Adonis agentic sales intelligence dashboard — L3 AI Native Project.
- Backend Service: FastAPI service, hybrid rules-and-AI classification engine, Supabase database client, and Slack notification dispatches.
Sales intelligence automation for the Adonis hospital prospecting team. Pursuit L3 AI Native Demo Day · June 24, 2026 at Blackstone
An automated system that monitors a curated list of hospital prospects for buying signals — new revenue leadership, M&A, vendor changes, Epic go-lives, regulatory events — and delivers them to the Adonis team in three forms:
- Monday morning email digest to Danielle, who forwards to each AE
- Web dashboard with territory-filtered views per AE
- Persistent hospital profiles that accumulate signal history over time
The tool replaces Glean (the internal AI tool the team uses reactively, only before scheduled calls) with continuous, proactive monitoring. See docs/PRD.md for the full product requirements.
| User | Role |
|---|---|
| Danielle Ferdon | Admin view across all accounts; compiles and forwards the Monday digest |
| Michael | AE — territory-filtered dashboard, his accounts only |
| Jeff | AE — territory-filtered dashboard, his accounts only |
| David | AE — territory-filtered dashboard, his accounts only |
- NewYork-Presbyterian
- UMass Memorial
- Ascension
- University of Arkansas (UAMS)
- CommonSpirit Health
- Juan Franco — frontend, design system, dashboard UX, Iris AI chatbot
- Michael Chabler — partner communications
- Joel Philip — backend, agent pipeline, scoping, project management
Partner: Adonis Technology, Inc. · Primary contact: Reed Kalash · SME: Danielle Ferdon
adonisagent/
├── apps/
│ ├── web/ Next.js dashboard, hospital list, profiles
│ └── agents/ Mon/Wed/Fri scrape pipeline + Monday digest email
├── packages/
│ ├── db/ Supabase schema + query helpers
│ └── shared/ Types, signal taxonomy, seed data
├── docs/
│ ├── PRD.md Final product requirements
│ ├── ARCHITECTURE.md System design and data flow
│ ├── AGENTS.md Instructions for AI coding agents
│ └── ROADMAP.md Week-by-week plan
├── .github/
│ ├── workflows/ci.yml Lint + typecheck on every PR
│ ├── ISSUE_TEMPLATE/ Bug + feature templates
│ └── pull_request_template.md
├── README.md
├── CHANGELOG.md Week-by-week build notes
└── CLAUDE.md Pointer to AGENTS.md
Requirements: Node 20+, pnpm 9+, a Supabase project.
pnpm install
cp .env.example .env.local
# fill in env values - see comments in .env.example
# In your Supabase SQL editor:
# 1. run packages/db/schema.sql
# 2. run packages/db/seed.sql
pnpm devThis starts apps/web at http://localhost:3000 and the agent worker in watch mode.
Manual jobs:
pnpm --filter @adonis/agents scrape:once # one full pass through sources
pnpm --filter @adonis/agents digest:preview # render digest to stdout- Repo structure + tooling
- Supabase schema for hospitals, contacts, signals, users, digests
- Signal taxonomy from kickoff (urgent / standard / noise)
- Five seed hospitals loaded
- Three web pages (signal feed, hospital list, profile)
- Agent worker scaffold with cron scheduling
- Email digest renderer (HTML + plain text)
- Real scraping logic per source (hospital newsrooms, serper.dev, Becker's)
- LLM classification wired to Claude
- Digest delivery wired to Resend
- Per-AE territory filtering (Phase 2)
- Persistent hospital profiles with signal history (Phase 2)
- Co-pilot chat interface (Phase 3 stretch)
These do not move:
- No HubSpot connection. Hospital list comes in manually; output is a CSV they upload.
- Public sources only. No PHI, HIPAA-protected, or paywalled data.
- No LinkedIn scraping. Sales Navigator complements this tool; it does not replace it.
- No internal Adonis systems. Tool is fully standalone for the build window.
| Date | Milestone |
|---|---|
| May 19 | ✅ Danielle workflow walkthrough |
| Weeks 1–2 | Digest + dashboard live for 5 seed hospitals |
| Tue Jun 2, 3:00 PM ET | Midpoint review with Reed |
| Weeks 3–4 | Full dashboard, per-AE territory views |
| Week 5 | Refinement + co-pilot prototype if ahead |
| Wed Jun 24, 6 PM | Demo Day at Blackstone |
- All external email to Adonis goes through Greg first. Period.
- Internal: GitHub Issues for tracked work, Slack for coordination.
- Reed prefers email; 2–3 status emails/week + monthly meeting.
Private — Pursuit capstone project. Not for distribution.