Roadmap

Where we are.
Where we're going.

Arbiter is in active development. This page shows what's been built, what's next, and the thinking behind the product.

What Arbiter does today

Everything in Phase 0 and Phase 1 is built and working.

Real-time transcriptionInline AI analysis (fact-checking, logic flow)Interactive conversation graphLive meeting status panelContext gap detectionMeeting health indicatorsMeeting purpose & effectiveness trackingComposable widget system with meeting type presetsParticipation meter

Phase 1

Complete

5

AI analysis tools

443

Tests passing

9+

Research areas explored

Development phases

0

Foundation

Complete

Solid, deployable codebase

100%
  • Real-time transcription + 2 analysis tools
  • Backend test suite
  • Dockerize backend + frontend
  • Error handling & reconnection
  • Debounced continuous analysis
1

Conversation Graph

In progress

Making conversation structure visible in real time

100%
  • Anchor analysis to transcript segments
  • Inline transcript annotations
  • Conversation graph visualization
  • Conversation ontology shared schema
  • Enhanced LogicFlowAnalyzer for graph edges
  • Resizable pane layout
  • ConversationStateManager backend
  • Live status panel UI
  • Graph visual polish — state & edge encoding
  • Meeting health indicators
  • Context gap detection
  • Meeting purpose & effectiveness tracking
  • Composable widget system & meeting presets
  • Participation meter widget
2

Pilot-Ready

Shareable, usable in real meetings

0%
  • Persistent storage (database)
  • Meeting bot integration (Recall.ai)Join Zoom, Meet, and Teams with a link
  • Authentication (invite codes)
  • Cloud deployment
  • Speaker diarization
  • Integration framework + webhook API
  • Google Calendar integration
  • Slack push integration
  • Meeting summarizer tool
  • Export (PDF, markdown, JSON)
3

Product-Market Fit

Learn from pilots, build integrations

0%
  • Entity recognition frameworkDetect Jira tickets, CRM deals, GitHub PRs in speech
  • Linear / Jira integration
  • HubSpot CRM integration
  • GitHub PR/issue enrichment
  • Notion structured export
  • Decision → Action pipeline
  • Zapier connector app
  • Full auth + multi-user
  • Organizational knowledge bases
  • Business model + pricing
4

Platform & Scale

Extensible meeting intelligence platform

0%
  • Salesforce deep CRM integration
  • Microsoft ecosystem (Teams, SharePoint)
  • Analysis tool marketplace / SDK
  • Multi-language support
  • Browser extension for system audio capture
  • Mobile companion app

Key decisions

Strategic choices we've made and why.

Target user?General-purpose

Too early to narrow. The extensible tool system serves many personas — let real usage guide focus.

Meeting bot approach?Third-party API (Recall.ai)

Fastest path to pilot. Can migrate to self-hosted later to reduce per-meeting costs.

Conversation state?Backend-driven (ADR-003)

Multi-client consistency and foundation for an organizational knowledge graph.

Integration architecture?Abstraction layer + webhook API + Zapier

Build the framework first. Individual providers are small once the abstraction exists.

Timeline?Pilot in weeks, not months

Validate the concept with real users before over-building infrastructure.

Research & thinking

Areas we've investigated to inform the product direction.

ResearchOntology

Conversation graph prior art

IBIS argumentation mapping, Graphiti temporal knowledge graphs, and how they inform our ontology design.

MarketStrategy

Competitive landscape

Analysis of Otter, Fireflies, Fathom, Gong, Read.ai, Granola, and the gaps in the market.

ArchitectureIntegrations

Integration strategy

Prioritized integration roadmap — EntityRecognizer framework, webhook API, and why HubSpot before Salesforce.

ResearchInfrastructure

Recall.ai integration spec

Full API spec for meeting bot infrastructure — joining Zoom, Meet, and Teams programmatically.

LegalEthics

Privacy & consent framework

How to handle AI analysis of live conversations responsibly — consent models, data handling, regulatory landscape.

ArchitectureDecision

Backend-driven conversation state (ADR-003)

Architecture decision: backend is the authority on conversation state. Frontend is a thin render layer.

ProductResearch

Meeting purpose & effectiveness

Four-layer design for optional success criteria, real-time tracking, purpose drift detection, and post-meeting scorecards.

InfrastructureDecision

Deployment & hosting architecture

Evaluating Fly.io, Railway, and AWS for the pilot deployment. Cost modeling and container orchestration.

ProductUX Research

Real-time meeting UX surfaces

Competitive analysis of 7 meeting tools, HUD patterns for real-time intelligence, composable widget architecture for different meeting types.

Want to follow along?

Request early access and we'll keep you in the loop as we ship.

Get early access