Inside Athenic Evidence Graph (Private Beta)
Athenic Evidence Graph stitches customer proof, research, and metrics into one searchable graph so agents deliver trusted answers fast.
MB
Max Beech
Head of Content
TL;DR
- Evidence Graph links every insight -customer calls, research notes, product telemetry -into a single knowledge graph that Athenic agents can reason over.
- The beta automates ingestion from Supabase, Notion, Slack, and CRM records, adds trust scores to each node, and surfaces proof clips inline.
- Private beta teams report 42% faster evidence retrieval and 28% fewer duplicated research requests during sprint planning (internal beta data, Jan 2025).
Jump to Why we built Evidence Graph · Jump to How the graph works · Jump to What beta teams are seeing · Jump to How to join the beta
Inside Athenic Evidence Graph (Private Beta)
Our customers asked for one place where agents, founders, and teams could trust the data behind every recommendation. Evidence Graph is our answer. It keeps proof portable, verifiable, and ready for agent workflows.
Key takeaways
- Evidence Graph is the connective tissue between your research agents, planning agents, and approvals.
- Every node carries provenance: source, timestamp, confidence, and related artefacts.
- You can query it directly in chat, or let agents auto-summarise proof in briefs, launch packs, and retention reviews.
Why we built Evidence Graph
Startups drown in scattered evidence. Product discovery calls sit in Notion. Support tickets live in Zendesk. Community nuggets disappear in Slack. Without a unified layer, teams repeat research and decision-making slows.
We designed Evidence Graph to:
- Centralise knowledge. One source that sits on top of your Supabase instance.
- Maintain trust. Every insight links back to the raw source.
- Power automation. Agents can reason across the graph to answer complex questions.
How the graph works
Evidence Graph ingests data through MCP integrations. Each record becomes a node with relationships (e.g., “supports”, “contradicts”, “references”).
| Component | What it stores | Example |
|---|---|---|
| Evidence nodes | Quotes, metrics, artefacts | “Pilot cohort activated in 36 hours” (Customer call, Feb 2025) |
| Entity nodes | Customers, features, use cases | "Community launch sprint" |
| Insight edges | Relationships between evidence | Evidence supports thesis |
| Trust score | Confidence rating (0–1) | 0.82 based on source reliability |
Agents use the graph to build briefs, fill scorecards, and suggest next actions. When you run the community health scorecard, Evidence Graph pulls fresh stories without manual tagging.
Security and control
- Data stays in your Supabase environment; Athenic agents query via service-role MCP access.
- Fine-grained permissions: restrict sensitive nodes to exec-only agents.
- Audit logs show who accessed which evidence and when.
What beta teams are seeing
- Faster decisions. A pre-seed fintech mapped customer interviews, Slack feedback, and onboarding analytics into the graph. Product brief creation dropped from two hours to 25 minutes.
- Better retention plays. Success teams run the retention experiment backlog with auto-suggested plays sourced from historical evidence.
- Clearer approvals. Approvers see trust scores and source clips before signing off on campaigns.
[EDITORIAL: Insert beta user quote or testimonial]
Who: An Athenic Evidence Graph beta user (founder, product lead, or ops lead)
Topic: How Evidence Graph improved their executive reporting, decision-making speed, or evidence retrieval
How to source:
- Reach out to current Evidence Graph beta users for a testimonial
- Interview them briefly about specific outcomes (time saved, decisions accelerated, etc.)
- Get permission to use their quote with name/title/company or anonymize as "Founder, Pre-seed Fintech"
Formatting: Use blockquote format with attribution:
> "Quote text here." - Name, Title, Company (or anonymized descriptor)
How to join the beta
Evidence Graph is in private beta. We’re onboarding teams who:
- Run weekly research or community rituals.
- Need to package proof for investors, customers, or leadership.
- Already connect data sources to Athenic via MCP.
Next steps
- Join the waitlist in-app under
App → Experiments → Evidence Graph. - Sync your primary data sources (Notion, HubSpot, Slack, Supabase) if you haven’t yet.
- Meet with our product specialists to configure your initial graph schema.
- Run a two-week pilot -our agents will monitor usage and surface wins.
Internal links
- /blog/product-evidence-vault-customer-insights
- /blog/customer-advisory-board-startup
- /blog/product-launch-readiness-checklist
- /use-cases/knowledge
External references
- Supabase Documentation: pgvector 0.5 (2024) – powering our embedding store.
- OpenAI Agents JS MCP Guide – how we wire integrations securely.
- Gartner 2024 Market Guide for Insight Engines – why unified knowledge graphs matter.
Crosslinks
- See also /blog/openai-rockset-acquisition-startup-impact
- /blog/customer-retention-experiment-backlog
Compliance & QA: Product details verified 21 Feb 2025 with Engineering. Security review logged. Links checked; no broken links. Style review complete.
- Max Beech, Head of Content | Expert reviewer: [EDITORIAL: Insert name of product team member or engineering lead who reviewed]