Data democratization is a hot topic now. Spokespeople from Google, SAP, Capgemini and other tech companies have spoken and written about how making data available to as many people as possible will both unleash the power of technology and prevent abuse of closely held data. Microsoft CEO Satya Nadella sprinkles his talks and interviews with references to democratization. TBR agrees that data access is critical to achieving artificial intelligence’s (AI) considerable potential, but access is not enough. For AI to do what it can do, business people with domain knowledge, who are regular users, must be able to work directly with data, without intervening layers of developers and data scientists.
Data access is a conundrum. Ensuring appropriate privacy and security while making data available to as many people as possible is a challenge, one that is inhibiting the growth of widespread AI-driven data analysis. This post will not address that challenge, however. It focuses on one of the other growth inhibitors: the current need for data experts, scientists, engineers and janitors, as well as developers, to extract the value from data.
Business users might see the hierarchy of data experts as a priesthood or a bureaucracy, standing between them and the data, but that is not really what is happening. Currently, there are no tools with which business users can conduct their own analyses, at least not without a lot of preparation by the data experts. Better tools are coming; there are R&D efforts worldwide to make data more directly accessible, which is part of what Nadella and other spokespeople are talking about.
Before these democratic tools are made available, there is strong growth in AI and the utilization of data analytics, because the value is there. But the need for experts greatly increases the cost of analysis, so only analyses with the highest potential value are performed. As more democratic tools become available, many more analytic projects will be worthwhile and the use of analytics will grow much faster.
The impact of democratized analytics tools will be huge because the costs associated with the data expert hierarchy are great. Those costs go beyond just personnel. Communication between business users and data experts is time-consuming and expensive, and it lowers the quality and value of the analyses. Business users and data experts live in different worlds and have different vocabularies; misunderstandings are common. Misunderstandings are expensive, but what is worse, working through intermediaries slows the iterative process by orders of magnitude. The potential value of data lies in insights, and finding insight is an iterative process.
The history of business technology is a progress propelled by increasing democratization of tools. The PC itself is prime example, making computing directly available to business users. The internet rode a wave of disintermediation and self-service to its current global saturation. In terms of democratization of AI analytics, the best parallel is the PC spreadsheet, which made it possible for business people to create and tune their own quantitative models of business activities. Before the spreadsheet, creating those models required coding.
“Spreadsheets for AI,” one of which may well be a version of Microsoft’s Excel, will accelerate growth in analytics, data access, storage, cloud services and the Internet of Things. AI spreadsheets will not retard growth in the use of data experts; they serve a different market. Even with very good first versions, broad adoption will take years, so the acceleration of growth will not be sudden. Over the years, however, the ability of business users to directly analyze their data will contribute substantially to the revenue of IT vendors and to that of their customers.