August 8, 2025
5 min read
Aytekin Tank
3 Lessons Learned From Building Our First AI Agent
Having led a business for nearly 20 years, I’ve overseen many product launches—and still learn something from every single one. Recently, my company, Jotform, ventured into entirely new territory with the release of our first AI agents. Our journey into agentic AI began in 2023, well before "AI agent" became a common term. The last few months have been filled with hype—from grand promises about AI agents’ disruptive potential to fears about their impact. Regardless of the discourse, AI agents are reshaping how business is done. Here are three lessons we learned from creating ours.Listen To Data, Not Hype
Like any product, your original idea may transform significantly by launch. When Jotform first explored AI integrations, generative AI was the hot topic. During one of our regular hack weeks, our team developed an idea: AI-powered forms users could generate and customize by interacting directly with the form. We launched a beta version and gathered real-world feedback. It was a hit—users engaged with the product and retention was strong. But data revealed an unexpected insight: 90% of users weren’t using the tool for forms; they used it for customer service. This was not our initial vision, but the data was clear. We pivoted to support this use case. The lesson? Success depends on your ability to listen to customers and adapt, even if it means changing your original plan. The data never lies.Efficiency Over Friendliness
AI companions have sparked concerns about replacing human relationships. This might lead leaders to think their agents need charming personalities to succeed. However, users care more about efficiency than personality. Historically, customers preferred humans over chatbots due to frustration with the latter. Our testing showed users actually prefer agents—not just for faster responses and 24/7 availability, but because agents remove social pressure. Users don’t worry about asking “dumb” questions or repeating themselves. Without the need to be pleasant, they get the help they need. Research from Salesforce supports this: overly human-like AI agents can disappoint users and erode trust if expectations aren’t met. Yvonne Gando, Salesforce’s senior director of UX, explains it’s better to keep AI helpful and approachable without creating false human connections. In short, make your agent helpful and pleasant, but don’t try to make it human.Reach Your Customers Where They Are
We realized the importance of integrating AI agents across multiple channels—websites, phone lines, messaging apps like WhatsApp, and more. Their 24-hour availability and multilingual capabilities break down barriers that traditionally prevent users from getting answers, making agents accessible. Knowing your customers and meeting them on their preferred platforms ensures a cohesive experience no matter how they interact with your AI agent.AI agents aren’t a fleeting trend—they’re transforming customer engagement. By listening to data, prioritizing efficiency, and meeting users where they are, companies can build AI agents that truly add value. Like any product, maintain flexibility and adapt to changing realities. You won’t regret it.
Originally published at Forbes on August 7, 2025.