Vective

Evidence first AI-diligence copilot for VCs and Angels

Comment 1

🚀 Vective

The Evidence-First Diligence Copilot That Remembers Everything and Verifies Every Claim

Vective turns a pitch into a verified IC-ready memo with proof behind every line.
It ingests pitch calls, email threads, Notion notes, decks, data rooms, and web research, stores it in long-term memory via Supermemory, and runs Browser Use Skills to verify claims on sites that do not have APIs.

Rule: No claim without proof.
Outcome: Faster decisions, fewer misses, and diligence that compounds over time.

🔥 Inspiration

The idea for Vective came from a blunt moment at Drive Capital.

I was speaking with a VC partner and they went on a real rant. Not about market risk or valuation. About something more basic and more painful: the diligence process itself.

They described the same loop they live through on every deal:

  • A pitch call happens.
  • Notes live in Notion.
  • Decks are in email.
  • Data rooms are separate.
  • Partners ask questions in Slack.
  • Research lives in 50 browser tabs.
  • The memo is stitched together manually.

And then they said the part that hit me:

“We keep redoing the same diligence. We keep losing context. And half the time, the final memo still doesn’t have proof.”

That is the problem. Not intelligence. Not effort. Fragmentation.
So we built Vective to fix it.

🧠 The Big Problem (Why Vective Exists)

Investors run diligence across pitch calls, email threads, Notion notes, decks, data rooms, and random web tabs.
The same checks get redone every deal, key context gets lost, and AI summaries are not trusted because they do not show evidence.
That makes diligence slow, inconsistent, and risky, so deals stall and red flags slip through.

What this creates in real life

  • Deals take weeks not because teams are slow, but because the workflow is broken.
  • Analysts spend hours on repetitive checks instead of real judgment.
  • Founders get asked for the same info multiple times.
  • Partners lose confidence in “AI summaries” because they cannot audit them.
  • Funds lose institutional memory. When a founder returns 6 months later, they start from zero.

Vective is the opposite: one place to remember everything, and one system to verify everything.

✅ What Vective Does

Vective converts messy deal inputs into a clean, evidence-backed output.

Inputs

  • Pitch call transcript or audio
  • Pitch deck
  • Data room docs
  • Email threads
  • Notion pages and notes
  • Links and web sources

Core engine

1) Deal Memory (Supermemory)

  • Stores everything across deals
  • Builds a timeline of claims, documents, and decisions
  • Makes it searchable instantly

2) Claim Extraction

  • Pulls structured claims like:
    • revenue, growth, customer count
    • partnerships, launch dates, traction signals
    • security and compliance claims (SOC 2, audits)
    • team size, hiring claims, technical credibility

3) Verification Skills (Browser Use)

  • Runs reusable web workflows as APIs
  • Captures screenshots, citations, timestamps
  • Verifies claims on websites that have no public API

Outputs

  • IC-ready memo
  • Red flags list
  • Open questions + missing docs checklist
  • Evidence Ledger (Proof Pack): citations, screenshots, and timestamps for every key point
  • Follow-up drafts: emails and checklists to close the loop

🧩 How We Built It (System Overview)

1) Authentication + Roles

  • Investor accounts and business accounts
  • Secure auth using JWT
  • Email verification, password reset, profile management

2) Integrations (Composio)

We use Composio OAuth to connect and access data from:

  • Gmail
  • Slack
  • Notion
  • Google Calendar
  • Google Sheets
  • HubSpot

This lets Vective pull real deal context without forcing teams to change tools.

3) Transcription + Q&A

  • Audio transcription using OpenAI Whisper
  • Store transcripts in memory
  • Ask questions like:
    • “What revenue did they claim?”
    • “What did they promise to send?”
    • “What changed since last call?”

4) Verification Skills (Browser Use)

We created high-signal Skills like:

  • founder verification
  • company footprint and start date
  • domain age and security signals
  • competitor pricing screenshots
  • customer sentiment sweep
  • SOC 2/compliance presence checks
  • GitHub activity checks (when relevant)

5) Evidence Ledger + Memo Generation

The Evidence Ledger is the backbone. Every major memo point includes supporting artifacts:

  • cited sources
  • screenshots
  • extracted snippets
  • timestamps

⚔️ Challenges We Faced

1) Trust is the bottleneck

AI summaries are cheap. In diligence, cheap summaries are worthless if they are not auditable.

We had to design Vective around a hard constraint:

If we cannot point to evidence, we do not say it.

2) Speed vs depth

Verification can take time, especially across many sites. So we structured the experience as:

  • quick first memo
  • live evidence upgrades
  • parallel execution of Skills

3) Integrations and permissions

OAuth is easy to demo, hard to make reliable. We focused on a stable, minimal set of connectors that map to real diligence workflows.

📚 What We Learned

  • The bottleneck in diligence is not intelligence, it’s fragmentation.
  • The only AI investors trust is AI that can prove its work.
  • Memory is the real moat. Diligence should compound, not restart every deal.

🧮 Math: Why Vective Feels Fast

Let \(t_i\) be the time for each verification Skill.
If Skills run in parallel, total verification time is roughly:

$$ T \approx \max(t_1, t_2, \dots, t_n) $$

Instead of waiting for tasks sequentially, Vective executes them concurrently and upgrades the memo as results arrive.

For trust, each claim must satisfy:

$$ \text{Claim is usable} \iff \exists\; \text{evidence} \; (\text{source}, \text{screenshot}, \text{timestamp}) $$

🛠️ Built With

Languages & Frameworks

  • TypeScript
  • Python
  • Next.js 16
  • React
  • Flask

AI + Memory

  • OpenAI GPT (analysis, memos, reports)
  • OpenAI Whisper (transcription)
  • Supermemory (persistent deal memory, retrieval, knowledge graph)

Automation + Integrations

  • Browser Use Skills (web workflows as APIs)
  • Composio (OAuth integrations: Gmail, Slack, Notion, Google Calendar, Google Sheets, HubSpot)

Data + Infra

  • SQLite (default)
  • Firestore (optional)
  • Plaid (bank linking)
  • ScreenshotMachine (web screenshots)

🎬 Demo Flow (What to Try)

1) Upload a pitch deck + transcript (or record and transcribe) 2) Run verification Skills on key claims 3) Open the Evidence Ledger to see proof artifacts 4) Generate an IC-ready memo 5) Generate a follow-up email draft + checklist

Vective turns messy deal work into verified decisions.

Share this project:

Updates