Measuring value can be challenging – sometimes even impossible. But engaging in the discussion will go a long way to ensuring the business fit for your AI/ML initiative.

The key here is to ask good questions. Here are a few questions that may be helpful to measure value and align the AI initiative with strategic goals:

• What is the problem? What objective are we serving? What are we trying to achieve?
• What kind of decisions do you want to support? Why is that important?
• What does success look like?
• Who is the end-user? What will change once the solution is implemented?
• What is the difference between measuring outcomes – and measuring value?

Drill Down into the Complexities When Defining the Goal

Typically, organizations are able to align on big-picture goals. The business users may define a goal such as:

• Discover customer behavioral trends that will predict future purchases
• Automate the order-to-fulfillment process
• Extract a subset of product use cases for immediate customer service interactions
• Predict worker absenteeism to improve scheduling

It’s drilling down into the details of the next level that is more difficult – and more critical. In the predicting worker absenteeism example above, it is important to understand what data is truly available. The amount of data required would be different if you want to predict absenteeism on a national level versus a single location. If the data is not sufficient to predict absenteeism for a single location, you could look for acceptable alternatives, such as: Grouping the data using existing categorical attributes (e.g., geography, gender, age, etc.); Using clustering techniques to derive groups using similarities and differences; or Developing a technology solution like an app to collect more data points.

Again, this process comes down to asking good questions. Sometimes this process can be mind-bogglingly complex. It may require visual tools for mapping resources and their relational connections, or graphing process flows – including the dependencies for sequential and parallel tasks. You might find that you need to engage resources from other departments to answer some of the questions.

Yet these focused details will help your technical team align the scope of work according to your available budget and resources. They will be able to evaluate the tradeoffs between out-of-the-box and custom solutions, rank-order the implementation tasks in terms of ROI, and prioritize tasks that add the most value to the end solution.

Here are a few probing questions that can be helpful during this stage of the process:

• What level of accuracy is required?
• How timely do insights need to be?
• What is available? What data can we create? What data can we buy?

Actively Manage Change

By definition, AI/ML initiatives will change a process. It could be a simple change such as providing a supporting input into an existing decision-making process. Or you could fully automate a routine decision-making process in order to elevate the role of the human decision-makers.

One of the most overlooked reasons that AI/ML initiatives fail is that the organization does not adapt internal processes to account for the change precipitated by the AI/ML initiative. Organizations may fail at realigning work, training staff on how to use the new process, or coordinating work or handoffs between teams. Often, that’s because these types of internal shifts often require authority above a single line-of-business manager.

Equally as common, teams accustomed to traditional ways of working often find the transition to a data-driven mindset unsettling. In order for your AI/ML initiative to be successful, you must build trust that data-driven decisions are, indeed, superior to gut feel or tradition. If business managers or executives cling to old wisdom or refuse to yield their decision-making authority to an analytics process, there will be no transformation or value gained.

Here are a few questions that you can ask to facilitate the change-management process:

• What teams will be impacted, and how?
• What kinds of decisions could be impacted by this data?
• What KPIs will be affected?
• What change-management processes will be included in the solution?
• What should executive sponsorship look like?
• What will we communicate to our employees and stakeholders? How will we build awareness?

Plan for the Long Term

Some organizations look at an AI/ML initiative as one-and-done. A better analogy would be an investment in a car, which needs to be maintained, oiled, and polished. AI/ML models need to be maintained, retrained, measured for quality concerns like drift or class imbalance, and tested for accuracy, precision and confidence. Data isn’t static over time, so models may need to be tweaked or updated.

In addition, the more precise the model becomes, the more value you can gain over time. It’s like fine wine – the more you invest in learning and experimentation, the deeper the flavor. The insights that you may develop from your AI/ML initiatives can well become your unique competitive advantage – like a Google or Netflix algorithm.

Finally, even if AI/ML initiatives aren’t applicable to you today, it would behoove you to consider what data you may want in the future. For instance, if the volume of data you need to make useful decisions is not available today, start thinking about how you could start gathering such data. Some data cannot be collected retroactively.

Ask the Right Questions Up-Front, and Listen to the Answers!

In short, the better your questions are upfront in the planning stage, the better you will be able to align your AI/ML initiatives with your strategic goals. And while each business problem requires its own unique process breakdown, having a standardized approach to aligning AI initiatives to strategic goals drives cohesion and reproducibility across your teams and organizations.