AI is Not the Panacea for Performance for Government IT Managers
By James D'Arezzo
Here we go again. Government IT managers (and for that matter, Corporate IT managers as well) are being asked to do more with less. Indeed, at the Federal level IT directors are being mandated by the Federal IT Acquisition Reform Act (FITARA) to reduce their budgets by as much as 25% in the current fiscal year."
The latest perceived "solution" to this problem is the broad adoption of Artificial Intelligence to reduce costs, improve performance and provide better service. No question, when properly implemented AI can do all of these things. If a piece of technology can reduce staff workload or walk citizens through a routine process or form, so the thinking goes, it could enable management to effectively multiply the output of a workforce without having to add new people - a clear cost/performance benefit. However, the reality is far from the perception. A recent white paper from the Harvard Ash Center for Democratic Governance and Innovation, in fact, finds that AI could reasonably be expected to be helpful to government in only a handful of areas-resource allocation, large data sets, expert shortages, predictable scenarios, and procedural and diverse data. The author of the study, Hila Mehr, notes that most of these are big-data applications in which sophisticated algorithms, i.e. artificial intelligence tools, are already routinely employed.
And, there are problems. A 2016 report by ProPublica, for example, found that an algorithmic system for criminal sentencing was biased against black people-not by understanding the color of an individual defendant's skin, but by using flawed data correlated with race. Recently, in the Houston, Texas Independent School System, where teachers' job performance is evaluated by an algorithm, a circuit court has found that the software violates the teachers' 14th amendment right to due process. Meanwhile, the very companies fostering the development of artificial intelligence are struggling with some of the unintended consequences of its application. Both Facebook and Google, whose algorithms are given autonomy over what users see on their respective sites, are entrenched in public battles over misinformation and propaganda gaming their systems.6, 7 Thus, there are unknowns that should be considered as the nascent AI solution set comes into the market.
So, what other options do government and corporate IT managers have to improve performance and lower costs? Certainly, the idea of a "forklift" upgrade of hardware to the latest all-flash storage arrays, hyper-converged systems or wholesale client/server enhancements are out of budgetary bounds for many organizations. There are alternatives to extending the life of systems and getting more performance at the same time. The ideal solutions would be non-disruptive and well tested. One example that can significantly improve performance of existing hardware and prolong its useful life is I/O reduction software (after all, if I/O to storage can be reduced by 50%, performance goes way up). Other enhancements include a modest investment in DRAM on key systems to better leverage I/O reduction software instead of going down the very expensive and disruptive path of a complete re-architecture of the storage backend. A third approach is more disruptive, but a move to cloud based environments can ensure the fastest configurations that can be guaranteed by Quality of Service contracts. There are drawbacks to this last approach as network bandwidth and security issues can be a concern.
The bottom line is that IT managers are constantly asked to do more with less, so it behooves a professional to consider all of the options prior to making the leap.
About the Author:
Jim D'Arezzo has had a long and distinguished career in high technology. First serving on the IBM management team that introduced the IBM Personal Computer in the 1980s, he then joined start-up Compaq Computer as an original corporate officer and helped the company grow to over $3 billion as VP Corporate Marketing and later VP International Marketing. Seeing the technology trend toward networking, Jim joined Banyan Systems in the early 1990s as VP Marketing and helped that global networking software leader grow rapidly and eventually go public on NASDAQ. He then moved on to computer-aided design software leader Autodesk as VP Marketing and multiple Division GM for data management, data publishing and geographic information systems. D'Arezzo later served as President and COO for Radiant Logic, Inc., the world leader in virtual directory database solutions. Jim holds a BA from Johns Hopkins University and an MBA from Fordham University.