Photo by Google DeepMind via Pexels
Every business collects data. Many businesses collect and store that data within a data warehouse. In fact, your business has likely benefited from basic reports generated from the data in your data warehouse.
None of this is too surprising: Data warehouses are designed to store and integrate data gathered from across a large organization, making it readily available—and that in turn drives Business Intelligence (BI) for the organization. But what if your organization could “turbocharge” the insights it got from its data warehouse?
This is where Artificial Intelligence (AI) and Machine Learning (ML) have a role to play. AI refers to technologies meant to mimic cognitive functions such as perception, learning and problem solving. ML is simply a subset of AI that focuses on learning over time. AI generally (and ML specifically) can empower organizations to use their data in ways that go well beyond basic reporting and querying. For example, AI/ML can use centralized data to create models which in turn can forecast future sales, customer sentiment, market trends, or potential business risks…just to name a few of AI’s many potential talents.
We recognize that there has been much hype around AI, and not all promises have been delivered on with this technology. Most business owners are simply wondering whether AI/ML can help with their specific business…and how they can get started using it.
Fortunately, the value of AI/ML for database and data warehouse applications is stunningly straightforward, with known use cases and proven value. By understanding what AI/ML is actually good at, we can more easily see how it can bring value, especially to BI and decision making.
Example: Using AI/ML and a Data Warehouses to Find Market Segments
To get a grasp of what AI/ML does with the data in a data warehouse, it helps to have an example. Here we will explore finding new market segments as part of overall BI.
In this case, a company can gather data from existing customers and then deploy AI/ML to tease out the most important characteristics of its best customers. This information can be used to build a customer model, which can then help the sales team discover new potential customers that fit the profile. The model can also be queried, allowing users to ask specific questions to inform a marketing strategy going forward:
- Which segments represent the best opportunities for the company? Which segments are most open to being up-sold, or cross-sold? Which ones have the greatest lifetime value?
- Are there differences in the way that men vs. women shop for our products? Or between Gen Xers and Gen Zers? Urban customers vs. rural? (Are these even good ways to segment the market?)
- Is there a noticeable gap in who is being addressed by our marketing and advertising? Does this “gap” represent an unrealized opportunity?
- Are there any indicators that a customer will cancel/quit (i.e. can we predict churn)?
It takes just a little imagination to apply this example to different industries. Maybe the customer profile is built from purchase data across a series of physical retail stores together with website purchases. Or maybe they are built from account profiles at a large bank. Or patient profiles held by a medical device distributor. In each case, existing data can be used to find patterns and trends, which can then inform business strategy.
Knowing What AI/ML is Good for by Knowing What it is Good At
Part of the hype around AI came about because the term seemed to imply an almost-human level of intelligence ("GAI" or General AI). Tech visionaries, for their part, did little to dispel that notion. The truth is that AI is not yet at that level of sophistication…but it can do some things extremely well.
Specifically, AI/ML is extremely good at pattern recognition. And it can recognize patterns even when they are hiding in mountains of raw data.
This means that any BI applications requiring recognition of a pattern in a sea of otherwise noisy data will benefit most from AI/ML. These might include operations like…
1) Trends and Forecasting
AI/ML is great at analyzing historical data in order to tease out trends, and then make predictions based on that historic data. For example, AI/ML might be able to alert a company when a product suddenly becomes popular (or loses popularity), and even recommend courses of action when those trends threaten to strain the current supply chain.
Another great use of AI/ML is to look for seasonal variations in the sales cycle or business cycle. This is especially helpful for predicting product demand for different geographic areas and keeping products (and services) flowing.
2) Anomaly Detection
Just as AI/ML is good at finding patterns and forecasting future trends, it is also good at noticing when an event or metric wanders far outside expected values. This kind of “anomaly” can be an early sign that something has gone wrong and needs more focused attention from a human being.
A great example of this is the spate of new fraud detection tools used at banks and financial institutions. Banks and FIs are using AI/ML to analyze transaction data in their data warehouse, looking specifically for suspicious patterns, in real time. When such a pattern is detected, the bank can require further authentication, cancel transactions pending further investigation, or even put a freeze on the account. Traditionally, such patterns might go unnoticed, or only be noticed well after the bad actor has already perpetrated their scheme—but today, billions of dollars of fraud is prevented before it makes much of an impact.
3) Recommendations
Our tastes and preferences are not entirely random; we tend to like similar things, and we also tend to share tastes and preferences with other people who are like us (or have similar occupations, hobbies, etc.) This means that past information about our likes and dislikes can provide a wealth of information about what other kinds of products and services we would likely enjoy (or find useful).
Product recommendations have become the secret weapon for eCommerce platforms, streaming services, and other similar industries that rely on understanding user preferences. Amazon, Netfllix, Spotify, and Airbnb would not be the companies they are today without understanding the preferences of their user base in very fine detail. That level of detailed segmenting is possible with the use of AI/ML. Making good recommendations that can steer users to explore more of your products or services not only helps you to sell more, it also creates a better customer experience.
4) Modeling
Data in a data warehouse can be used to create an ideal “model” of how something in the real world is supposed to act (or could ideally act), and then use anomaly detection (see above) when that thing behaves in a way that is not predicted…or is predictive of something going wrong.
One area where this kind of modeling is used is for predictive maintenance in manufacturing. Manufacturing plants can use AI/ML to analyze sensor data from various machines and look for changes that regularly predict equipment failure. It can then “stay on the lookout” for those changes, warning operators when equipment is near its failure point and allowing them to schedule downtime to fix the issue proactively.
Another example is customer churn analysis. Telecom companies, for instance, have been using AI/ML to identify customers that are at high risk of churn and take proactive measures to retain their business. Since retaining customers is much less costly than acquiring them, this helps companies get a better ROI on their marketing dollars—and helps ensure satisfied customers.
Getting Started with AI/ML with Your Own Data
There is no doubt that the addition of AI/ML tools can greatly aid in strategic decision-making. This makes sense, as AI/ML is so much better than we are in picking out patterns from noisy data and then using them to create models against which to make predictions. These activities allow organizations to find straightforward ways to innovate and go to market.
But the #1 question for any AI project is not “What can it do?” nor “what will we learn?” The biggest question most organizations have is “How do we even get started?” Having a data warehouse and an IT team in-house might not be enough. Depending on your need, you might have to build a dashboard or application from the ground up.
This is where the team at CSG can help. We have been creating solutions in the BI and Data Visualization space for decades, and we have been at the forefront of using AI tools to augment and extend those kinds of solutions. We would be happy to meet with you to assess your current tech stack and discuss what kinds of possibilities AI tools could unlock.
