How to use AI in Business Analytics

Impact areas for augmenting your work as an analyst

Hi there,

Today, we'll find out how AI can improve business analytics and augment your work as a business or data analyst.

Contrary to some fears, AI isn't here to take away jobs — it's here to enhance them. The challenge (and opportunity) is for professionals like us to assess impact areas and use this technology responsibly.

AI for business analytics isn't one-size-fits-all. So we're going to take a fresh look at each type of business analytics and how AI can enhance each. Plus, we'll make it more concrete with some real-world examples.

Ready to dive in? Let's go!

📣 Event announcement 📣 

Join Steve Rosvold, founder of CFO.University, Prashanth H Southekal, founder of DBT Institute, and me for a LinkedIn Live on the role of data, analytics, and AI in transforming finance - attendance is free!

Recap: The Four Pillars of Business Analytics

You're probably already familiar with the four main types of business analytics: Descriptive, Diagnostic, Predictive, and Prescriptive.

Each one has its unique role, challenges, and opportunities. And, each one can benefit from the right application of AI.

  1. Descriptive Analytics answers "What happened?" by providing a clear view of past events.

  2. Diagnostic Analytics goes a step further, asking "Why did it happen?" by analyzing data across variables to reveal underlying patterns.

  3. Predictive Analytics uses historical data to predict future events, answering the question, "What's likely to happen?"

  4. Prescriptive Analytics ties everything together, using insights from the other three types to answer the question, "What should we do?"

AI can enhance each type of analytics, but its role varies depending on the specific requirements of each stage.

We'll break down each type, discuss how AI can enhance them, and use a concrete example to illustrate these points.

AI in Descriptive Analytics: Boosting the Basics

Descriptive analytics is the backbone. We use it to get a clear view of past events and to lay the foundation for more advanced concepts. Surprisingly, AI can be extremely useful right from the start.

To make it more concrete, let’s consider a practical example.

Suppose we're a telecom company dealing with a common issue - customer churn. We have churn rate data from the past six months as follows:

This table looks innocent, but depending on the data, it could have been a lot of trouble to get here.

Here’s how AI could help us in this context:

  • Visualization: This table is a good start, but wouldn't it be nice to see the trend over time? Generative AI can suggest and create appropriate visualizations based on the data we have.

  • Exploratory data analysis: For more advanced descriptive analytics, Generative AI could suggest a structure for our EDA and perhaps even write out the code for this.

  • Q&A: Instead of just showing a table, we could use NLP to ask questions about our data like "What was our churn rate last month?” - Much more fun!

  • Data Narration: To communicate our findings, we can use NLP to create a precise narrative of the observed insights. This is a great help – who likes writing out table summaries anyway?

  • Unlocking new data: Using computer vision or speech-to-text services, we can turn unstructured data such as PDF documents or call transcripts into actionable data - maybe we needed to process some cancellation letters to create the report above?

  • Data access: With the help of Generative AI we could write, debug and document complex SQL queries that were perhaps necessary to get this data.

  • Data preparation: Generative AI can help us to create data preparation scripts, or suggest suitable data cleaning methods.

Key AI Archetypes for Descriptive Analytics

For descriptive analytics, there are potentially the following AI Archetypes at work:

  • Generative AI

  • Computer Vision

  • Speech Recognition

  • Natural Language Processing (NLP)

Next, let’s move to Diagnostic Analytics.

AI in Diagnostic Analytics: Reduce Time to Insight

Diagnostic Analytics takes us a step further by answering "Why did it happen?".

Continuing our telecom company example, we noticed an upward trend in customer churn rates. But we're not content with just knowing the "what"—we want to know the "why."

So, we dig deeper and enrich our churn data with more customer information, such as age groups.

Here's what we find:

This table reveals that younger customers are more likely to churn, and this trend increased in Q3.

So how could AI have assisted in this diagnostic process?

  • Pattern Detection: We could use tools like Auto ML to check for important " features " (= key influencers) that drive churn. In this case, "customer age" might have emerged as the most important attribute among a plethora of other variables.

  • Q&A: Just like above, NLP could provide us with more intuitive ways to interact with our data, e.g. “What was the churn rate for customers below 30 years”.

  • Data Narration: NLP could help to automatically generate explanatory text to communicate our findings.

  • Methodological support: Generative AI could help guide us through analysis frameworks (e.g., 5-Why, Issue Tree) so we can do more effective number crunching. It could also help us design and set up experiments like A/B tests for further data collection.

Key AI Archetypes for Diagnostic Analytics

For Diagnostic Analytics, we'd primarily leverage:

  • Auto ML

  • Generative AI

  • Natural Language Processing (NLP)

Now, let’s leave the past behind and get into Predictive Analytics.

AI in Predictive Analytics: Make Better Predictions

Finally, we’ll want to answer "What could happen?" using past data patterns.

Let's go back to our telecom company scenario. After analyzing the factors that influence customer churn risk, we could use them to predict the actual risk for each customer. Here's what a (very simplified) churn prediction table might look like:

There are multiple ways AI could help us in this space:

  • Predictive Models: With Auto ML, we could quickly build, evaluate, and deploy a model that predicts churn for each customer - potentially outperforming simpler models used previously.

  • Simulation & Scenario Analysis: We could use Auto ML to build multiple models to simulate different scenarios, such as the impact of a new loyalty program, and see their potential impact.

  • Forecasting: Auto ML can also be applied to time series forecasting problems. For example, we could predict next month's revenue based on predicted churn rates (among other variables).

  • Methodological support: Generative AI can suggest suitable models for prediction and evaluation criteria.

  • Documentation: When we create customer churn predictions, there are a lot of assumptions we make: how we define churn, what the data should look like, how accurate the predictions are, and so on. Generative AI could help us keep track of these experiments and generate automated documentation, leading to better transparency and governance.

Key AI Archetypes for Predictive Analytics

In Predictive Analytics, the following AI archetypes mainly come into play:

  • Auto ML

  • Generative AI

  • Natural Language Processing (NLP)

With a better understanding of the past, present, and potential future, we're now equipped to determine the best course of action.

Enter Prescriptive Analytics!

AI in Prescriptive Analytics: Guiding Action

Prescriptive Analytics is the final step that helps answer "What should we do?" based on the insights gained so far.

Returning to our telecom company, we now have a table of customers, along with their predicted churn probabilities. The next step is to decide what we should do to reduce churn, especially for those high-risk customers.

This is what our updated table could look like at the end of a prescriptive analytics process:

So, how could AI support us in this prescriptive stage?

  • Next Best Action Models: Auto ML can be used to suggest the next best offer or action to take for each customer based on their churn risk and historical data, such as special promotions, upgrades, etc.

  • Simulation: Auto ML can simulate the outcomes of different actions to assess their potential impact before implementing them.

  • Methodological support: Generative AI can suggest suitable approaches and optimization techniques

  • Documentation: Generative AI can document assumptions and the analytical approach.

Key AI Archetypes for Prescriptive Analytics

As we have seen, the most relevant AI archetypes in Prescriptive Analytics include:

  • Auto ML

  • Generative AI

Conclusion

AI can play a crucial role in every stage of business analytics, and it's up to us, data professionals, to leverage these tools and improve our work.

Remember, AI is not about replacing us — it's about augmenting our capabilities and allowing us to do what we do best, but even better.

That's it for today.

I hope you've gained some insights on how AI can augment business analytics every step of the way.

Don't hesitate to dive deeper into each area and explore how AI can improve your specific workflow.

See you next Friday!

Tobias

Want to learn more? Here are 3 ways I could help:

  1. Read my book: Improve your AI/ML skills and apply them to real-world use cases with AI-Powered Business Intelligence (O'Reilly).

  2. Book a meeting: Let's get to know each other in a coffee chat.

  3. Follow me: I regularly share free content on LinkedIn and Twitter.

Reply

or to participate.