Automation 2015: How do you decide between algorithmic judgment and human judgment? You deploy WorkFusion.

Imagine a banking operation with 20,000 human data analysts, each spending 9 hours a day, 5 days a week hacking away at hundreds of terabytes of unstructured PDFs, emails, images, and websites in search of patterns and value. 50% or more of that collection, extraction and enrichment work could be performed better and cheaper by algorithms. How, though, does a data operations executive decide which 50% is optimal for machine automation and which needs human judgment?

Alan Turing: the father of computing

The machines vs. humans debate is picking up steam but it’s been around since the birth of computing. Christopher Mims published an article in yesterday’s WSJ about Alan Turing’s pioneering work in mathematical computing called, “Why Humans Needn’t Fear the Machines All Around Us.” According to Turing’s leading scholar, Professor S. Barry Cooper, the ‘father of computing’ knew even decades before the first circuit was wired that machine computing would change the world but that humans could solve problems that would always stump machines. Mims concludes the article with a vision that describes precisely what WorkFusion does: “The future of technology isn’t about replacing humans with machines, says Prof. Cooper — it’s about figuring out the most productive way for the two to collaborate.”

The key to optimizing machine / human collaboration is understanding both the ideal domain and the outer limits of automation.

One of the most common questions our customers ask about our platform is, “how does WorkFusion know what work can be automated and what work should be done by humans?” The answer took our product team about three years to actually build, but famed professor Andrew McAfee does a nice job of succinctly describing it in an HBR blog post called, “When Human Judgment Works Well, and When it Doesn’t.” His answer: by listening to the data.

An unabashed believer in the power and place of machine productivity, Andrew cites research that proved algorithmic judgment on data superior to human judgment in all but 8 of 136 cases. The best conditions for algorithms are the worst conditions for humans, where data is “often lacking, low-quality, or conflicting.” Algorithms excel over human data analysts because machines are more adept at detecting nuances in noisy data environments. In other words, the data typhoon environment that is most enterprise data operations.

The rapid feedback and learning loop is the second key advantage that algorithms have over human experts. Whereas human experts are disconnected from the impact of their output by space or time (or both), “well-designed algorithms can and do incorporate feedback and results over a long time frame,” which leads to better accuracy.

These 136 cases of fastidiously fed algorithms were executed by researchers whose academic goal was to prove the superiority of algorithmic judgment. In the age of aggressive cost cutting, how does a data operations exec far outside of lab research find the time and budget to deploy a team of data scientists, architects, and engineers to listen to the data, find patterns, and test and train algorithms to replicate them? The answer is never, because that kind of time and budget doesn’t exist at the enterprise level.

So the data ops exec shortcuts this expensive and time-consuming ordeal and deploys WorkFusion.

One of the highest benefits of our platform is its native ability to evaluate human data collection and enrichment work for automation viability and seamlessly shift between man and machine resources. Matched against every human data analyst performing a data collection or enrichment task on a WorkFusion GUI are various models of machine learning algorithms, each attempting to find and replicate the pattern of work that yields quality data.

The biggest eCommerce, financial services and retail enterprises deploy WorkFusion not only rapidly and effectively automate the work that slow people down and mire large enterprises in cost but also to find those opportunities for automation to begin with.

Imagine how much easier it would be to leverage automation if a business didn’t have to look for automatable processes or test and train algorithms? Imagine how much more profitable that 20,000 person data operation would be if 10,000 of those data analysts were tasked with actual analysis rather than commodity collection and extraction? Imagine how much more valuable human intelligence would be if it were applied only to the tasks that absolutely required it? WorkFusion’s customers don’t have to imagine.


Originally published at www.workfusion.com on December 2, 2014.