The Journey to Enterprise AI
by Mark Stadtmueller, VP Product Strategy at Lucd
“but I believe that we, as humans, have an insatiable demand for products and services that will help us to achieve even more of our potential.” John Hagel
“Always watch where you are going. Otherwise, you may step on a piece of the Forest that was left out by mistake.” Winnie the Pooh
Enterprise AI is a unique subset of AI that is all about achieving new growth through responsibly learning from and leveraging data. While AI suffers from too much over hyped future capabilities described as if they are available today and very valid concerns about proper data usage, AI enabled products and services can safely improve our lives. Businesses will rightly create transformational value and define the digital future by combining their unique market knowledge with responsible data usage to offer new capabilities.
But, as Pooh says in the quote, businesses need to watch where they are going, because creating new value from anything including AI is “new” and without being thoughtful and careful, that “new” thing might just end up being more like “same old, same old”. This paper describes the phases of maturity that a business goes through to achieve Enterprise AI.
Phase 1: Plain Old BI
Phase 1 starts with where Businesses are already at. Business Intelligence (BI) is a powerful capability and the tools for BI are very impressive. In addition, Modern BI usually includes what Gartner calls “Augmented Analytics” link. Indeed, Modern BI platforms often work to back fill AI capabilities into its considerable heavyweight infrastructure.
When in the “Plain Old BI” phase, businesses only leverage data for traditional reporting purposes and with traditional tools from spreadsheets to expansive licensed reporting applications and web-based CRM tools. The process is somewhat regimented where data is fit into a BI application and reports are generated that often provide some insight but do not exactly capture the intent or intuition that the business seeks.
The purpose of Plain Old BI is to report on aspects of the business and in so doing, improve decision making. Plain Old BI is a business support function. A current view of AI usage in business is to improve decision making. AI can be leveraged to take BI analysis to the next level, but even when Plain Old BI is leveraging AI, it is focused on reporting. Enterprise AI is focused on responsible data usage and superior data analytics is a way that creates new things, a new way of doing business, a new way to transform your business.
Phase 2: Isolated Craft
High school statistics often now include R. Online data science, machine learning and deep neural network courses from edX, Coursera, fast.ai, and others utilize Python. Business publications are full of articles describing both the shortage of data science skillsets and the personal opportunities associated with people that obtain these skills. Because of this, there usually exists some people somewhere within a business doing something or looking at using something related to learning from data/AI. They can be formal machine learning teams; lines of business that have “citizen” data scientists working to improve themselves or areas they work; or people within the business who are reading about data science, exploring business data, and actively educating themselves.
When these types of efforts exist within a business, the business has entered the Isolated Craft phase. People or Groups within Business work on data to find new insights and potential new business outcomes. The people or groups leverage their own separate computer “beasts” or “closet clusters”. People leverage data they can get their hands on that often prove to be insufficient and there are little data controls. Interaction with lines of business is loose or informal and business-related results are isolated or random. Even if there is a new business outcome, there is a higher risk of bias, improper, or insecure use of data because of the lack of data management.
Because of the higher risks and lower chance of rewards in the Isolated Craft phase, if it does not yet exist within a business, this phase can be skipped. Businesses can migrate from Phase 1 to Phase 3.
Nevertheless, Isolated Craft groups can be a powerful force for change within a business. They have willingness and intuition on how transformational objectives can be created and achieved when leveraging data with the capabilities of AI. These people or groups also struggle through the challenges of creating and managing an end to end AI pipeline and associated infrastructure. In many ways, they are the “Enterprise AI” pioneers. If businesses skip from Phase 1 to Phase 3, it is still important to bring forward the “Enterprise AI” pioneer mindset.
Phase 3: Responsible Data Factory
Forbes points to a study back in 2016 where 80% of data science time is spent getting and cleaning data and as far back as 2009, data scientist Mike Driscoll popularized the term “data munging,” describing the “painful process of cleaning, parsing, and proofing one’s data” link. And back in December, pre-eminent AI leader Andrew Ng stated: “if you're a big company and you have 50 databases under the control 50 different vice-presidents then it'll be impossible if an engineer needs to combine data from multiple data silos to create value and if they need to get approval from 50 different vice-presidents it's just not going to happen” starting at 21:52 link.
The challenge doesn’t end with the above. There are.... Read more..