Honestly? Not very. At first, I had a lot of doubts, not about AI’s ability to deliver but my ability to create something worth using. Could AI really deliver accurate, region-specific HR advice? Would people trust it?

One major concern was accuracy, because for those who are familiar with AI it’s very important to fact check the information it provides. HR decisions impact people’s careers, well-being, and financial security. If AI provided incorrect or misleading information, it could create serious issues. This was especially clear during testing, when I realized the GPT I’d built was mixing up regional benefits information giving teammates the wrong details depending on their country.

It wasn’t ideal! We needed to ensure accuracy to truly support our global team. It was important to us that we empowered new hires by providing instant, accurate responses about company policies, benefits, and processes.

To avoid overloading the GPT and confusing users, I experimented with moving away from bulk uploading all documents into a single model. During testing, I found that separating the content into topic-specific GPTs significantly improved clarity and accuracy, especially when handling region-specific information. Based on those results, I’ll be building out the approach using dedicated GPTs for each topic area moving forward.