While consultants are keen to tout machine learning (ML) to gain more work, they may find it will ultimately have the opposite effect. As a former consulting leader, I believe ML-based disruption will have a greater impact on consulting than the globalization of work. This is because eliminating work altogether is more disruptive than lowering its cost.
Inspiration Versus Perspiration in Work
Edison said his work was “one percent inspiration and ninety-nine percent perspiration.” Why is consulting work so vulnerable? Not all consulting is pure “inspiration.” It’s the “perspiration” part of the work that is vulnerable. In fact, a lot of consulting work is: 1) highly repetitive, 2) information gathering that can be automated, and 3) problem solving that is common and predictable. Namely, the perfect ingredients for machine learning. Here are some major activities that may be vulnerable.
Firms spend millions modifying software workflows that are outdated because conditions change. At OnCorps, we have developed apps that use ML to automatically change workflows (i.e., how the app actually appears to the user and what it asks the user to do next) based on algorithms. In our first system, we used algorithms to score and prioritize possible errors (also known as anomalies) in mutual fund transactions. When the algorithm detects an anomaly that represents a big potential loss, it triggers a workflow urging deeper investigation and approvals. How will this disrupt consulting? When systems begin automatically sensing conditions and modifying themselves, consultants will no longer be needed to make the same modifications.
Cloud architectures are more easily integrated than traditional systems. This is because open source communities, the primary drivers of cloud technologies, are contributing millions of lines of code and countless hours to better integrate components. While this is still an early feature, machine learning will begin playing a vital role in managing these new contributions - automatically testing them, and offering faster ways to make disparate databases work together.
Consultants are fond of using surveys and interviews to gather information. These datasets are often stored in spreadsheet files, only to be exported as charts in slide presentations. At OnCorps, we have replaced spreadsheet surveys with cloud-based diagnostic tools that draw on existing data and add new diagnostic question data to help improve predictive values. The data are stored in collaborative filtering databases, which can trace every answer back to each individual user.
Consultants spend hours reading, classifying, and correlating data to gain insights. When they find meaningful insights, they prepare slides to present to client executives. While much of this insight work may never be replaced, those familiar with machine learning will recognize a pattern. Machine learning algorithms learn how to interpret data and make predictions. OnCorps uses machine learning algorithms to classify historic decisions so that our systems can guide decision makers in future decisions. Our algorithms automatically read data, classify data into meaningful splits or clusters, and determine combinations of factors that lead to the highest payoffs.
Building a Business on the Disruption
Edison’s ratio of inspiration and perspiration should be welcome news to those who want systems development work to be reduced. At OnCorps, we have built a business that offers disruptive customized solutions without the extra fees of consulting. By applying a combination of agile methods, cloud virtualization, and machine learning, we have consistently delivered complex customized decision guidance systems in about 1/10th the time of conventional projects. As the diagram below illustrates, our customers pay a fixed pilot fee to obtain a customized production system. They then pay a subscription for support of the system. We provide a dedicated high-end data and systems support team essentially for the price of a SaaS subscription. We are able to do this at SaaS-level profits because the work of modification and integration is reduced dramatically.
As ML becomes more pervasive, it will inevitably reduce the perspiration work of systems development. We intend to drive this faster than others, offering substantial time and cost savings to our customers in the process.