This Insights post is a part of a series about the use and adoption of AI within the government. Read the other Insights posts here: “Yes. We do need Chief AI Officers“, “Managing for AI in the Federal Workforce“, and “AI Talent Isn’t Easy to Keep“.
Okay, dashboards aren’t actually dead. They are visually appealing business tools that will likely be around for a long time. However, emerging and dynamic capabilities within artificial intelligence (AI) threaten to the use of dashboards in the future.
As data scientists, we spend a lot of time in dashboards – building the ETL, writing (sometimes) complicated measures, and designing engaging visuals – which ultimately become simple-to-use tools that attempt to satisfy business needs. However, the visualization component of dashboards can also be their downfall. As an end user, it can be incredibly difficult to get desired answers from static visuals. And while the capability isn’t quite here yet, the use of artificial intelligence (AI) will inevitably change the way we interact with data and visuals. We will soon be able to ask AI to dynamically find the exact answers we are looking for and build corresponding visuals.
Before we get there, however, organizations need to ensure they have clean data. If AI cannot ensure the accuracy of the data it is pulling and deriving answers from, it will not be successful. Today, it is rare to find a data set that does not face data quality issues. Data scientists know too well the feeling of dealing with data entry issues, inconsistencies in the data, and incomplete data—all of which can have significant impacts on the quality and usefulness of a dashboard. Currently, data scientists spend significant time developing Master Data Management (MDM) solutions and cleaning data before it goes into a final report. MDM ensures a set of business rules for what your data means and how it is used, and writing code to clean up data allows dashboard users to have confidence in the final product and what it’s showing them.
There should be greater emphasis on the actual data collection and entry process. For example, organizations should build more rules around how fields are used and ensure that the data in their systems is high quality. In an ideal world, there would be no data clean up needed. More realistically, AI will be integrated with MDM tools and will learn how to detect data quality issues and deal with them. AI’s ability to understand the unique business rules of an organization, and how certain measures and KPIs are calculated, will also be critical. If AI accurately does these things, one day dashboards may just die.
