4 Accelerators for the AI Pipeline in Business
Updated: Jun 6, 2018
February 7, 2018
by Mark Stadtmueller, VP Product Development
I will date myself, but one of the first books I remember loving was "Four Stars from the World of Sports." Now also loving to empower decisions & innovation by unleashing AI for Business, I decided to try to title this blog similar to the title of that book and align to those four stars. Whether you call it “Digital Transformation” or “AI Pipeline” or “Data Science” there are basically four stages to generate business benefit from data:
2. Data Management
3. AI Innovation
The problem often is that each of these four areas take too long. Here is my input on making each phase quicker (with the corresponding sports star):
Deliver results before talking too much. Hank Aaron. Hank Aaron literally epitomizes talking softly but carrying a big stick. Too much of AI is over-hyped pie in the sky and where most public innovation on AI is specifically around computer vision, speech recognition, and natural language processing, it is difficult for a specific business to map that to their efforts on new products and services, better customer interactions, and improved ways to do business. Ideas to leverage AI in business come from people or from people after exploring data. But, often, how to link data to AI innovation is immature at best. Quite telling is an MIT Sloan article that points out: “Most organizations represented in the survey have little understanding of the need to train AI algorithms on their data”. Before hiring long and expensive digital transformation engagements, a means to rapidly explore business results from data needs to happen. At dis.ai, we use our AI Business Jumpstart to accomplish this within one month. If something takes longer than this, then it is too slow.
Integrated Big Data, Security, and AI in one platform. Roger Staubach. Roger Staubach literally did things the right way. He committed to his service before succeeding in the professional football. And doing this right may seem harder, but is simpler than doing this wrong. In a lot of AI initiatives the focus is all on the data model. While that is critical, business specific innovation usually requires business specific data and the critical importance of data governance, controls, and security. By performing AI Innovation within an integrated end to end platform that provides for the ETL, compliance, and proper big data handling and controls, making business meaningful results is greatly hastened.
Automation and Scaling in AI Model Building. Kareem Abdul Jabaar. The Sky Hook revolutionized and transformed basketball. But, not all businesses have the skills and compute power to revolutionize their business with AI the way Kareem Abdul Jabaar revolutionized basketball with the Sky Hook. One thing in AI seems to be holding true: “it seems like given enough data & compute we can design systems that can match or exceed humans’ capabilities at narrowly specified tasks.” But, for business, that means handling large data with finite access to compute and ability to build and select models specific to that business data. With integrated automation and scaling these roadblocks that slow everything down are removed.
Dynamic Learning in Production. Bobby Orr. Bobby Orr produced and for any AI innovation to produce for business it needs to be easily and rapidly deployed into a new or existing production operation that leverages that capability integral to that process. But, often, the process of transitioning a model from training and testing into production is time consuming and initially produces less than stellar results. The long lead time associated with editing a model and putting back into production impedes AI innovation’s ability to keep up with the speed of business. Dynamic Learning allows for production data to feedback into the training and testing rapidly allowing an AI model to converge on business results.
Those are my four accelerators/stars. What are yours?