More and more utilities are realizing that they have a problem of too much data and not enough ways to use it all.
Some executives are wondering if they shouldn’t just disconnect their new sensors and data-gathering devices and walk away. Others, however, are searching for solutions, and have found their answers in AI technologies. Machines can compile, organize, and analyze data at large scale and great complexity. AI technologies provide utilities with the insights they’ve been searching for from the piles of data they already have.
Once they’ve assessed whether or not machine learning is the right approach, these utility companies are faced with yet another question: How do they obtain machine learning technology for themselves? Should it be built internally, or should companies partner with an external vendor?
The temptation for many utilities is to try building and implementing a machine learning solution internally, since most utilities now have their own data scientists. However, this may not be the best path forward in terms of getting value from the investment. Here are the three questions to ask when assessing whether to buy or build artificial intelligence:
1. How long will implementation take?
A carefully vetted partner brings a team with solid AI experience that most companies will have difficulty attracting (not to mention time lost recruiting and onboarding). Vendors can also dedicate far more resources and staff to the project, as these implementations are their primary function. An AI partner provides ready-made solutions to a company’s specific issues, while using its understanding of the technology to tailor where needed. Depending on the project and the partner company, working with a partner may offer a significant return on investment within three to six months. With most internal implementations, it takes years to see this same level of progress.
2. Who is going to maintain and upgrade the product?
Even after the product is built, staff must be retained to identify problems and upgrade features. A vendor is continually working to upgrade the product, and issues that arise from another client may provide ideas for efficiencies applicable across the board.
Internal data science teams do have a key role in this process as well. Machine learning companies still need people inside the client utility who understand the analytics at hand. A successful machine learning implementation depends on the collaboration between vendors who have experience with machine learning and internal data scientists who have intimate knowledge of their own company’s operations. An external machine learning company is not going to replace in-house data scientists, but augment them.
Of course, this all depends on selecting a good company to work with. A high quality machine learning company will collaborate with all personnel within a utility to ensure smooth implementation. The vendor must know the importance of fostering a cooperative relationship with all levels of the client’s organization.
3. How are vendor solutions specialized to a client?
Some utilities are hesitant to work with an outside company because they believe that vendor solutions are simply pre-built products that may or may not fit the utility’s needs and can be easily resold to competitors.
These concerns are unfounded. While the precise approach may vary from company to company, often about 70% of what a machine learning company does is using its own IP and platform, and the other 30% is customization to meet each client’s needs, such as plugging in the client’s data and creating a custom user interface. This 30%, which is strikingly different for each client, is where the real work of the project is. The first 70%, while replicable, is a framework that the machine learning company leverages. The end result is unique to each client.
Ali Raza, President of Dover Energy, Explains his decision to partner with an external vendor.
If a utility has a large, dedicated data science team, it may be possible to implement an internal machine learning solution. But any executives attempting this need to take the time to fully understand the infrastructure needed for a successful machine learning implementation.
Machine learning brings a shift to utilities, and strong change management is needed to guide the whole company through the transition. Buy-in from top management is essential, as those managers help spread the vision of machine learning—and the benefits it brings—to the utility as a whole. Implementing AI in utilities requires good leadership regardless of the method used.
Utilities have already started to make the changes that are critical to their continued success. Now they need to take it one step further, and find the right way to move forward as a company and as an industry.