Is AI the Right Solution For My Problem: AI Decisioning Framework and Flowchart

In today’s tech-driven world, artificial intelligence (AI) is often hailed as the magic solution to all of our problems. But let’s be real—AI isn’t always the answer. It’s easy to get swept up in the hype, only to end up using a bulldozer when all you really needed was a shovel. 

As a technical consultant and Microsoft MVP in Azure and AI, I work with companies every day to explore how technology can effectively address business challenges. When it comes to AI specifically, a recurring theme across organizations is helping them navigate the complexities of AI decision-making and tackle the key question: “Is AI the right solution for my problem?”

When I set out to find a resource online to use with companies, I was surprised to come up empty-handed. So, I made my own. This flowchart is designed to help you cut through the noise and determine when to embrace AI, what tool you should investigate… and when to walk away. 

AI Decisioning Flowchart Overview

In order to help navigate the flowchart I have included a narrative around two sample business challenges and how the organization would flow through the chart to determine whether AI is a good solution or not.

  1. Internal Data Catcher: Providing employees information about internal processes and systems.
  2. Proactive Monitoring and Anomaly Detection: Lessen the likelihood for unplanned downtime of machinery.

If you want to skip the narration, you can download the flowchart here!

Use Case #1: Internal Data Fetcher

Problem Overview: The problem we are trying to solve in our fictitious organization is providing employees information about internal processes and systems. Like many organizations, sometimes there are many sources of truth, or they need to go to different sources to find different information. The goal of this is to deliver information to employees as soon as they need it instead of having to ask around the company.

Will this success influence an Objective & Key Result (OKR)? Yes! We believe that delivering this tool will not only help educate our company about AI technology, but it will also increase employees’ access to information and decrease the messages sent to HR asking where information is located.

Can this be done with traditional tech? No or not easily or cheaply. We have a lot of different data sources and information scattered across PDFs, Docs, HR systems, etc. Using all of those tools we also don’t have a unified search today so we can’t easily, for example, have a few if statements and say if it’s an HR question go ask the HR system’s search, because it doesn’t exist.

Is a probabilistic answer acceptable? Yes. One of the key principles of responsible AI from Microsoft is human in the loop. Since this is an internal tool our employees are able to find the information and evaluate it before acting on it. Worst case scenario, they are going to reach out to HR as if the tool never existed. 

Azure or AWS? Azure. As a Microsoft MVP in AI and Azure I am totally biased, but you can fairly easily do this in AWS land too. 

What are we hoping it will do? Answer questions/Chat bot. This is one of the most common use-cases for AI. Robots are really great at searching massive amounts of data that people could never remember or dig through in a timely manner. 

Custom Data? Yes! We are using internal data like HR documents, benefit information, peoples’ job titles and descriptions. Most, if not all, of that data is internal data. 

How much coding? Lots! This question is more along the lines of how comfortable are you writing code and maintaining it, or how custom do you want the look and feel to be? We are wanting this tool to be able to be dropped into any tool or UI so we’re saying lots, but if you’re not overly comfortable with coding you can go the Copilot studio route and still get some customization! 

Can you connect to the internet? Yes! We are expecting this tool to be used by employees with an internet connection in their daily work. This means they don’t have to download a language model to their device or an app to their phone just to ask for information. In this case Azure Open AI is our solution of choice! 

Use Case #2: Proactive Monitoring and Anomaly Detection (Manufacturing Example)

Problem Overview: When our customers’ equipment experiences unplanned downtime the costs of lost productivity, repair costs and loss of business can have enormous negative impacts. Procuring specialized replacement parts in a hurry is often not possible, or at minimum associated with increased costs. 

Will success influence an OKR? Yes. If we have a landing space for our data we can start reporting on performance of our machines and trying to proactively detect anomalies to keep machines running and doing work for our customer. 

Can this be done with traditional tech? No or not easily or cheaply. We could train a traditional model to start monitoring our data and then use the outputs to build a graph in something like react, but that would take a lot of time training the model, dev time to build the UI, and would increase the cost of doing this at all in dev hours and AI computer cycles. 

Is a probabilistic answer acceptable? Yes! We have a human in the loop monitoring anything that is being built before it contacts a customer directly.

Azure or AWS? AWS. I know what I said above about loving Azure, but I have to acknowledge AWS is a good option too! So we’ll flow through that branch. There is a similar offering in Azure that would get us to an actionable end goal too.

What are we hoping to do? Anomaly Detection/Reporting. As stated in the problem statement, I am wanting to keep an eye out for any machines acting odd, try to find trends in machines acting odd before having issues, or trying to find peak performance of the machine. 

Is just reporting enough? No. We are wanting a place where we can monitor our data and label it if needed too. Not just create a dashboard to visualize it.

Do you want to build it from scratch? No. In this case this will be an internal tool that we want to roll out quickly to start getting insights on data. If it were a reporting and workspace product we were selling, possibly building your own makes sense, but in this case the value add is getting insights from our data. In this case we are going to look at Redshift for our solution!

Key Takeaways

AI can be a game-changer, but it’s not a one-size-fits-all solution. This flowchart was created to help navigate the complex decision-making process around AI use, serving as a reminder that not every challenge requires a high-tech answer. 

As you consider your own projects, remember:

AI should enhance your work, not complicate it. 

The key takeaway? Be strategic. Use AI when it makes sense, and don’t be afraid to opt for simpler solutions when they’ll do the job just fine. After all, it’s not about chasing the latest trend; it’s about finding the right tool for making a meaningful impact for your business.