Question about Managing Latency on AI Tools

Hi everyone! Hope you're all doing well. I was wondering how makers of AI products deal with latency? At WorkHack, this is a big challenge for us, especially since we make something as critical as online forms - which everyone expects to be as fast as possible. We've come up with ways to manage the latency on the backend by making some specific architecture choices. We also are dealing with this by innovating on the wait time in the product itself. How do other makers manage this? Insights and feedback would be super valuable. Also, we're launching our first product tomorrow - the AI Form Roast. Its a free AI tool to audit and optimize online forms. Here's our teaser page: https://www.producthunt.com/products/workhack-form-tuner If this sounds valuable, we would greatly appreciate your feedback and be grateful for your support tomorrow. It's our first launch, and we are very excited to show makers what we've been working on :)