Follow Datanami:
January 26, 2018

Are Our Expectations For AI Too High?

Roman Stanek

(Vasilyev Alexandr/Shutterstock)

AI has been in the news for decades and its capabilities continue to become more exciting and futuristic. However, few companies seem to be taking advantage of everything that AI has to offer.

While reading a recent study titled, AI survey by MIT Sloan Management Review, I noticed only 14% of respondents stated they believed AI is currently having a large effect on their organization’s offerings. However, 63% said that they expect to see AI have an effect on their organization within the next five years. These high expectations held true across every industry, company size, and geography represented in the report, even though few respondents had actually seen any substantial benefit from AI thus far.

Given the current state of AI adoption, are these expectations too high? Quite frankly, I do believe that to be the case but not due to any fault of the AI technology. In reality, most companies are completely unprepared to begin implementing AI.

That said, if companies expect to see any measurable progress in their AI capabilities, they need to first recognize and start addressing the parts of their data “foundation” that are broken. Things like improving data quality and timeliness, effectively harmonizing data, ensuring data privacy and security, and having a strategy in place for the cloud are essential to successful AI implementation. These steps are integral if an organization expects AI to function the way it’s been promised to.

Clean data is a basic requirement for AI (Profit_Image/Shutterstock)

In a recent meeting with some of our customers, they shared that while they are very interested in seeing how AI can help their business, they recognize that they’re still fighting fundamental data quality issues. This is a perhaps the biggest hurdle for companies looking to successfully implement AI, as a machine is only as good as the data it has been provided with. To see the kind of results a company is looking for, it first needs to ensure the data is as accurate as possible.

While we already have the necessary tools to help organizations improve data quality and flow, realistically it can take a couple of years to be vetted, implemented, and actually begin improving the data. Time is of the essence here, especially because this is a crucial first step to implementing AI.

Once data quality has been addressed, then the focus needs to be on getting data to arrive in a timely manner so it can be harmonized to help create a solid foundation. Data that comes in too late becomes essentially worthless, and when it does come in, it’s coming from many different sources. Ensuring that a company has accurate, current, and properly harmonized data is critical.

With that initial foundation built, the focus needs to be on properly safeguarding data according to regulations and security standards while also efficiently storing data. Moving data to the cloud has a clear advantage here when you consider the sheer volume of data that AI requires; it would be impossible to acquire or build new data centers fast enough to handle all of it. A cloud storage solution is a scalable option that frees companies from having to choose between finding enough storage space and continuing to use AI.

Determine your storage requirements before embarking upon machine learning (robuart/Shutterstock)

However, most companies just aren’t following these entry-level steps. I often hear from our analysts that a large percentage of their calls are from customers who are just now asking about cloud storage. While this is an important start, there’s much more that needs to be addressed first. These companies are so far behind the bare minimum that AI needs to succeed, but are quick to blame AI technology instead of addressing basic foundational issues.

Deploying true AI and machine learning within an organization is a complex ask, from the skills required for deployment to the technology needed for workflow improvement and collaboration, and it needs to be treated as such. We need greater investment in our data foundation across the board if we expect to move toward digital decision making.

Going back to the expectations in the survey, very little will happen in five years unless we focus on building a good data foundation. Companies need to be cognizant of the issues that need to be addressed today and of the goals that need to be part of their overall long-term strategy to set themselves up for a successful AI implementation.

About the author: Roman Stanek is the CEO of GoodData, which he founded in 2007 with the mission to disrupt the business intelligence space and monetize big data. Prior to GoodData, Roman was founder and CEO of NetBeans, the leading Java development environment (acquired by Sun Microsystems in 1999) and Systinet, a leading SOA governance platform (acquired by Mercury Interactive, later Hewlett Packard, in 2006).

Datanami