are getting started with their #AI journeys? Would love to learn (10/10) (end)

@threadreaderapp unroll