An AI Strategy Does Not Always Mean ChatGPT

It is not uncommon for ChatGPT to be the first thing people think about in relation to Artificial Intelligence (AI). According to Accenture Research, the number of mentions of AI in earnings calls has increased by 6x since ChatGPT was released. Therefore, it might feel like if you’re not using ChatGPT in your corporation-wide strategy, you’re not doing it right.

As AI technology has evolved, it can be difficult to keep track of the problems it is best suited to fix. In fact, according to surveys performed over the past year, 90% of Americans have heard of AI, but only 30% of U.S. Adults correctly identify the different types.

With all the talk about ChatGPT and generative AI, it can be easy to fall into the trap of trying to apply it to every problem. Yes, ChatGPT is extremely powerful. It can be a great tool to summarize text, draft emails, or brainstorm ideas (No, I didn’t use AI to write this blog post… maybe I should have). Still, it’s important to note that not all companies depend on the specific solutions provided by ChatGPT.

How To Best Incorporate AI Into Your Business Strategy

Identify an expensive step in your process and try to mitigate it early with Predictive Analytics.

If you are struggling to envision how ChatGPT-like tools will power your company, maybe because you create finished goods not content, consider that there are other styles of AI that have a longer track record, such as Predictive Analytics.

Predictive Analytics uses data about your environment to predict whether something will happen. Imagine you have been using custom-developed software to run your manufacturing process. You have been collecting data that relates to your operations for years. What if you could be alerted to a potential issue before it is too late to fix?

In our experience, the way to get the most impactful result from AI is to identify an expensive step in your process and try to mitigate it early with Predictive Analytics.

We have worked with clients to build systems which collect many data points: measurements every second (also called time series data), quality approvals, transition timestamps, etc. The initial need for all this data is to manage manufacturing operations, collect audit history, or deliver notifications. With all that data, executives inevitably start to wonder how else to use it for better decisions. We have helped clients identify those areas where their data can be used to predict anomalies.


Here are a few key factors in determining if you can leverage Predictive Analytics to greatly benefit your existing process.

Does your current process gather results data?

Predictive Analytics must be trained on real outcomes that exemplify your environment. Not only must your software track data about your process, but it must also track the results.

Do you track whether a finished good goes through an approval process? What about whether or not a lead turns into a sale? If so, it’s possible this could be used to create good training data because your system would know if the events leading up to it ended up in a positive outcome.

Alternatively, you could have someone manually go back and enter in the results after the fact.

Do you believe that the data you collect is an indicator of what causes the outcome?

The data you collect must be related and causational for what is going to happen. As a counter-example, previous prices of Bitcoin are not a good indicator to determine the future price of Bitcoin.

What types of events cause the behavior you are looking to predict, and do you track that data? Do colder temperatures tend to cause pipes to burst? Do tasks that take longer tend to create lower-quality products?

Can you enrich your data?

You may not have all the causational data that you need for accurate predictions. However, It may be possible to enrich your data. There are services that can be queried based on a key indicator like date and time or an identifier that can be used as a lookup. You can add more detail about your environment,  such as stock market, weather, Multiple Listing Service (MLS) records, background checks, etc.

Do you have an urgent pain you are trying to resolve?

Implementing an AI solution can often be an expensive endeavor. That is why we recommend starting with the most painful need for a more immediate Return on Investment (ROI). For example, what would it be worth to be able to intelligently and proactively maintain heavy machinery, reduce rework, or appropriately score the probability of an outcome?

Do you have the in-house talent or a trusted partner to advise on and implement a solution?

Whether you need someone to develop your solution or to be a second opinion, Unstoppable Software is here to help you detect anomalies, protect against undesirable edge cases, and predict outcomes.


Join our newsletter to receive future articles about how to best leverage data in your business.



Accenture’s Technology Vision 2024

Americans’ views of artificial intelligence

Start typing and press Enter to search