Artificial intelligence (A.I.) use is one of the fastest growing technologies in the construction industry, which I think is a good thing. It is greatly improving productivity and efficiency with the assisting writing emails, helping with spreadsheet formulas, and reviewing and summarizing large documents, just to name a few common uses.

One of the most useful cases is creating custom documents through HTML. HTML is basically what every page of the internet is built on, but it is not limited to just online use. You can create a local HTML page that lives on your computer but acts as a webpage. Contractors are using this for custom built A.I. tools such as time tracking, project landing pages and custom dashboards, just to name a few. Along with this ease of A.I. assisted coding comes the lack of knowledge on safeguards to ensure the accuracy of the output and quality of the data used as the source.

Let’s use a simple prompt such as, “build me an HTML file to turn this spreadsheet data into a dashboard” as an example. It will result in a professional-looking dashboard that pulls the data from the spreadsheet, organizes it and structures it as the user asked.

Now let’s take a look at a couple of shortcomings with the result.

What happens when the data on the spreadsheet changes? Is there a button to load or refresh as changes are applied to the original file? The A.I. will not automatically build that functionality in, as it was not instructed to account for updates.

Is the dashboard linked to the spreadsheet as a live document? If you share the dashboard through an email, will the data reflect the information pulled from the original Excel file? How will this affect versioning once the changes are accounted for and other dashboards are created to show updates?

A better prompt would be “build me an HTML file to turn this spreadsheet data into a dashboard, account for updates in the spreadsheet for live tracking and take into consideration this HTML may be shared company wide.”

Here is another example: “Build an HTML file for field personnel to upload timesheets.”

The A.I. will build a great looking interface with an upload or submit button that appears to function as it should. One slight problem here: where is it submitting to? The user will not get an error saying that there is no storage location assigned, or that the file was not uploaded because these are handled in the error-handling portion of the code. The A.I. did not account for that since it was not prompted to do so. The user will fill out the form, hit the submit button and close the page, never knowing that the information was never transmitted. Chances are that they will not make a paper copy as they were submitting through an app. This causes a false sense of security believing that the timesheet was submitted digitally and no paper record was needed.

A better prompt for this example would be, “Build an HTML file for field personnel to upload timesheets. The data will be stored on our company’s OneDrive and be structured in folders using Employee/date format. Include warnings if the data is not transmitted properly. Ask me clarifying questions to ensure that this HTML file is functioning properly before its release.”

Even with this rewritten prompt, there will need to be some back end work to ensure that the company’s OneDrive is accessible by an outside source and tested from outside of the company’s network.

A.I. is bringing in a lot of opportunity, and for the most part, is bringing a lot of value to electrical contractors. But the quote, “With great power comes great responsibility” really applies here. By the time that the error is found, the potential for lost information can be huge, and in recreating that information you just doubled the amount of work you were trying to improve. You also just lost the trust of technology; you only get one chance for a “first release” make sure it works.