12 hours ago
The Best Way To Improve Your AI Agent? Use It Yourself
Photo by Solen Feyissa on Unsplash
As an entrepreneur, my goal has always been to build products I want to use. Back when I was working at a media company in New York, I created a form builder that saved tons of time. I used it myself because it made my life easier—it also served as the basis for what would become my company, Jotform.
The world looks a lot different now than it did in 2006, when Jotform was born. If you'd told me then that in 2025, we'd release an AI agent capable of talking with customers in a human-like way, autonomously carrying out tasks, and customizing forms on its own, I probably wouldn't have been too surprised. Or rather, I'd be amazed by how much had changed—after all, this was back before the iPhone had even made its debut. But the idea that I'd be leveraging the latest technology to make tedious tasks simpler? That's right up my alley.
To this day, I believe my form builder became a hit because I used it myself—that meant I also experienced its bugs and imperfections, and saw opportunities for features I thought would make it better.
Fast forward to this year, when we released our new AI agents. Not only are our agents super helpful for our customers, but they're incredibly helpful for us, too. Of the 5,000 conversations our agents engage in daily, 2,000 are with Jotform support staff.
This high level of engagement throughout the company presents us with an amazing opportunity to improve our product. Here's why.
We Understand Areas For Improvement
As a business owner, I've always encouraged our teams to use our own products. After all, if we're not using it, it's hard to get a sense of the issues our customers are facing. While there were obviously occasions where we would use our own forms organically, we mostly had to contrive use cases ourselves.
With our AI agents, the opposite is true. We're using them not because we have to, but because they're a huge time-saver for our customer support staff. Using our agents as much as we are has been instrumental in allowing us to understand areas for improvement.
For example, in reviewing the conversations our agents are having, I've noticed its responses are good, but they're not always able to autonomously solve a problem without a human member of the support team intervening. If a customer has a localized technical issue—say their access was suspended and they need it restored, or they have a very specific question about their account—a human team member will ultimately still need to get involved to address the problem. Knowing this now, we're committed to dedicating an entire support team to making sure the agent can solve such problems on its own.
If you're not really using your own product and getting to know it, issues that arise are mostly theoretical. We do our best to address problems and add features our customers want, of course, but our understanding of their effectiveness is incomplete.
But by using our agents ourselves, on a daily basis, we really understand any pain points our customers are having—because we're having them, too.
We're Training Agents In Real Time
One of the greatest debates in the tech world now is the extent to which agents' actual utility will outweigh the hype. While there have been plenty of early adopters, the truth is that the general public isn't clamoring to try them out—for the most part, people just want their problems solved. According to data from Salesforce, 54 percent of consumers don't care how they're interacting with a company, as long as their problems are fixed quickly.
Our user base is still growing, but because we're using our own AI agents so much, we're simultaneously training them to be more effective. The beauty of generative AI is that it learns from every new query it tackles, improving upon and refining its answers continuously. That we're constantly engaging our agents with real data from real customer inquiries means our agents are getting better at their jobs with each passing day.
Our agents aren't perfect yet. No one expects them to be. But we're reviewing every single one of the conversations they're having, identifying problems and improving iteratively. We're carefully monitoring performance, identifying mistakes, tracking patterns and providing new learning data. As agentic AI gains traction and more users sign up, our AI will continue evolving, offering smarter, faster, and more accurate solutions while reducing the burden on human support teams.
We're Training Our Human Employees, Too
As AI takes over mundane tasks, our customer support teams will be able to use their time solving more interesting problems. But as agents get more adept at handling support on their own, our teams' responsibilities will shift to training, effectively upskilling them for a new AI-powered world. This not only helps them grow their own skills, it also keeps them energized and engaged.
Without a doubt, the best way to build a great product is to use it yourself and to keep refining it based on real-world feedback. By embedding AI agents into our own daily operations, we're not just improving the technology; we're shaping the future of customer support.
