My Framework For Finding Impactful AI Use Cases

Read time: 3 minutes

Today, I won't share a concrete use case, but a framework I use to find good AI use cases.

Good AI use cases are characterized by two features: They have a big business impact and they are feasible, meaning you have the data, skills and other resources to realize their potential.

Today, we'll look at how to find use cases with a big impact. 

The framework is called story mapping and is inspired by product management techniques.

Here's how it works:

If you like this content, check out my book AI-Powered Business Intelligence for a more detailed walkthrough of today's use case and many more!

Story Mapping Framework 

The core idea of AI story mapping is to contrast the current implementation of a process with an AI-assisted implementation of that process.

This way, you get a comprehensive overview of what would be different, what you'd need to change, and most importantly, it gives guidelines to structure your thought process.

Creating a storyboard is quite simple:

Step 1: Create the map

Take a blank sheet of paper and divide it into a table with four columns and two rows.

The top four boxes represent your current process, and the bottom boxes describe the future, anticipated implementation.

Name the columns from left to right: setup, actions, outcomes, results:

Step 2: Fill in the map with the current process

To create your storyboard, you need to fill in the columns from left to right.

You start with the first line, where you describe how the current implementation of a particular process works along the following dimensions:

• Setup: Describes how the process begins and lists your assumptions, resources, or start criteria.

• Actions: Includes all tasks and action items performed by or with the resource described in the setup.

• Outcomes: Describes the actual artifacts of the process. What exactly is generated, created, or modified?

• Results: Contains the impact that the outcomes have on the business and/or the subsequent next steps for the outcomes. For example, displaying a report in a dashboard is an outcome, but in and of itself has no impact. The impact is what happens based on the information displayed in the dashboard.

Let's run an example:

Imagine we're working on a use case to prevent customer churn. Customer churn happens when customers cancel their contract or stop buying from our company.

Preventing customer churn sounds like a simple AI/ML use case in theory, but in practice it can be quite difficult to turn customer churn predictions into business value.

Let's take a look at how customer churn is detected and prevented in our fictitious example:

Setup

Currently, customer churn is detected by sales reps receiving feedback during their regular calls with existing customers, or by customer service reps receiving complaints from customers.

Actions

In the next step, customer service or sales reps try to resolve the issue directly with the customer - for example, by offering help with onboarding.

Outcomes

The main outcome of this process is that customer support (hopefully) resolves the customer's existing pain points and complaints. Recurring pain points are reported to the management level or a complaint management system.

Results

The result is that the customer hopefully stays with the current service after the issue is resolved.

Step 3: Compare the AI-assisted approach

In the next step, you do the same exercise for the anticipated future (AI/ML) implementation.

This gives you a direct comparison between the old and the new approach, providing more clarity about how things will change and what impact those changes might have.

Let's continue our example.

How could an AI-assisted customer churn prevention process look like?

Setup

We would collect historical data on how customers used various products and services, and we would flag customers who have churned. We would also bring in sales and customer service people to share their expertise with the analyst in the loop.

Actions

Next, we'd analyze historical customer data to determine if key drivers of customer churn can be identified in the data. If so, we would develop a predictive model that would calculate an individual churn risk for each existing customer in our database and provide context on why churn is likely to happen.

Outcomes

As an outcome, these churn risk scores and churn reasons would be presented to the business. The information could be combined with other metrics, such as customer turnover, and presented in a report in Customer Relationship Management (CRM) or in the BI system.

Results

With this information, customer support could now proactively reach out to customers with a high churn risk and attempt to resolve the issue or remove obstacles before the customer raises a support ticket or turns away without even opening a ticket.

As a result, the overall churn rate could decrease over time because the company can better address the reasons for customer churn at scale.

Conclusion

With both storymaps in place - the existing process and the new one - you should feel more confident in describing what a possible AI solution might look like, what benefits it might bring, and whether it makes sense to take the new approach at all, either replacing it with the existing process or blending it with it.

The purpose of a storyboard is to create a simple one-pager for each use case that intuitively contrasts the differences and benefits of the existing solution and the new one.

A storyboard helps you structure your thought process and is a good starting point when it comes to prioritizing AI use cases. 

Of course, you need to have a good understanding of both how the business works and what AI/ML capabilities exist in order to create good story maps.

As an exercise, you can use this storyboard template and map two or three AI use case ideas.

Which idea seems most promising to you?

Which might have the largest impact?

I hope you enjoyed today's issue.

I'll see you again next Friday!

PS: Want more content around frameworks for AI in Business? Reply with "More" and I'll address them in future issues of this newsletter.

Resources

This content was adapted from my book AI-Powered Business Intelligence (O’Reilly). You can read it in full detail here: https://www.aipoweredbi.com