Blogs

Top 3 themes: winning use cases in data science and AI

Senior Portfolio Marketing Manager, Data Science and AI, IBM

I am working on a project that requires a lot of research on how people are using artificial intelligence (AI) and machine learning (ML) in real-world business cases. While AI and ML are two of the hottest trends in technology, they don’t seem to be proliferating as widely or as quickly as we would expect.

In fact, according to an Economist Intelligence Unit survey, 70 percent of business executives rated analytics as “very” or “extremely important.” But just 2 percent say they have already achieved “broad positive results.”

There are many reasons for this low success rate. Long development cycles, shortage of the right skillsets, and high data scientist turnover rate are just some of the factors. But what if you could get a quick win to prove the value? I've compiled hundreds of AI and ML use cases and found that there are basically three themes that people can start with to get that quick win.

1. Gather insight.

What happened or is happening, and how can a business measure or monitor it? This is quite evident in many use cases relating to customer engagement.

In a traditional company – like in financial services, banking or retail – there is a huge set of historical data. But this data can be sitting in different databases and locations, and in different formats, structured or unstructured. By using techniques like data mining, machine vision, natural language processing (NLP) and text analysis, the business can find relationships, patterns and drivers to get a 360-degree view of what's going on. Things like customer segmentation – who bought what, when, where, and how – can be the starting point for the business to be able to take actions and easily see the ROI.

Two examples: you can see AMC Networks in how they analyzed data in seconds that unlocked insights to win new viewers and advertisers. And Fifth Third Bank used data science to optimize marketing analytics.

2. Make predictions.

Based on historical data, what is the likelihood of an event occurring in the future?

You can see predictive analytics being used in places like product recommendation, pricing, risk and fraud detection and predictive maintenance. This can be a good, quick win because often times the process of doing such activities are manual, error-prone or slow. By using data mining techniques, AI, or predictive algorithms, the business can quickly realize process automation, better predictions and ROI from the investment.

Here are three case studies to explore how companies are using predictive analytics.

In the financial services sector, Nedbank implemented a solution combining predictive models and decision optimization to predict machine outages and the nature of issues and to schedule repairs. And Accenture is bringing real-time analytics to fraud detection.

Finally, one of the world’s largest perfumery and cosmetics franchisers, Grupo Boticário, uses predictive analytics to enable smarter sales, marketing and production planning.

3. Take action.

What is the best way your business should allocate resources, plan, schedule, and determine your next best action? This type of use case can stand alone or be built on top of the first two.

This use case applies to situations where:

  • It is difficult to make multiple simultaneous decisions, including plans, schedules and next best action
  • There are limited resources or conflicting objectives
  • There are multiple possible future scenarios

By using mathematical programming, ML, and scenario-based optimization, the business can get optimal recommended actions or make scenario comparisons on their KPIs, such as profit, cost, revenue or delivery time. Based on data-driven recommendations, the business takes actions that can directly impact revenue.

Two examples of this use case: a large bottling company was able to reduce plastic waste and improve production efficiency through ML recommendations. Electric company Red Eléctrica de España was able to strengthen how they forecast their demand and optimize production.

Machine Learning and AI are solely the tools that can help a business to do more with the data it gathers. It is up to the team of data scientists, business users, and the IT organization to collaborate, define use cases, and show ROI.

Ready to explore how AI and ML could help your business score a quick win and deliver ROI? Explore more data science use cases and capabilities.