AI is everywhere. By now, you’re familiar with ChatGPT, Gemini, Copilot, and several more third-party tools. And while they’re useful, they do come with some considerations.
First, the data risk is a real thing. About 48% of respondents from Arcoro’s AI in Construction HR survey cite data privacy as their top concern with AI, roughly triple the next most common concern, accuracy. Routing sensitive HR information through free, unaffiliated, third-party AI tools comes with a lot of risk because unless you opt out, some third-party tools can use your information to train their models.
Second, adding another tool to your already full tech stack might be the proverbial straw that breaks your HR team’s back. It’s another tool to learn, another tool to monitor, and another tool to track. Your overloaded team doesn’t want another solution. In fact, 23% of survey respondents said having AI built directly into their existing HRIS would be the single biggest accelerator toward adopting AI faster. But most construction HR teams aren’t using integrated AI tools… yet.
Given that so few construction HR professionals surveyed are using integrated AI tools (only 5%), it’s probable your HR software provider currently doesn’t include a robust offering. Talk to your provider about where the roadmap is going and the plans for integrated AI.
Here are five questions to ask before you trust any HR tool with AI.
Is our company data used to train your public or shared AI models?
Where are the AI models hosted, and what encryption standards protect sensitive employee records?
How do you test and audit your algorithms to prevent discrimination in hiring, pay, or performance reviews?
Can we see the exact logic or data sources behind an AI recommendation, or is it a black box?
Can human managers easily turn off specific AI features for certain roles, locations, or processes?
AI gets smarter the more it learns, and that’s a good thing when it’s learning to serve you. Your data should help the tool understand your company, your crews, and your workflows. It shouldn’t be used to train a public or shared model that serves everyone else.
Some AI tools use the data they learn to train their public or shared models, which puts your company at risk of exposing your data to other users, according to, according to Stay Safe Online.
If you’re inputting private or proprietary company data, like jobsite bids or employee Social Security numbers, it could show up in other users’ results. Once your employee data is exposed, the legal repercussions could be severe, and data that’s been absorbed into a model is very difficult to remove.
NO. Your data is used only to deliver results for your company. It is never used to train public or shared AI models.
Your software provider should be able to tell you where its AI models are hosted and how it protects your data.
Firewalls and secure system design are a good start, but they can’t stop everything. As Microsoft points out, if someone physically steals a hard drive or backup tape, they could load that database onto their own system and read everything on it.
That’s where encryption comes in. Transparent data encryption (TDE) scrambles the data stored in a database and locks away the keys needed to unscramble it. Without those keys, stolen data is unreadable. Arcoro uses TDE to protect customer data.
Arcoro uses industry-recognized models that are made available through our hosting provider, Microsoft Azure. We do not built our own models. As far as encryption goes, all customer data is encrypted with AES-256 at rest (database transparent database encryption, storage, disks, and backups) and TLS 1.2 in transit across our production endpoints.
AI is being built into hiring software to help quickly write job descriptions, scan resumes, and scour job boards for candidates. While these tasks are getting done faster, they’re still prone to bias, even from AI, and you’re liable.
According to the American Bar Association, AI bias can occur throughout every stage of the employment relationship and can result in discrimination lawsuits. For example, in 2022, the EEOC sued a tutoring company because the AI hiring program it used “automatically rejected female applicants age 55 or older and male applicants age 60 or older,” resulting in screening out over 200 applicants because of their age. The company had to pay $365,000 to the group of automatically rejected applicants, adopt antidiscrimination policies, and conduct training to ensure compliance with equal employment opportunity laws.
Your software provider should regularly test its tools for patterns that could create bias or disparate outcomes across protected groups.
Arcoro regularly tests its tools for patterns that could create bias or disparate outcomes across protected groups. We also advocate for human review in the process and periodically audit both the data going into the system and the outcomes it produces to identify and correct potential discrimination.
Many AI tools work like a black box. You see the recommendation, like a ranked list of candidates or a suggested pay range, but not how the tool got there. That’s a problem in HR, where every hiring, pay, and promotion decision needs to hold up if an employee, applicant, or regulator asks why.
When you can see which data and factors shaped a recommendation, you can catch errors, spot patterns that might signal bias, and explain your decisions with confidence.
When you can’t, you’re relying on a tool you can’t check for decisions you’re still accountable for. Your software provider should be able to show you what goes into its AI, not just what comes out.
Depending on the feature, Arcoro’s AI architecture is built with the ambition to provide insights into AI-based recommendations so you can be informed.
AI is moving fast. Arcoro’s research found 72% of respondents use AI every week but only 8% have standardized AI across HR. This lack of oversight opens your company up to risk as your teams are likely using AI faster than you can manage it (see our article on creating an AI usage policy).
Until you have safeguards in place, you need to have an option to turn off the features that could cause data leaks or privacy concerns. You should also always be in control of your AI, not the other way around.
YES. Arcoro allows its customers to turn off the AI features they’re not ready to incorporate, including by role, location, or process.
AI can give your HR team back real time, but only if it’s built on a foundation you can trust.
Before you adopt any AI-powered HR tool, ask your provider these five questions and expect clear answers. Vague responses tell you something too.
At Arcoro, we build AI directly into the platform construction HR teams already use, so you get the efficiency without routing sensitive employee data through third-party tools.
Ready to see AI that’s built for construction HR? Schedule a demo to see how Arcoro puts you in control.