By Myles Suer
A few years ago, I asked CIOs about data science and it turned into a yawner of a discussion. However, in the last few years as chief data officers have made their mark at more and more enterprises, CIOs have needed to build their data chops. Given this, it was time to assess where CIOs are today.
To do this, I ran a #CIOChat on AI and Data Science. From this discussion, it was clear CIOs are spending more time considering the "I" part of their titles. For this reason, they get not only the business potential from driving analytical organizations, but also what gaps need to be fixed to deliver it.
Where do You See the Biggest Promise in Your Industry?
CIO Wayne Sadin started the conversation by saying “we need to make our employees smarter by helping them quickly zero in on making better decisions, especially when the data is uncertain or incomplete.” He went onto say that “because AI is good at pattern-matching and remembering lots of facts, it seems reasonable that it can help most organizations.” To succeed, Sadin wants “software that can make inferences, detect patterns, and identify anomalies.” Personally, Sadin sees “AI as augmented intelligence that can make employees and clients (appear) smarter.”
For these reasons, former CIO Tim McBreen claims “pretty much every industry needs each to perform thinking tasks while freeing up staff to come up with new value based information that can come from the current data as well as collected data through AI or ML. This will be huge in distribution, logistics, transportation, etc. A number of companies have already made it part of their DNA. Where we need to see it grow is in financial and insurance services beyond traditional warranty related services.”
Meanwhile, CIO Jason James claims “we will see advancements in patient monitoring in healthcare from advancements in AI. Areas such as wound monitoring, blood sugar predictions, and infection prevention are already seeing solutions brought to market.”
CIO Milos Topic claims that “AI will initially see success in repetitive, high volume requests that would save everyone's time and enable people to focus on more creative and innovative things.” For higher education, CIO Carie Shumaker favors AI for “answering high volume questions such as dates, deadlines, FAQs, locations, etc.
She sees value coming through real time, accurate answers, even off hours, with an interface that doesn’t judge you for not knowing the answer already.” CIO Paige Francis also sees potential for AI “around customer service, rapid responses, repeat transactions/processes, and equitable remote hands on experiences and visits.”
Concluding this conversation, former CIO Tim Crawford suggested, “we have long-since passed the point where managing data through traditional means is possible. Automation and intelligence are key. If you look at Amazon, machine learning is part of their core…not an add-on to consider. It’s part of their DNA to business operations and has been for a while.”
What Types of Problems are Best Suited for AI/ML/Data Science?
Crawford believes that “AI/ML/Data Science is broadly applicable across the enterprise. With the amount of data increasing, we can expect to see these technologies more widely used. Cybersecurity is a huge opportunity for IT.” McBreen agrees and takes a step further by “saying it might be a savior in cybersecurity. People can't keep up with the way it runs today.”
CIO Steve Anthanas also sees this opportunity for AI/ML/Data Science when he says, “for the problems where humans add too much bias or cannot process data impartially. The thing that really interests me about AI in cybersecurity is the real-time nature of correlating tons of data that no human could ever do. Think about crowdsourcing security data from trillions of transactions in real-time and applying that in your infrastructure. The thing that helps is that unless you give the algorithm the information it cannot see people. AI doesn’t see color, mannerisms, attire, etc. and doesn’t make decisions based on factors that aren't germane to the situation.”
McBreen, in contrast to Anthanas says, “we will have to watch for human bias also in the developers of the AI or ML engine. Either one can get way off track based upon bias and provide long range answers that are way off base. You need reasonability checks for both ML and AI. Meanwhile, James says listening to vendors and their solutions will solve all problems. Shoes not fitting? AI problem. Deliveries running slow? ML will fix it. Perhaps it's not the type of problem they are best suited for, but the leaders willing to make the necessary personnel/data changes.”
CIO Milos Topic likes these tools, however, for “larger volume, things that don't scale as easily with one-on-one type of support.” Sadin gets more specific and discusses problems with “large datasets and pattern recognition: helping people find the zebras ("when you hear hoofbeats, think horses not zebras" was part of medical education because a new doctor couldn't know enough about everything).”
As an example, former CIO Mike Kail sees the opportunity to use these approaches for financial behavior and transaction monitoring. Schumaker says “here there is too much data and too many data streams for a human to quickly and accurately assess. I think AI/ML often over promises, though-it can be dirty, biased and infer causation where none exists.”
Fawad Butt, Chief Data Officer at United Healthcare & Optum, says, “in his experience while AI is being thrown around as the panacea, it does have some useful applications. In pattern recognition, simulating complex environments, recommendations etc. AI is for real. To be clear, the algorithms aren’t new, but we have data and compute today.”
What are the Biggest Things You Need to Put in Place to Establish Proficiency at AI/ML/Data Science?
Sadin believes “AI isn’t good with bad data”. He, also, suggests that “data cleansing represents a good problem domain for AI.” Sadin goes onto say, he “agrees with importance of getting to clean, consistent data. To make AI/ML help us, we need to work with dirty data and identify, isolate, correct data issues? Intelligence should mean the ability to reason in the real world of inadequate/incomplete/inconsistent data.”
CIO Adam Martin agrees and says “dirty data inside of organizations is a huge problem for sure”. For this reason, Sadin suggests that “the elephant in the room is dirty data. That dirty data is our own fault. Much of that data comes from applications that we developed without sufficient consideration for how it's digital exhaust might need to be used in the future. Another problem that we created was technical debt! Much of what an organization classes as dirty data comes from data entering from outside the organization. Additionally, there is a lack of interoperability--semantic even more than syntactic.”
CIO Jason James agrees with Sadin when he suggests, “data is like oil in the sense; it is unusable unless it's refined. Much like oil, it is expensive to store when it is not used. Dirty data is extremely common across all industries. Do you want data equivalent of kerosene or diesel fuel? Both come from refinement, but you have to know what you want from data.” Sadin agrees and says “data, being like uranium, has a half-life.” Sadin suggests that “the ability to Identify the half-life for a class of data is an important skill.” James agrees and claims “there is a data lifecycle. Data lives, it's used, it ages out, and eventually, it dies. Retaining data forever can also come with risk in cases of breach or data exfiltration. As Tom Davenport put things several years ago, “you can’t be really good at analytics without really good data. (Analytics at Work, Tom Davenport, Page 23)”
Getting to Data Sufficiency
In terms of getting to data sufficiency, former CIO Tim Crawford says,” the biggest challenge isn’t in understanding the technology. It is in understanding your business. Understanding your business provides the necessary context to understanding your data.” For this reason, Butt says, “do the basic stuff, first. AI isn’t something a company typically does, it is something a company typically enables via data management, data governance, and other friction reducing approaches.”
This means that organizations need to think about their data processes. CIO Carrie Shumaker, for example, says,” create data definitions and clean data. Document your desired questions/outcomes. Clarify your privacy policies. Note that none of this is really technology.” Mc Breen suggests, “besides governance, you need to hire extremely good talent that truly understands all aspects of data science. Too many times, I see people that are barely good enough at data warehouse or reporting being thrown into this area.”
This means for James, “most organizations need to bring in new resources with experience to bear. This includes the ability to spend the time and investment in upskilling current staff. There is a limited talent pool for those with proficiencies in these technologies in most industries.”
What Problems are You Having Integrating Your AI/ML/Data Science teams?
CIO Steve Athanas says he “isn’t really close to this yet, but we are working jointly on a few projects with the high-level goals of getting better alignment between the teams. The biggest challenge now is aligning the shared goals of security and stability with apps and data consumption.” For Topic, he says that his “organization doesn't have teams dedicated to these initiatives yet, but he has clear understanding of expectation and communication are always of great importance.”
Continuing with this thought, James says, “many organizations haven't even gotten to that level yet. They are still in the phase of determining what issues they are trying to solve and which solutions best fit.” To get after this CIO Paige Francis says, “problems are training all to learn and embrace. For that, we need less specialized language and jargon. Talk about how the sausage is made in the kitchen please, not the dining room.”
Parting words – Upcoming Webinar
CIOs are clearly are clearly getting after data. There is work to do and partnerships to make data valuable. And data needs to be refined just like oil to have usefulness. With this, it is possible with the right top leadership to build a data-oriented organization. For many, this will be a big step forward. To get more perspective on data and analytics please join us for our panel of experts on October 2nd—Including myself, Tom Davenport, Marco Iansiti, and Dion Hinchcliffe – "Data Analytics for Competitive Advantage." Register here.
ABOUT THE AUTHOR:
Myles Suer is the Head, Global Enterprise Marketing at Boomi. He is also facilitator of the #CIOChat, and is the #1 influencer of CIOs according to LeadTails. He is a top 100 digital influencer. Among other career highlights, he has led a data and analytics organization.