Project 04 / 05

PITCH AI

Role
Frontend Team Lead
Year
2024–2026
PITCH AI interface
Overview

AI-powered pitch intelligence. Sell smarter.

This platform gives sales teams AI-generated pitch intelligence drawn directly from their CRM data. Decks that adapt to each prospect's context, real-time collaboration on live pitches, and engagement analytics that tell you which slides actually close deals.

Next.jsTypeScriptAI APIsCRM integrationsWebSocketsZustand
Platform
B2B SaaS · Web
Intelligence
AI pitch generation from live CRM context
Collaboration
Real-time multi-user deck editing
Analytics
Slide-level engagement tracking
Team
Frontend team of 4, led as team lead
The Problem

Pitch decks were static. AI was disconnected.

Static decks

The same deck goes to every prospect. No adaptation to company size, industry, pain points, or deal stage. A pitch for a 10-person startup looks identical to one for an enterprise account, and both underperform because neither feels personal.

Disconnected AI

AI writing tools produce generic output because they have no access to deal context. They don't know what the prospect cares about, what's already been discussed, or what objections have come up. The suggestions feel hollow because they are.

No feedback loop

Teams had no visibility into which slides prospects engaged with, which got skipped, and which triggered follow-up questions. Without that data, there was no systematic way to improve the pitch. Every rep was flying blind.

Approach

Context-aware AI. Real-time teams. Data that closes deals.

01

Context-aware generation

The AI pulls from live CRM data (deal stage, company size, rep notes) before writing a word. Every suggestion is grounded in the actual deal, not a template.

Technical note

CRM data normalized at ingestion: Salesforce, HubSpot, and Pipedrive fields mapped to a common schema. Per-deal context window includes contact history, opportunity notes, and previous proposal versions. Suggestions streamed via SSE for perceived speed; partial results render as they arrive.

02

Real-time collaboration

Multiple reps and managers work on the same deck simultaneously: changes propagate instantly, comments and approvals happen in-context, no version conflicts.

Technical note

WebSocket-based operational transforms for conflict-free concurrent editing. Presence indicators show who is on which slide in real time. Role-based permissions: editors, viewers, approvers. Full undo/redo stack with per-session history. Offline edits queued and replayed on reconnect.

03

Engagement analytics layer

Every shared pitch is instrumented: time per slide, return visits, click heatmaps. Teams see exactly where prospects engage and drop off; the AI uses that signal to improve future decks automatically.

Technical note

Anonymous tracking via edge function: zero overhead for the viewer. Events streamed to a time-series store, aggregated per deal. Heatmap view per slide, funnel visualization per deck. AI recommendation engine fine-tuned weekly on aggregate engagement signals across all pitches.

Feature 01

AI Pitch Generator

The AI reads the deal's CRM history before it writes a word. Suggestions are specific to this prospect, this stage, this rep. Not a template filler, but a context-aware co-writer that knows what's already been said.

CRM contextSSE streamingAI APIs
AI Pitch Generator
Deal Rooms
Feature 02

Deal Rooms

A shared space for each deal: deck, comments, approvals, and version history. Multiple stakeholders on both sides. Real-time presence, concurrent editing, and role-based controls. No email chains, no "latest version" confusion.

WebSocketsOT conflict resolutionRole permissions
Feature 03

Engagement Analytics

Time per slide, return visits, drop-off points, and click heatmaps, tracked without a SDK install. The AI surfaces insights automatically: which slides hold attention, which lose it, and what to change before the next pitch.

Edge instrumentationTime-seriesAI insights
Engagement Analytics
AI
Context-aware pitch
RT
Real-time collab
3x
CRM integrations
4
Team size led