Skip to content

brainstorm.md

Latest commit

 

History

History
142 lines (102 loc) · 5.01 KB

brainstorm.md

File metadata and controls

142 lines (102 loc) · 5.01 KB
Raw

Formless — Brainstorm & Ideation

Saved from initial planning session — 2026-03-12

Origin

Started from a prototype called nlp-cms (Dec 2024) — a proof of concept showing that natural language input could be classified by an LLM and mapped to validated, typed JSON objects (events, articles, notes) using Zod schemas.

The core loop:

natural language input
  → AI classifies type (event | article | note)
  → AI extracts structured fields
  → Zod validates against typed schema
  → clean JSON stored in DB

The narration written for the prototype demo said it best:

"Bridging the gap between how humans naturally communicate and how computers need data to be structured."

The Core Insight

Forms are a hack. They exist because computers can't understand intent. With LLMs, they don't have to.

WordPress's genius was abstracting HTML into posts/pages/fields. Formless abstracts intent into structured data — replacing forms with conversation.

Initial Brainstorm (Three Angles)

Angle 1: No-Code AI Site Builder (Path B — future)

Target: Small businesses, freelancers, non-technical founders

  • You describe your site in plain English
  • AI builds pages, content model, and layout
  • You edit via chat: "Make the hero headline more urgent", "Add a testimonials section"
  • Publish to static hosting (Vercel/Netlify) in one click
  • Closest comp: Webflow + Framer, but conversation-first

Angle 2: AI-Powered Headless CMS (Path B extension)

Target: Dev teams who want Contentful/Sanity but faster schema design

  • Describe your content model in plain English → AI generates the schema
  • Chat with your content: "Find all blog posts about React published this year"
  • AI auto-generates migrations when schema changes
  • Delivers a REST/GraphQL API
  • Closest comp: Sanity + Contentful, but AI-native schema design

Angle 3: Enterprise Form Replacement (Path C — THIS PROJECT)

Target: Internal ops/IT teams drowning in form maintenance

  • Replace internal forms, ticketing systems, intake workflows with natural language
  • "Ticket systems, document management, knowledge bases — anywhere we need structured data input"
  • Zero training required for end users
  • Closest comp: ServiceNow, Jira forms — but conversation-first
  • Revenue: $500–$5000/mo per team

Why Path C First

  • Highest ACV (annual contract value) of the three paths
  • Clearest pain: ops/IT teams maintain dozens of brittle forms
  • Fastest adoption signal: replace ONE internal form, prove ROI immediately
  • Enterprise buyers pay before you're perfect
  • Defensible: schema enforcement + validation = not just ChatGPT in a box

What Makes Formless Defensible vs. Just Using ChatGPT

The critical differentiator is the schema enforcement + validation layer:

  1. Guaranteed structure — validation means output always fits the schema, or it fails gracefully
  2. Queryable data — typed JSON means you can filter/sort/aggregate like a real database
  3. Audit trail — every AI action produces a diff, not a black box
  4. Idempotency — same input reliably produces the same schema shape
  5. Integrations — connects to existing enterprise systems (Jira, ServiceNow, Slack, email)

ChatGPT gives you prose. Formless gives you a structured, validated, API-ready data record.

The Conversational Edit Loop (Key Unsolved Problem)

The prototype handles: input → structured output. A real product also needs editing:

  • "Change the priority to urgent" → patches only priority field
  • "Assign this to the infrastructure team" → updates assignee field
  • "This is actually a billing issue, not a tech issue" → reclassifies type + re-validates

This requires stateful conversation with schema awareness — the AI needs to know:

  1. The existing record
  2. The schema constraints
  3. The patch to apply (not a full rewrite)

This is the core R&D challenge of the product.

The Bigger Vision (2–3 year horizon)

"Describe your data. We'll build the database."

Any domain where structured data input exists today can be replaced:

  • IT helpdesk tickets
  • HR onboarding/intake
  • Legal contract intake
  • Insurance claims
  • Patient intake (healthcare)
  • Customer support tickets
  • Procurement requests
  • Incident reports

The schema becomes the product. The natural language is just the interface.

Competitive Landscape

Competitor Weakness Formless Exploits
ServiceNow Expensive, complex, requires consultants
Jira Forms Developer-centric, not conversation-native
Typeform Still forms — just prettier ones
Google Forms Zero intelligence, manual structure
Microsoft Forms + Copilot Surface-level AI, not schema-native
Notion AI General purpose, not structured data focused

Revenue Model Ideas

  • Per workspace/seat: $500–$2000/mo for team plans
  • Per schema/workflow: $X per active form replaced
  • Enterprise contract: Annual deal, custom pricing
  • Usage-based: Per AI parse/classification above a free tier