AI agent creates a Typeform demo request form with four question types
A demo request form sits between interest and a sales conversation. It needs to collect enough to qualify the lead without adding friction that kills completion rates. Typeform's one-question-at-a-time format helps, but question order and field type selection still matter. The video above shows Demosmith's AI agent building one from scratch — four questions, each using the right field type.
Step 1: Sign in and create a new form
The agent was given login credentials for a test Typeform account. It navigated to admin.typeform.com/login, entered the email and password, and landed on the dashboard. From there, it clicked "Create typeform" and selected "Start from scratch" instead of a template. Templates work for generic use cases, but a demo request form benefits from a controlled question sequence — you want low-effort qualifying fields before the open-ended question that takes more thought. The agent titled the form "Product Demo Request."
Step 2: Add a short text question for Company name
The first question uses Typeform's Short Text type for "Company name." Short text is correct here because company names are brief and don't need validation beyond basic input. Starting with this question is deliberate — it's low effort and immediately qualifies the lead. A long text field would invite unnecessary length, and a dropdown is impossible since you can't pre-populate every company name.
Step 3: Add a dropdown for Company size
The second question is a Dropdown for "Company size" with four options: 1–10 employees, 11–50 employees, 51–200 employees, and 200+ employees. Dropdown is the right field type because the options are mutually exclusive ranges — a respondent picks exactly one. The ranges map to typical sales segments: solo/startup, small team, mid-market, and enterprise. This lets a sales team route leads automatically based on the response without a human reading each submission.
Step 4: Add an email question
The third question uses Typeform's Email field type for "Work email address." The email type validates format automatically — it checks for an @ symbol and a domain, which reduces junk submissions without a separate validation step. Placing email after company info is intentional. By the time someone has entered their company name and size, they're committed enough to provide a real work address. Asking for email first increases drop-off.
Step 5: Add a long text qualifying question
The final question uses Long Text: "What problem are you trying to solve?" This is the qualifying question that gives context before a sales call. Long text is correct because the question asks for a multi-sentence response — short text would truncate useful detail. Placing it last is deliberate: it's the highest-effort field, so the respondent should already be invested before reaching it.
Step 6: Preview and verify
The agent opened the form preview to see it as respondents would experience it. Typeform's conversational format shows one question at a time, which is why question order matters more here than in a traditional multi-field form. The preview confirmed the dropdown options rendered correctly, the email field had validation, and the long text field gave enough space. The agent then returned to the editor.
How Demosmith generated this video
That's Demosmith's AI agent navigating Typeform — not a screen recording of a human. We gave it admin.typeform.com, login credentials for a test account, and the prompt above. It handled authentication, created the form, added each question with the correct field type and configuration, previewed the result, and returned to the editor. The final video includes captions. No screen recorder, no editing session.
This demo required the agent to handle a login flow, navigate a dashboard, interact with a multi-step form builder, select question types from a menu, configure dropdown options, and use the preview mode. Demosmith handles authenticated flows — you provide the credentials, and the agent signs in and runs the flow like a real user would.
