FormFlow

I built something I’ve been wanting to exist for years. FormFlow is a tool I’ve been developing that lets you record any conversation — a phone call, a Zoom meeting, a desk-side chat — and have AI automatically extract the important details into structured form fields. In real time.

In this demo, I’m running a simulated sales discovery call using AI-generated voices. As the conversation covers pain points, budget, timeline, and current stack — FormFlow is listening. It’s pulling out each of those data points and dropping them into the right fields on a fully customizable form template. No copying. No post-call note scrambling. No “let me circle back with what we discussed.”

The way it works: the audio is transcribed in small batches using Whisper, and every few seconds GPT-4 analyzes the rolling transcript against the form template’s extraction instructions. Each field has its own prompt telling the AI exactly what to look for. That means a “Budget” field knows to look for dollar amounts and financial context. A “Decision Maker” field knows to listen for names and titles.

The result is that by the time your call ends, your CRM data is already done.

This isn’t theoretical. This is production software running in a browser. No plugins, no downloads, no Twilio integration required. Just your mic and a browser tab.

If your team is spending 15–20 minutes after every discovery call typing up notes, that math adds up fast. FormFlow gives that time back.

Every team has the same meeting problem: great discussion, mediocre documentation.

In this demo I’m running a simulated project planning call through FormFlow. The AI-generated conversation covers scope, milestones, task assignments, blockers, dependencies — a typical kickoff or sprint planning conversation. FormFlow is extracting it all into structured boxes: an ordered list of milestones, action items as an unordered list with owners, project risks as tagged pills, timeline as structured date fields.

When the meeting ends, I have a couple options for what happens next.

First — PDF export. FormFlow generates a multi-page report with a cover page showing session metadata (date, duration, participants), followed by the organized content from each box. Transcripts are color-coded by speaker. It’s a professional deliverable, not a raw brain dump. Send it to stakeholders who weren’t on the call and they get the full picture.

Second — webhook integration. The structured data fires off to your project management tool. Task titles and assignees go to Jira or Azure DevOps. Meeting notes go to Confluence or Notion. Risk items go to whatever risk register your PMO uses. The payload is clean JSON with every field labeled and typed.

Third — the session itself is saved and searchable. Go back to any previous planning session, review what was discussed, see the full transcript, and compare it against what actually got done.

The thing I kept running into before building this was the gap between “we talked about it” and “it’s tracked somewhere.” That gap is where things fall through the cracks. FormFlow closes it by making the conversation itself the data entry mechanism.

Your meetings already contain the information. You just need something smart enough to pull it out.

Healthcare has a documentation problem that everyone knows about and nobody has solved cleanly.

In this demo I’m walking through a simulated patient intake call using FormFlow with AI-generated voices. The conversation covers symptoms, medical history, current medications, allergies, insurance info — the usual intake workflow. But instead of a nurse or admin staff member toggling between an EHR and a phone, FormFlow is handling the structured capture.

What makes this work is the template system.

FormFlow templates are built with “boxes” — modular containers that each have a type and an AI extraction prompt. For patient intake, you might set up:

A “Chief Complaint” box (plain text, extracted from how the patient describes their issue), a “Current Medications” box (unordered list, pulled from mentions of drug names and dosages), an “Allergies” box (pills/tags for quick visual scanning), and an “Insurance” box with structured input fields — policy number with an input mask, group number, provider name.

Each box tells the AI exactly what to listen for. The extraction prompts are fully customizable per field, so a “Date of Birth” field knows to look for dates in context, not just any date mentioned in the conversation.

There are 30+ field types with input masking — phone numbers, SSNs, dates, currency — so the data comes out clean and formatted. Not raw text that someone has to reformat downstream.

The template builder is drag-and-drop, no code. An office manager can set this up. And once a template exists, every staff member in the organization can use it. Templates can be scoped to personal use or shared across an entire organizational unit.

Documentation should happen during the conversation, not after it.

Hiring is one of those processes where everyone agrees the notes matter but nobody actually takes good ones.

This demo shows FormFlow running during a simulated job interview with AI-generated voices. The conversation covers behavioral questions, technical scenarios, culture fit questions — and FormFlow is extracting structured assessments in real time. Technical proficiency? Captured. Communication skills? Tagged. Red flags? Noted.

But here’s where it gets interesting beyond the single session.

FormFlow has a campaigns feature. You can create a form template — say, your standard interview scorecard — and send it out as a campaign to every interviewer on the panel. Each person gets their own secure link with a unique token. They can fill it out via voice mode (FormFlow asks the questions and they respond conversationally) or traditional form entry.

As responses come back, you get campaign-level analytics: open rates, completion rates, response rates, time-series data showing when people engaged. You can see at a glance that 4 of 5 interviewers have submitted their feedback and the fifth hasn’t opened the link yet.

All that data can fire to a webhook — meaning your ATS gets the structured feedback automatically. No chasing people for written summaries. No inconsistent formats across interviewers. Everyone uses the same template, the same scoring criteria, and the data rolls up cleanly.

FormFlow turns interviews from “we’ll get back to you” into “we already have everything we need.”

This applies to any workflow where you need structured input from multiple people: 360 reviews, incident reports, site inspections, compliance audits. Campaigns scale the collection.

One of the most common questions I get about FormFlow: “Cool, but how does the data actually get into our system?”

This demo answers that.

Here I’m running a simulated insurance claim intake call with AI-generated voices. The conversation walks through the details — date of incident, policy number, damage description, parties involved — and FormFlow is capturing and structuring all of it live.

But the real story here is what happens after the call ends.

Every FormFlow template can be configured with a custom webhook endpoint. When a form is saved or submitted, the structured data fires off as a JSON payload to whatever URL you point it at. Your claims management system. A ServiceNow ticket. A Salesforce record. A Slack channel. An internal API that kicks off an adjuster workflow. Whatever your stack looks like.

You configure the HTTP method, custom headers for auth, retry logic if the endpoint is flaky, and timeout thresholds. There’s SSRF protection built in so you’re not opening up security holes. Payload includes the form data, respondent info, and session metadata — all structured and ready for your downstream system to consume.

What this means practically: an insurance agent finishes a call and the claim is already in the system. No double-entry. No swivel-chair integration. No “I’ll file this after lunch” backlog.

The gap between conversation and system of record should be zero. That’s what this does.

This demo shows FormFlow handling IT support ticket creation by extracting structured data from a simulated help desk call with AI-generated voices. As the conversation covers the issue — device type, error messages, steps already tried, urgency level — the system captures and categorizes everything into a structured ticket format. By the time the call ends, the ticket is ready to route to the right team.