Article originally written by Scott Emigh. Find him on LinkedIn here.

The two letters heard most frequently at Microsoft Inspire this year were “A” and “I”. In this Partner Tech Perspective, I’d like to take you through some of the approaches we believe will be most effective for our partner community in establishing their Artificial Intelligence (AI) practices and solutions. Let’s tackle this by addressing the three biggest challenges facing successful AI projects:

  • AI solutions are only as effective as the quality of the questions asked. All too often, our customers don’t know what questions to ask. Even savvy business leaders can struggle with the decomposition of desired business outcomes into digital signals, insights, and automated action—but the Listen, Think, Act model can help structure AI solution consultations with your customers.


  • A lack of commitment or endorsement of AI from top leadership as a decision-making alternative to “how it’s always been done” results in projects failing to turn into funded initiatives. Instead, infuse AI into every project/service you deliver as a subtle way of introducing your customers to the power of AI.


  • The condition of a customer’s data estate can often become a serious obstacle to operationalizing an effective AI solution. A host of data problems—including poor processing, lack of completeness and relevance, poor integration, and non-compliance—can pose serious risks. Leverage the opportunity to modernize as you migrate and create the conditions for effective AI scenarios after moving to cloud.

There are certainly other challenges in successfully implementing AI solutions for customers, including access to AI and Data Science talent, which we’ll touch upon at the end. For now, we’ll use a series of three posts to tackle the main challenges, starting with this one.

Winning new AI projects

Solution: The Listen, Think, Act model

I like to structure a digital transformation discussion using the model of Listen, Think, Act. This thought experiment begins by determining what act would have to take place to address the business problem. For example, do we need to create a field service request for a maintenance issue which we think is imminent? Do we need to issue a special stock grant to a critical employee who we predict is about to leave the company? Do we need to make a new service bundle offer to a customer that is ready for more?

With that end in mind, consider what data we might listen to that could inform a solution to the problem? Be sure to consider data and signals that may not currently be available to the customer, because virtually anything can be instrumented at this point.

Next, think about that data—this is where we would directly apply AI. By considering the data, what unique insight could we derive that addresses the business problem?

Then, back to the action. We must act on that insight; to truly digitally transform the business, the action should be automated to the extent possible. That might mean workflow configuration through Microsoft Dynamics 365 or Microsoft Office 365, or it could be low code/no code via Flow and Power Apps, or a fully custom and integrated workflow that results in an action occurring at the edge on a mobile device.

Let’s walk through a real-world example of this approach. Say you read an article in the Wall Street Journal about American Airlines cutting unprofitable international flights in response to higher fuel prices and uncertain demand. Now, canceling international routes is not without significant cost for an airline, both in terms of operational expense and opportunity losses. The business problem could be described as uncertainty in route demand vs. fuel costs resulting in expensive guesses about routes. One way to address this problem would be to generate a heatmap of 6-month forecasted passenger profitability between cities (which we do for Azure capacity demand planning). As inputs (Listen), take indicators of passenger traffic, such as exchange rates or hotel occupancy rate trends, and indicators of fuel prices, such as oil reserve levels. Then, one could develop an index of future passenger profitability by route (Think). This heat map, implemented as a dynamic Power BI map, for example, could then feed the planning team with an overlay to current route contract negotiations (Act). The end result would be more predictability on route commitments and improved profitability.

Listen, Think, Act is an easy way to get the conversation going with our customers, and helps to keep the conversation at the right business altitude—focused on business outcomes and not simply experimenting. It also ensures that you’re working toward a holistic solution that’s inclusive of data acquisition/IoT, AI & analytics, and business process automation/edge computing.

Once you have used the Listen, Think, Act approach with your customer to establish an AI solution scenario, I encourage you to explore the Team Data Science Process (TDSP), a methodology for delivering and managing effective AI solutions throughout the lifecycle of your data science projects. It helps you ask the right questions and addresses the gaps and challenges throughout the AI lifecycle by providing best practices, examples, and templates. You can find more details on TDSP here.

Keep an eye out for the next post, in which we’ll tackle the problem of AI sponsorship and executive buy-in.

Learn how we can accelerate your business. Visit the Microsoft NZ Partner Hub at