Virtualization Technology News and Information
Creating a Data Strategy for the Future: Why AI Is Only as Good as Its Data Strategy

By Thomas LaRock, SolarWinds Head Geek

As I've written in the past, artificial intelligence (AI) is a broad term: ask five different people and you'll get five different definitions.

At its most basic level, if you've ever written a piece of code with IF-THEN-ELSE logic, you've written AI. Any computer program following rule-based algorithms is AI. If you ever built code to replace a human task, then yes, you built AI. However, AI can also mean more advanced applications like image recognition, text analysis, speech translations, and even predictions-based outputs like expected loan default rates.

What binds all these use cases together is data. And while AI is likely to have a profound impact on all our lives in the future (nearly every product and business process, and almost everything we buy will be touched by it in some way) no business can afford to implement it (or any other emerging technology) without a solid data strategy in place.

At the end of the day, we're all on the path to adopting AI. Like I said, you can't afford not to. The question is really, "how quickly are you moving down that path?" Is it a walk, a jog, or a full-on sprint? And how are you preparing your data for its integration? If you implement AI with bad data, it'll cost you. Therefore, before anything else, have an effective data strategy.

Here are some quick tips to help you while you're still at the starting block.

  • Your data strategy can't be an island. The data strategy underpinning an effective AI-enabled project must have appropriate buy-in and approval from key business stakeholders. Specifically, as companies continue exploring the value they can extract from existing data, they should look to involve a chief data officer as a key inflection point. The CDO (or other top data manager) should be the arbiter of the organization's "single point of data truth" to ensure data is properly governed and sufficiently approved for use in any project (AI-based or otherwise). At the end of the day, the output is only as good as the data that goes into this type of data-based project, so creating models and insights based on customer data without proper permission, for example, won't do anyone any good. Delivering business value from an AI-based project requires the implementation of strict data policies and strategy at the outset.

  • Meet with your team. Before determining which project to work on, meet with your data and analytics team leaders. They should know the emerging technologies and be able to assess how machines can solve business problems. Your Chief Data Officer will need to communicate with all stakeholders to figure out your single source of truth, or SSOT. However, data rarely resides in a single master file, but in multiple information systems. Deciding upon the SSOT, therefore, isn't always easy.

  • Make results manageable and measurable. For a first-time effort, it's best to ensure the stakes are manageable while you work out the structural and managerial kinks. There should be metrics to gauge success or failure, because deciding a project "feels" like a success is a timewaster. Don't go on instincts-go on metrics. At first, the conversation can be open-ended, like: Wouldn't it be great if X could happen... It should end up at the point where you decide on something more specific, such as using insight gleaned from AI-based models to improve customer segmentation by a factor of 10.
  • Remember your ethics. Many businesses don't have explicit permission to use customer data in their AI projects. Part of your data governance will be to ask for this permission. While you may be able to get away with using your data any way you'd like (in the U.S., data is still mostly the Wild West) ethically, you shouldn't. Let's be honest here-AI scares people. They know what a Ring doorbell is and what it records-for better or worse. They realize how pervasive facial recognition technology has become and its implications for personal privacy. Don't damage your brand by exceeding boundaries.

At the same time, we're at a crossroads for data privacy laws, which means if you go forward on a project with specific permissions at the outset but then look to leverage other aspects of a data set, you're either out of luck or in an ethically dicey situation. It's therefore vital to practice good data governance. Good data governance also calls for your CDO to monitor your data for validity and bias on a continuous basis. There should also be a system of checks and balances, so the data is kept as clean as possible.

  • Don't forget security: Data is the fuel for AI-and if the data isn't secure, the probability you'll experience a breach is high. According to the World Economic Forum's 2019 Global Risk Report, cyberattacks are a major worldwide threat. Adopting threat monitoring and detection tools is a key best practice for effectively managing and protecting a tech environment. Regardless of the type of threat, having the right IT security solutions in place to combat these issues, like automated response capabilities on threat monitoring tools, can help tech pros respond faster to events and better protect data.

And Finally... Give It Time.

The C-suite may want this project completed by yesterday, but that's not possible. While the execution phase has sped up in the last few years, the discovery phase will still slow you down.

You'll still need qualified personnel on board. You'll need clean data and permission from your customers to use it. You'll need a workable plan with real metrics. You'll also need to make sure all the stakeholders are on board. These matters (and many others) will take time.

But still: get started.


About the Author



LaRock has over 20 years of IT experience holding roles such as programmer, developer, analyst, and database administrator. He is a Microsoft Certified Master, VMware vExpert, Microsoft Certified Trainer, and a ten-time Microsoft Data Platform MVP.

LaRock has spent much of his career focused on data and database administration, which led to his election as a Technical Evangelist for Confio Software in 2010, where his research and experience helped create the initial versions of the software now known as SolarWinds Database Performance Analyzer (DPA).
Published Thursday, January 09, 2020 9:02 AM by David Marshall
Filed under: , ,
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
<January 2020>