Every few weeks we have an in-house AI meeting with our programming and engineering teams. We could almost have that meeting every day as quickly as things are moving. The first few meetings were centered around complaints by our developers about how bad the GitHub Copilot is. Generating garbage code that took more time to fix than it would have to write. Then we shifted to Claude which we found writes better code.
Next, we started honing in on parts of our coding that were readily taken on by AI and had a high chance of success. Reporting was the first and most obvious choice. Many times, over the years we have taken on a project where building the database and the data collection screens was pretty economical, and straight forward. But the amount of reporting the customer required was a deal killer.
Next building database queries was an easy AI task to delegate. Recently we realized that taking PDFs of label templates and having AI generate label formats in BarTender or Loftware was readily done by AI. Saving potentially hundreds of hours on a project. Now we are starting to question the basic tenants of our project sales process. Do we even need to search for a middleware product to fulfill a customer’s project needs.
For decades we have had the same project methodology when it comes to software. First, will the existing software the customer already owns do the job? If not, is there an off-the-shelf software product that will do the task? Finally, and only last, should we consider writing the code needed for a project ourselves?
The size of a project that we can economically complete on behalf of a customer grows in scope every month. There is possibly a scenario where we write the database and data collection portions for the customer, and they use AI to extract all of the needed reporting or even write their own API to push the collected data into business systems.
Yes, there are security concerns. Feeding a customer’s data into an AI engine is not acceptable. But we are talking about building the framework for the code, not using AI to manipulate or streamline data. There are now AI agents that live in a silo that we are looking at. That way, we are using the AI engines capabilities but not sharing the results with anyone else.
So, if you want a technology partner who is exploring the use of AI to complete project work faster and more economically, reach us at www.emptechgroup.com to learn more.
