This post originally appeared in Stephen Goldsmith's Better Faster Cheaper blog on Governing.com.
Recent advances in artificial intelligence have enabled many promising technologies, from sophisticated voice recognition to virtual personal assistants. While there is much excitement about how these technologies might lead to better public services, we will achieve more profound gains if we first focus on building governing intelligence -- the capacity for effective data-driven governance.
Just as artificial intelligence enables machines to be responsive in ways beyond simple pre-planned procedures, governing intelligence can empower governments to be effective in the dynamic modern era in areas ranging from customer service to responding to unseen terror threats.
We must recognize that intelligence and analytics are not simply pieces of code to attach ad hoc onto our governing systems. Instead, governing intelligence emerges from connectivity among civil servants, their information systems and citizens. As we build the capacity of public workers and increase their connectivity through better tools, we build governing intelligence.
Systems theorists describe a continuum of organizational intelligence using the acronym DIKW: Data > Information > Knowledge > Wisdom. Data in its raw form ("building address: 200 15th St.") supports information, including processed data and relations ("the owner of the building at 200 15th St. is the XYZ Corp.") Raw data and information, in aggregate, support knowledge ("20 landlords hold more than 150,000 of the city's 2.2 million rental units"), which over time and within the appropriate context leads to wisdom: "Lack of transparency in ownership records makes it hard to determine exactly how many units are owned by each entity, and the largest owners are not individuals but development funds and private equity firms."
This is just the sort of governing intelligence we have long needed. When I was deputy mayor of New York City, we sought to stop issuing building permits to developers that owed the city money. To check whether the entity was in default, we needed to extract data from the city's IT systems. Ultimately, we did not succeed at automating this process because the finance department and the buildings department were storing addresses in different formats. The structural differences in the data, along with missing links to metadata such as the relationships among shell companies, prevented us from making appropriate connections to act with wisdom in our permitting process.
Could we have built analytics to match addresses, check financial records, look for commonalities among shell corporations and flag deadbeat developers? At the time, the cost of doing so just wasn't justified in light of the other challenges we were addressing each day. Today, with the emergence or cheaper, more powerful tools to mine and organize data from multiple sources, the answer is different. Many municipalities are developing this very architecture as they build master addressing services to supply building street-location data to many applications simultaneously. To deny tax avoiders building permits, for example, tax-delinquency status could be read from such a master data service when a city employee is reviewing a permit application.
While new analytics might seem like the only solution to thorny problems such as deadbeat developers hiding among complex shell companies, a focus on foundational data systems over time will eliminate the need for much of this auxiliary analytics work and dramatically reduce costs and the timeline required to implement the analytics that will eventually need to be built.
Common data formats and unified data services lay a foundation for organizational intelligence. Sound data stewardship avoids duplication of data across systems and physically links disparate departments. Such a focus on data fundamentals prepares an organization to unlock the enormous potential of artificial intelligence. But for now, a connected and informed workforce is the basis for governing intelligence.