Jotform AI Agents tested: do they beat a form?

A static form has one obvious flaw: hit a question you don't understand and it's easy to just close the tab. Jotform AI Agents try to fix that by replacing the form with something closer to a conversation.

Instead of a wall of empty fields, the agent talks to whoever's filling it out — answers questions about your business, pulls the same information a form would need, and in some setups even takes a payment mid-chat.

Setting one up

Building an agent doesn't take any coding. The whole process is conversational, which is a little on the nose but does work:

1. Describe what it should do: something like "Create an intake agent for a residential solar installer that qualifies homeowners, checks roof viability, and books a site visit."
2. Feed it a knowledge base: FAQs, PDF brochures, price sheets, or URLs. It leans on these to answer questions instead of guessing.
3. Map it to your form fields: as the conversation goes, the agent pulls out the relevant pieces — name, address, current energy bill — and drops them into your backend.
4. Deploy it wherever your customers already are: a website widget, or straight into WhatsApp, Instagram DMs, SMS, or email.

Where it held up in testing

Two things stood out as genuine strengths rather than just a novelty, both tied to situations where a static list of fields tends to lose people.

1. Leads who don't know the technical answer off the top of their head. In something like home improvement or custom design, clients rarely know the exact spec upfront. Ask "What type of electrical panel do you have?" on a static form and a lot of people just bail, either guessing wrong or abandoning the form entirely rather than admitting they don't know. The agent, instead, can say: "No worries, check the main breaker box near your garage. Does the main switch say 100A or 200A?" and adjust its next question based on the answer. That's the real difference: it can coach someone through a question instead of just displaying it and hoping they figure it out, which in testing noticeably cut down on the "I don't know" and blank-field answers a static version of the same form collected.
2. Meeting people where they already are. For customers who live in WhatsApp or Instagram, sending them to an external browser form loses people at the link, every extra tap away from the app they're already in is a chance to lose the conversation entirely. Running the agent directly inside WhatsApp keeps the whole booking or intake conversation in the app they're already using, without asking anyone to open a browser, create an account, or even leave the chat thread they started in. For a small business whose leads mostly arrive as a DM anyway, that's a meaningfully lower-friction path than any web form, however well-designed.

Where a regular form still wins

None of this makes the conversational format a universal upgrade. A few things a plain form still does better, and did better in every side-by-side test run for this piece:

• Speed: someone filing an expense report or a nurse logging vitals gets through a structured form far faster than they'd get through a chat, because they already know exactly what information is needed and just want to enter it without a conversational back-and-forth slowing them down field by field.
• Reviewing what you entered: scrolling back through a chat history to check or fix a value is a lot more annoying than glancing at a multi-page form, where every answer is visible at once and a mistake two sections back is a click away instead of a scroll through transcript.
• Bulk or repetitive entry: anyone submitting the same type of data over and over, a daily site log, a recurring order, gets tired of a conversational tone fast; a form that remembers the last submission's structure and just wants updated numbers is a better fit than a bot re-asking the same pleasantries every time.

Bottom line after using it for a few weeks

What it costs to run one

The pricing question is the one most coverage skips, and it changes the calculation. An AI Agent conversation consumes against your plan's quota in a different way from a form submission — a completed conversation is a submission, but the back-and-forth that produces it is where the cost sits, and a chatty respondent is more expensive than a form filler.

The practical planning note: a conversational agent is a poor fit for high-volume, low-value intake precisely because the per-interaction cost is higher and the per-interaction value is not. It earns its place on the things where a form loses people — a complex quote, a qualification conversation, an enquiry that branches ten ways — and those are lower in volume and higher in value by definition.

Where we would not use one

Four cases, from the fortnight of testing.

• Anything regulated. A conversational agent paraphrases. A consent form, a medical screen or a legal disclosure has to present specific wording and record that it was shown. Use a form, which is a document.
• Anything where the respondent already knows the answers. A returning customer placing a repeat order does not want a conversation; they want eight fields and a submit button. Chat is slower when the friction was never comprehension.
• Anything on a bad connection. A form loads once and can be completed offline. A conversation needs a round trip per turn, and in a basement or on a rural site it simply stops.
• Anything where you need clean data. Free-text answers interpreted into fields are right most of the time, and "most of the time" is a reconciliation problem when the field is a postcode or a quantity.

Where it genuinely won was the long qualification form — the one with twenty questions where only six apply to any given person. Letting somebody describe their situation and having the agent ask only what follows from it is a better experience than a form with aggressive conditional logic, and the completion difference was the clearest result of the test.

Jotform AI Agents work better as a sales and support layer than as a form replacement. For top-of-funnel leads and anything that benefits from a bit of hand-holding, they noticeably cut down on how many people give up halfway through. If you're already running Jotform for forms, the lowest-risk way to test the format is turning on an agent for whichever single intake flow has the worst drop-off rate today, rather than replacing every form on the account at once and finding out which ones needed to stay static.