Problem
Service businesses collect intake through scattered forms and inboxes, then spend time manually reading, triaging, and chasing each lead. The signal - who to call first, what they need, and how urgent it is - gets buried in raw responses.
Selected work
Owned product · 2025 – Present
FormPilot
An AI-assisted form builder for service businesses: publish intake forms, persist validated submissions, then asynchronously turn each response into a structured, actionable summary through the OpenAI Responses SDK.
01 Scope
The product context behind the work.
My contribution
- Designed the build -> share -> collect -> summarize -> act flow end to end
- Built the form builder, public form rendering, and server-validated submission pipeline
- Implemented persisted-first AI summaries: after validation and storage, an asynchronous OpenAI Responses SDK job performs schema-validated extraction of lead intent, urgency, and budget
- Set up workspace auth, plan entitlements, billing, and integrations
02 Approach
Important engineering decisions.
FormPilot starts from one operating idea: the form is the easy part. The value is everything that happens to a submission after it arrives. The product is shaped around the full path from a published intake form to a decision the business can act on.
The build splits into clear workflow moments. A team can create a form, publish it to a clean public URL, and collect server-validated submissions into a dashboard that makes each response easier to review, filter, and move through.
The differentiator is the AI summary layer. Each server-validated submission is persisted before an asynchronous job uses the OpenAI Responses SDK for schema-validated extraction. That scoped path surfaces the lead's intent, urgency, budget, contact details, and useful tags so the owner knows who to follow up with first.
Around the core sit the product details that make it feel operational rather than just a demo: workspace auth, plan limits, billing, and integrations that help submissions move into the systems where a service business already works.
03 Evidence
Product screens from the shipped flow.
04 Delivered & verified
What the public proof supports.
Delivered scope
Workflow
Collapsed intake, triage, and follow-up into one connected flow
Time to signal
AI summaries surface lead intent and urgency without reading full submissions
Scope
Shipped builder, public forms, dashboard, AI summaries, billing, and integrations
Stack
Next.jsTypeScriptClerkNeon / PostgresPrismaOpenAIStripeTailwind CSS
Reflections
- The implementation uses the OpenAI Responses SDK directly for persisted-first structured extraction; it does not use RAG or the Vercel AI SDK.
- Most of the product value comes from the workflow around the form, not the form field editor alone.
Next step
Have a similar project?
If this looks close to the product work you need, book a short call and I will help scope the next step clearly.