ONLINEcascade://samwarren.iov0.1.0
all projects
POTENTIATING · V2 · UNIFICATION IN FLIGHT
SIGNAL CHAIN · TWO SURFACES· 03 / 06

Sales Intelligence Hub

Weekly digest + pre-call brief, one signal chain.

HomeDrop tells a rep what changed across their book this week. Intel Chat tells them what to do about one account right now. The unified product closes the loop between them — every digest card is one click from a seeded chat, every chat reshapes next Monday's digest.

7
DATA SOURCES
2
SURFACES UNIFIED
15 min → 90s
PRE-CALL PREP
targeting >25%
RELEVANCE LIFT
PROBLEM

Two tools that share a skin. Not a loop.

We shipped HomeDrop (the weekly Monday digest) and Intel Chat (pre-call prep, stateless) within a few weeks of each other. Same data sources, same users, same brand tokens. Different surfaces.

The failure mode was obvious from competitive research. It's what Gong, 6sense, and Demandbase all get wrong: two products sharing a skin, where reps eventually pick one and ignore the other because the tools don't know what each other did.

THE INSIGHT

Close the signal chain.

The best analog in the market is ZoomInfo Copilot's feed-with-inline-query pattern. Every passive signal card is one click from a scoped query surface. The framing to borrow from Clari: the active layer exists to explain the passive alert, not replace it.

One sentence for the team: HomeDrop tells you what changed across your book this week. Intel Chat tells you what to do about one account right now. The unified product makes each stronger by keeping context alive between them.

The active layer exists to explain the passive alert, not replace it.
ARCHITECTURE

One front door, two surfaces, a shared signal store.

Path-routed under a single subdomain via a thin Cloudflare Worker. One cookie jar. One access policy. Behind the routing: two existing deploys (the Python-generated digest and the Next.js chat) left intact. No Frankenstein rewrite.

The data flywheel is a single new table: signals_log(user_email, account_slug, signal_type, weight, occurred_at). Chat activity writes to it. The digest pipeline reads a 14-day rollup from it. Accounts a rep has been watching pin to the top of next Monday's digest, with framing that references the watching.

  • ·Week 1: gateway worker, unified access policy, cross-nav
  • ·Weeks 2–3: 'Prep me for this account' buttons on every digest card, URL-seeded chat context, 'Prep me for my 1:1' as a first-class chat prompt
  • ·Weeks 4–6: signals_log migration, entity extractor, digest rank boost, 'My watchlist' persistent view
  • ·Month 2+: Digest risk flags write back to the CRM, post-call intel ingested to signals_log
METRICS THAT MATTER

Five numbers that say the flywheel is working.

Before scaling past the pilot, the unification has to earn its keep. Five metrics, 12-week targets:

  • ·>70% of pilot reps open ≥3 of 4 weekly digests
  • ·>30% of account cards get ≥1 click per week (digest → chat deeplink)
  • ·Median rep runs ≥4 chat sessions per week
  • ·>40% of chat sessions start from a digest deeplink
  • ·>25% overlap between accounts flagged in the digest and accounts chatted about — that's the flywheel working
▸ components

What’s inside

HomeDropWeekly digest — what changed this week across your book
Intel ChatPre-call prep — conversational brief in under two minutes
▸ stack
next.js 15supabasecloudflare d1pythongithub actions
▸ clone it
internal · ETA next week
Weekly digest UI with 5 cards. One click seeds a streaming chat with that card's context. Shows how to close the loop between async signal delivery and on-demand conversation.
see /downloads →
▸ connects to
NEXT · 04 / 06
Target Account News Agent
A per-rep weekly digest, rebuilt for scale.