Katherine Johnson is director of emerging technologies at IT solutions provider Trace3. A featured speaker at the 2019 MIT Sloan CIO Symposium event in May, Johnson joined CIO Vince DiMascio of law firm Berry Appleman & Leiden; Bill Krakunas, national management consulting leader at audit, tax and consulting firm RSM US LLP; and Shane Jason Mock, vice president of research and development at American Fidelity on a panel about automation and AI.
In this video shot at the event, Johnson talks about some of the chief automation and AI challenges CIOs face today. Data quality is a "huge" problem, she said, but so are human workforce issues that arise when implementing AI and automation.
Hear her advice for overcoming these AI challenges.
Editor's note: The following has been edited for clarity and length.
What are some of the chief AI challenges CIOs face?
Katherine Johnson: The machines have to learn, just like humans learn, and the data quality is huge. But it is not just the data quality that is the problem; it is bringing the data together from all of the different systems that you're looking to learn from. So, you almost have to go back and rethink the way you're integrating the data, because it is not the way you did it before, if you want that data to [be acted on].
So the chief [AI] challenges are bringing the data together, identifying that data, cleaning that data and teaching off of that data. Doing all that prep work forces you to change the entire stack that you previously had.
Another [AI] challenge is the human adaption. Oftentimes [with AI projects], instead of a bottom-up [approach], where you're engaging the different layers of humans that do the work and understand the why, [AI projects] tend to be top-down. The people that are more impacted dig their heels in, and you tend to find that 25% to 30% [get to] the implementation phase and then it stops.
How can companies get the AI naysayers on board?
Johnson: The prevention key is clearer business drivers and use cases -- with goals. [These enabling technologies are] very sexy. People tend to bring in these enabling technologies [like AI] because they read about it somewhere or they think it is going do something great. But that something has to actually be defined. So, clear business drivers, clear use cases with clear goals help push past that 25%. A lot of times that 25% to 30% comes from poor solution choices because there were not clear business drivers and clear use cases.
How big a priority is operational efficiency for companies today?
Johnson: If you had to blanket the No. 1 reason for automation or bringing in machine learning technology is for operational efficiency. It doesn't matter how you apply that: If you apply that to how you're protecting your domain, how you're upleveling your workers, how you're able to quickly turn around something for a client -- all of these [require] operational efficiency to get there. Some of that efficiency can be found in process retooling -- maybe the process needs to be challenged, but for the most part it's all operational efficiency…
It's an odd paradox. We've continued to add emerging and innovative technologies, but it hasn't actually made us more productive. In fact, as a workforce, it's kind of made us less productive. So bringing in automation at this time is bringing together these systems to regain that productivity.
Is there a final point you'd like to make about AI and automation in the enterprise?
Johnson: If there is any one … takeaway, bringing in enablers like AI and machine learning and automation, one day that will be invisible to us. It's scary and new and we're not sure how to implement it, but at some point in time it will just become invisible to us because it is a part of everything we do. So an automation or machine learning strategy really should be about how you're going to uplevel all of your workers and not just because it is new cool or fascinating -- it will just become a fabric of your life.
How well do you think companies understand that point?
Johnson: They don't. They don't! I think [companies] form a lot of committees on how is the best way to bring [AI] in, and then it is indecision by committee, as opposed to looking for those quick wins … Companies tend to look for areas of pain instead of areas of gain. And [areas of gain] is key to [AI] success. Don't try to fix something that is a huge area of pain; start with the areas of gain.