Using an AI COE (Center of Excellence) to Bridge from Experimentation to Mastery - DataScienceCentral.com

Using an AI COE (Center of Excellence) to Bridge from Experimentation to Mastery - DataScienceCentral.com

Summary:  There is now sufficient experience among mid and large sized companies starting their AI journey to identify a single best practice for moving from AI experimentation to scale-up: the AI COE (Center of Excellence).

If you are a mid-sized business, government organization, or educational institution chances are pretty much 100% that you already have AI somewhere in your company. 

AI is embedded in so many modern applications that some version of NLP, machine vision, text recognition, or a recommender or behavioral predictor of some sort is already hard at work without you’re really having to do much beyond the original customization and implementation. 

Or it may also be that you’ve intentionally entered into the phase of experimentation to see what AI can do either on an enterprise-wide basis or perhaps in only one or two business units.

Whichever way you got started, it hasn’t taken you long between these experiences, what your competitors are doing, and what’s in the business press to realize the potential.  But the question of how you get from experimentation to mastery can be daunting. 

Issues of strategy, project sequencing, ROI, workforce integration, data acquisition and management, IT infrastructure, and even reputational risk from some yet undiscovered bias or privacy violation need C-level direction, planning and coordination.

But as Deloitte puts it so well in some of their literature: AI isn’t something companies need to do.  It’s something they need to be.

The Biggest Issues to be Addressed

No question there are some big issues to be bridged here over a period of years.  Those include an integrated IT infrastructure and data acquisition and management plans.  Don’t be deterred.  These shouldn’t stop you from moving forward and AI adoption doesn’t need to be terrifically front loaded with cost or changes to legacy systems.

A combination of the public cloud, a variety of AI/ML platforms, and intermediate solutions like data lakes can give you lots of hands on experience without solving those problems up front.

The primary challenge you’ll face in moving through experimentation to scaling up is your people.  First the issue of still relatively rare and expensive data scientists and data engineers who also understand both your industry and the processes you want to impact. 

Second is your existing workforce.  They need to be brought past the point of concern that AI is going to take their jobs to the new more important realization that people and AI can now work as a team to make your processes at all levels more productive. 

And finally, especially your management team needs to be sufficiently educated in AI so that they can recognize opportunities and effectively lead or participate in the implementation projects.

Current experience across a wide variety of industries and project types is that starting with a relatively compact AI COE (Center of Excellence) is the best way to grow during the experimentation phase and even as a method of scaling up.

Thomas Davenport wrote recently in the HBR: “…we believe that companies need to establish dedicated organizational units to entrench AI. This is an important business tool that cannot be left to bottom-up whimsy. Companies are devoting considerable financial resources to AI, and necessary skills and experience are too rare to assume that they will be scattered around the organization with little coordination or collaboration….In one recent survey of U.S. executives from large firms using AI, 37% said they had already established such an organization. Deutsche Bank, J.P. Morgan Chase, Pfizer, Procter & Gamble, Anthem, and Farmers Insurance are among the non-tech firms that have created centralized AI oversight groups.”

This means concentrating your AI resources within a single group.  Data scientists and data engineers are easier to recruit and retain in this structure where they can have adequate resources, recognition, and can share experiences more readily. 

At the same time the COE might also embed some of its members on a project or rotating basis directly in different business units where they can become more familiar with business methods and AI opportunities.

What Should You Expect From Your AI COE

If you are very early in your digital journey, you may not yet have formulated an enterprise wide vision or strategy for the role of AI.  You might use your AI COE to start here or alternatively you might choose to get some successful experiences first. 

In either case your AI COE should lead in identifying business driven use cases and guide you to an appropriate level of complexity based on your readiness.  They will always be your main cheerleading team for AI and should have a channel for advertising their success stories.

The AI COE is not only the core resource for implementing AI projects but also for educating your management team at large.  Most LOB units already have members who are AI-aware and anxious to participate.  These so called Citizen Data Scientist can be brought in on a project basis as valuable knowledge resources and then go back to their units as better educated AI evangelists.

While direct participation in projects is enormously effective as a training tool, you may want to more rapidly spread the skills and knowledge needed to identify and prioritize AI opportunities. You may then want to direct some of the AI COE resource to delivering a more formal training course to your broader management group. 

Your AI COE will either lead or be a key contributor in identifying and prioritizing projects.  They will also make those key initial decisions about common platforms, AI project management procedures, and begin to lay the groundwork for data management and model-lifecycle management.

The final critical goal is to create the working relationships and linkages with your legacy IT group.  The issue of how AI projects move from proof of concept to operations will vary widely by organization and by project.  It’s also a well-known source of project failure. 

As your organization grows more comfortable and experienced you can more knowledgably approach the issues of how much of the AI COE needs to remain separate and what parts might best live within your existing IT department.  Data engineering, the preparation and maintenance of internal and external data sources and the platforms to support them is an area that may well go either way depending on your particular circumstance.

When you reach the scale-up phase of adoption these AI COEs can be cloned and spun out geographically or by LOB as makes most sense in your business.  The best part is that you will have gained very cost effective and concentrated experiences to guide you to becoming a true AI-first organization.

About the author:  Bill is a Contributing Editor for Data Science Central.  Bill is also President & Chief Data Scientist at Data-Magnum and has practiced as a data scientist since 2001.  His articles have been read more than 2.5 million times.

Images Powered by Shutterstock