Jotform, VP of Product AI, Berkay Aydın’s Exclusive Interview with MarTech Pulse on Jotform AI

In an exclusive interview with MarTech Pulse, Berkay discusses how Jotform AI, AI App Builder, intelligent workflows, and agentic capabilities are reshaping no-code product experiences while keeping reliability, user control, and trust at the center.

These days, when we say AI, we usually think about LLMs. At Jotform, we always had different projects based on traditional machine learning, but in December 2020, we got early access to the GPT-3 beta through OpenAI, and we organized a hack week around it.

At that time, it was only a completion API, but it was still fascinating to me. That hack week was probably the moment I realized this was going to become much bigger than another technology we experimented with. I decided to dedicate some of my personal time to investigating generative AI, and in almost six years, the technology has moved incredibly fast.

You now oversee AI strategy across a monorepo with 400+ packages serving 35+ million users. How do you balance rapid AI experimentation with the reliability users expect from Jotform’s core form-building tools?

At Jotform, we move really fast, with hundreds of updates going to production every day. There are two sides of AI in our development cycle: AI we use ourselves, and AI we give to our customers.

We use AI across all kinds of tasks, including development, and AI tools have significantly improved our speed. If we have an idea we want to experiment with, we can build it quickly and open it to a small number of users. We were already doing this before AI through methods like prototypes and user testing. The difference is that with AI in our daily work, these cycles have dropped from weeks to days, and sometimes even hours.

But our users expect reliable products, so getting something into production is a very different problem. We closely monitor usage, feedback, and failures and continue improving from there.

When we put AI directly into the hands of our customers, the bar is much higher. Users can face edge cases and complex situations you never expected, so we use a lot of guardrails and testing, both with AI and humans. We also review sampled AI interactions and failure patterns to understand where the product succeeds and where it fails.

So we experiment aggressively, but we’re much more conservative about what becomes part of the core experience. We use evaluations, testing, fallbacks, observability, user-research and gradual rollouts before giving AI more responsibility.

One principle I like is: AI can be probabilistic, but the product shouldn’t feel probabilistic.

The Jotform AI App Builder launch signals a bigger shift. How does it fit into your broader 2026 vision for weaving AI through Jotform’s existing suite rather than launching it as a standalone product?

Jotform AI App Builder lets people create no-code, ready-to-use web and mobile apps. You might have a small company to run, or you might simply have a wedding and want to create an app for it.

AI opened a door that we had wanted to open for years. Instead of requiring users to understand every part of a builder, they can simply describe what they want (through written prompt), or talk directly with our AI using voice, and AI translates that intent into a working Jotform product.

That’s why AI App Builder is bigger for us than just an AI feature for Apps. It represents how we want people to interact with the entire Jotform platform.

I don’t really think about AI as another product in our suite. I think about it as a new interface to the entire Jotform platform.

I also believe Jotform’s capabilities should exist wherever our users already are. That’s why we’ve been bringing Jotform Apps into platforms like ChatGPT and Claude as well. Users shouldn’t always need to come to Jotform first. They should be able to create and work with Jotform products from the tools they’re already using.

Looking at 2026, how will AI features extend beyond app building into Jotform’s broader suite – forms, tables, agents, and workflows – to create one connected AI-powered ecosystem?

It would be easy to just put an AI button into every product.

The real opportunity is getting the products to work together around the user’s intent. A user might say, “I need an employee onboarding process.” From that one goal, Jotform should eventually be able to create the form, structure the data, build the approval workflow, prepare documents, and create an AI Agent that can answer questions or help people through the process.

Forms collect information. Tables organize it. Workflows move the process forward. Agents interact with people. AI can become the layer that understands the goal connecting all of them.

We have Podo, Jotform’s AI assistant and mascot, which is becoming the interface connecting these product experiences.

That’s where I think this gets really powerful. You stop thinking about which Jotform product you need to open and start thinking about what you want to accomplish.

And the good part is that you can even talk with Podo. You don’t need to only type something and wait for a response. You can interact naturally with voice and get things done while you’re on the go.

Which existing Jotform product do you expect AI to transform most dramatically in 2026 – forms, tables, or e-signatures – and what would that enhancement actually look like for a daily user?

You mentioned forms, tables, and e-signatures, but I would actually pick Jotform Workflows.

One thing we’re always proud of is that Jotform is the easiest online form builder to use. We hear that from our customers all the time.

But once you have hundreds of forms managing business or personal processes, things naturally become more complicated. With Jotform Workflows, you can build things like approval processes or recruitment workflows, but creating and managing workflows is naturally harder than creating a simple form.

So this year, I especially think AI will dramatically reduce workflow creation time.

Instead of manually setting up every step, connection, condition, and approval, users should increasingly be able to describe the process they want and let AI create much of that structure for them.

For users, that means going from describing a business process to having a working workflow in minutes instead of building every step manually.

Rather than building AI as a separate layer, Jotform seems to be embedding it directly into existing workflows. What’s the product logic behind enhancing what already exists instead of building net-new AI-first tools?

Because the value isn’t AI by itself. The value is AI combined with the context, data, permissions, and workflows users already have.

If someone has spent years building forms, collecting submissions, creating approval processes, and connecting their tools inside Jotform, asking them to start again inside a completely new AI product doesn’t make much sense.

There’s also a very practical product advantage. Existing Jotform products already contain the user’s context: their forms, submissions, workflows, permissions, integrations, and history. AI becomes much more useful when it can work with that context instead of asking the user to recreate everything inside a separate AI product.

And then the product logic becomes pretty simple: we watch how customers use that AI, what they ask it to do, where it succeeds or fails, and improve upon it from there.

AI becomes much more useful when it’s connected to the work users are already doing.

If 2026 is about weaving AI into Jotform’s existing suite, what’s the next frontier for 2027 – do you see AI agents eventually managing entire workflows end-to-end without human Setup?

Agents will definitely take on more and more responsibility.

For some tasks, our AI can already work with very little human setup, such as researching or analyzing information. But I’m less convinced about the “without human setup” part, especially for complex operations.

I think humans will remain in control for a long time.

For example, AI can handle a very large percentage of customer support cases, but you still need to figure out what happens with the remaining cases it cannot resolve. And even if AI eventually performs almost every step of an operation, humans will still need ways to review those actions and make sure everything is working as expected.

So I see the future as AI doing much more of the execution, while humans define the goals, constraints, and level of control.

For product leaders navigating AI adoption inside legacy platforms, what’s one piece of advice you’d give on balancing innovation speed with the trust users place in an established product?

My main advice would be to build a very simple version and create the tightest feedback loop you can with real users.

Look at what users requested, what your AI actually did, where it failed, and focus your teams on improving those gaps every day.

With daily cycles, you get speed. By continuously listening to users and observing the actual AI behavior, you build trust.

Don’t overengineer the first version, but remember that people have much higher expectations from AI than from traditional software. They expect it to understand them, adapt, and handle situations you may never have explicitly designed for.

And don’t add AI just because users expect an AI feature.

Use it where it removes real work.

Write to us [⁠wasim.a@demandmediaagency.com] to learn more about our exclusive editorial packages and programmes.