Evaluating Best Practices Before Evaluating AI

Lyndsay Wise
June 19, 2019

Words From the Wise

Digital transformation is a key initiative within many organizations. Understanding how to enable an automated and intelligent organization with the goal of getting the most out of data assets seems to be a struggle for many companies.

As a former analyst and consultant who has worked for a solution provider for almost two years, I feel I have gained a broad perspective about where companies excel and where they go wrong when implementing analytics and data management projects. From the solution provider side, we can definitely enable organizations, but sometimes our push for innovation impedes our ability to step back and ensure that organizations leverage both the technology and best practices that will support successful outcomes in relation to their digital transformation efforts.

Organizations should evaluate the latest technologies as a way to build applications that are attached to business outcomes. Because of the constant shifts in trends, it can be hard to keep up. There are times when organizations and solution providers may become too focused on trying to be early adopters and innovators in the market and neglect to align key trends with what organizations need to address business pains.

A great example is artificial intelligence (AI). Many organizations want to take advantage of augmented or artificial intelligence as an extension of their analytics environments, but doing so successfully is a challenge that requires many considerations and resources, and may not be conducive to every use case. 

Here are some initial considerations:

  • Data volumes. For AI to be successful, a system needs to have enough information to learn from the models built from it. If an organization lacks data, building models to leverage AI becomes a wasted effort, as no valid outcomes will be generated
  • Data diversity. In addition to volume, a certain level of complexity is required so that several scenarios can exist that will create the diversity needed for successful and valid AI applications
  • Skill sets. If AI is new to the organization, new skill sets might be required. These can be acquired through outsourcing, hiring, or educating the current team. In many cases, AI skills will be developed more fully over time, but getting started might mean outsourcing skills at first, to assist in the development of preliminary applications while the organization builds internal skills
  • Preparing for bias. Due to existing complexities, several examples of bias have already been identified – whether based on demographics, hiring recommendations, etc. – and should be taken into account when evaluating AI. This evaluation should be continuous, to ensure that bias does not creep in to models and applications being developed
  • Overall value proposition. In some cases, teams will want to leverage AI because it is a trend – in the same way organizations have jumped on Cloud First when cloud was becoming popular – without attaching cloud adoption to business outcomes
  • Business outcomes and applications. The most important aspect is the ability to define a project's needs based on business needs and not due to innovative capabilities or applications being developed. Technology should always support the need to address business challenges and not the other way around

When organizations fail to evaluate all of the business and technological requirements associated with emerging technologies and newer applications within analytics and business intelligence (BI), there is greater potential for projects to fail. As organizations leverage AI to create augmented environments, there needs to be an understanding of the complexities that exist beyond traditional analytics and BI initiatives.

For more on cutting through the hype around AI, check out this SeachCIO article featuring insights from myself and other industry experts. 

How does poor data quality impact artificial intelligence?