Organizational Impact of Machine Learning Transformation

Open Data Group February 27, 2019

             In a previous blog, we discussed how companies can enable a machine learning transformation within their organization. One key element for successful transformation is the organizational alignment to this goal. Leadership must ensure that each employee and department is aligned toward the goal of enabling machine learning within the organization. In addition, clear and demonstrable accountability is paramount.   It’s not enough that everyone in the organization is aware of the goals and objectives of machine learning, but they should also know the role that they play in it.

              A crucial part in both organizational alignment and accountability comes from the top: A diverse set of executives, such as the CIO and CDO and various business leaders of the company, who must decide that the machine learning transformation is a priority for the organization. While it may seem “obvious” that one department, usually the data science team, is “leading” these critical initiatives, it must be stated that they cannot succeed alone.  The old saying “it takes a village” will certainly be true for the analytics transformation ahead.

              Data science teams are generally responsible for the model throughout its development and will typically partner with other teams as the model moves into production.  The initiatives will succeed with the help and support of the operations, data, tech, and security teams. Each of these departments has a role to play throughout this process. Operations/ business units must ensure that the model will add value to the business. The data team must ensure the data input is up to date and accurate. The tech team must ensure the model has the proper technologies needed to be deployed into production and run successfully. Finally, the security team must ensure that sensitive information isn’t at risk throughout the process. Following the lead of the data science team, these departments must do what they can to push the machine learning transformation forward.  All of this work can be rolled up into the idea of a Model Development Life Cycle – a unified strategy and set of processes that are applied throughout each model’s useful mathematical life.

              One difficulty that companies often face when enabling a machine learning transformation is that there are so many diverse tools that can and should be used to support models. Because so many departments must be involved in enabling machine learning, many different tools may be used to achieve each department’s specific needs and goals. Depending on the use case, the technology needed to support machine learning models can be very different, and it can be extremely difficult to standardize these technologies.  Balancing the need for the diversity, but supporting the realistic requirement to have scalable, repeatable supportable processes is exactly why we developed FastScore.  Teams have to find the balance point, intersecting data science, data engineering and IT, and simultaneously enabling each teams own journey while ensuring these mutual efforts can scale. To learn more about FastScore, check out our product page here.  Each organization will have it’s unique journey to the data and ML/AI transformation, but it’s sure to include the need to align organizational focus, accountability and systems to realize the business outcomes desired.

Topics: Machine Learning, Model Development Life Cycle