David Menninger's Analyst Perspectives

The industry is making huge strides with artificial intelligence (AI) and machine learning (ML). There is more data available to analyze. Analytics vendors have made it easier to build and deploy models, and AI/ML is being embedded into many types of applications. Organizations are realizing the value that AI/ML provides and there are now millions of professionals with AI or ML in their title or job description. AI/ML is even being used to make many aspects of itself easier. Organizations that want to build and deploy their own AI/ML models need to be realistic about the capabilities that are available today. As a practical matter, organizations should anticipate that a robust AI/ML deployment in the current environment requires a set of specialized skills and operational processes, including data operations (dataops) and ML operations (MLops). Collaboration across these disciplines and processes is also required.

Our Machine Learning Dynamic Insights research identifies the top challenges organizations face in applying machine learning. Accessing and preparing data tops the list. Data preparation plays a significant role in AI/ML, and different algorithms require data to be prepared in different ways. For example, many continuous values need to be prepared as normalized values from 0 to 1. In other cases, data needs to be prepared in buckets or bins of data. Data needs to be checked for bias otherwise the models produced will reflect that bias.

Organizations seeking to deploy their own AI/ML models recognize the need for specialized skills, but struggle because they don’t have enough skilled data science resources to turn models into operations. Almost half (45%) of organizations report they have limited or no knowledge in applying machine learning. To be most effective, models must be deployed into business processes so the analysis can be performed in a timely manner and the results can be used to impact business outcomes. That is a cross-discipline process that requires the involvement of application developers. Once those models are deployed, they must be maintained because as soon as a model is developed, it starts to become stale. While machine learning algorithms can adapt to some degree, fundamental changes in the economy or competitive landscape may require a different set of algorithms.

The budget challenges organizations face likely result from these various other challenges that require specialized skills. Organizations that lack applied ML skills need to acquire them and fund them through wages or consulting fees. Participants in our research that were expert or knowledgeable in applying ML were significantly less likely to face budget challenges than those that had limited or no knowledge, 12% compared with 24%.

There are many ways you can use AI/ML today. The most common applications organizations report are front office functions of customer service, marketing and sales. Forecasting and fraud detection are also common among our research participants. The top benefits reported of effectively using AI/ML include achieving a competitive advantage and improving customer experiences. Participants also report increasing sales, lowering costs and responding faster to opportunities and threats.

Not all analyses require specialized skills. Some will be more easily enhanced by automated AI/ML analyses such as customer segmentation. However, if you want to take full advantage of AI/ML and achieve a competitive advantage for your organization, you should be prepared to invest in specialized skills and processes.

Regards,

David Menninger