Canonical synthesis. Detailed per-platform dossiers live alongside ascompetitor-dossier-*(cx-ai, smb-widgets, voice-ai, avatar, backend-baas, auth-identity). Source doc:openalice/docs/competitive-matrix-2026-06-21.md.
OpenAlice — Competitive Feature Matrix + Self-Assessment (0–5)
Date: 2026-06-21 · By: Tycho (5 parallel research waves + synthesis) · Method: 39 platforms deep-researched across standardized 16-category frameworks (sources cited in the per-wave reports). Honest + critical — scores are where we actually are, not aspirational.
Scoring key (our implementation quality): 5 = best-in-class / shipped + polished · 4 = solid, shipped · 3 = real but partial · 2 = minimal / rough · 1 = stub / barely · 0 = absent. Gaps are flagged with the concrete fix.
0. Executive read — where we stand
The product whitespace is real and it's ours. Across 20 AI-agent competitors:
- Avatar leaders (Tavus, HeyGen/LiveAvatar, D-ID) have best-in-class embodiment but no platform layer — no analytics, no HITL inbox, no A/B, no proactive, no lead capture ("build it yourself via webhooks"), and none are omnichannel or self-hostable.
- Platform leaders (Intercom Fin, Sierra, Decagon, Ada, Zendesk) have deep platforms but zero avatar/embodiment and are US cloud-only, enterprise-priced ($30K–$1.5M/yr), no self-host.
- Nobody combines (embodied avatar) × (full conversational platform) × (EU-sovereign self-host). That triple is OpenAlice's moat.
The backend/auth moat is EU-sovereign self-host. WorkOS is US-only; Clerk/Auth0/Stytch/Firebase are cloud-only (US-HQ → CLOUD Act). Only Supabase/Keycloak/Ory/Zitadel/Appwrite/Hasura/Encore self-host. We already self-host a Rust stack in EU — and (Atlas-confirmed) openalice-tenants already ships SAML 2.0 SSO + SCIM 2.0, which most SMB competitors lack.
The honest gaps: omnichannel (web-only), multilingual breadth (3 langs vs 40–90), FGA/ReBAC authz, integration marketplace, compliance certifications (we have the posture, not the SOC2/ISO paper), and conversational-model differentiation (we orchestrate frontier LLMs; the incumbents post-train their own).
Part 1 — PRODUCT matrix (vs the 20 AI-agent competitors)
| # | Category | Market leaders / bar | OpenAlice today | Score | Top improvement |
|---|---|---|---|---|---|
| 1 | Conversational AI / LLM | Intercom Apex (own model, 73% resolution), Sierra Constellation (15+ models), Decagon (80% in-house) | Orchestrates frontier LLMs (gpt-5.x via providers) + tools/function-calling; honest guardrails | 3.5 | Add a resolution-rate metric + an eval/sim harness (like ElevenLabs/Ada); consider a CX-tuned fine-tune later |
| 2 | Knowledge / RAG | Ada/Decagon unified KB + gap auto-detection; Sierra Expert Answers | openalice-rag: doc/URL/file KB, pgvector; widget knowledge editor | 3.5 | Auto-sync + citations to visitors + KB-gap auto-detection from unanswered convos |
| 3 | Omnichannel | Intercom/Ada/Sendbird (web+WhatsApp+email+SMS+social, cross-channel continuity) | Web widget only | 2 | WhatsApp + email are the highest-ROI channel adds (per the embed roadmap) |
| 4 | Voice | ElevenLabs (135ms), Retell (~600ms), Vapi; phone agents | STT/TTS/clone (Mistral voice) + 3-layer voice stack (openalice-voice/-agent/-client) | 3.5 | Quantify latency; add phone/SIP inbound; barge-in tuning |
| 5 | Avatar / embodiment | Tavus (Phoenix-4 perception), HeyGen (photoreal), D-ID (V4) — all platform-less | 3D VRM embodied avatar (three 0.184 / three-vrm 3), clothing-logo, voice-synced — and on a full platform | 4.5 ★ | Our differentiator. Push lip-sync/perception quality; keep the platform+avatar combo nobody else has |
| 6 | Persona / brand customization | Sierra Ghostwriter, Ada Coaching | Persona script + 5 dials + appearance/brand/VRM + custom CSS + modes | 4 | Multi-persona per account polish; plain-English persona builder |
| 7 | Lead capture / CRM / commerce | Intercom (qualify+CRM), Tidio (Shopify actions) | Lead capture + CSV export (GDPR Art.20, gated) | 3 | Native CRM sync (HubSpot/Salesforce) + Shopify actions |
| 8 | Analytics / insights | Ada AR rate + Reasoning Logs; Sierra Insights 2.0 | Real analytics/insights/intent/revenue/sentiment/CSAT (no mock); removed the dishonest funnel | 4 | Add per-conversation eval scoring; honest deflection/resolution metric |
| 9 | Human handoff / HITL / inbox | Sierra Live Assist, Zendesk Agent Workspace | Live operator takeover/release/operator-message + 401 recovery + auto-refresh (just shipped) | 4 | Multi-operator routing; mobile operator app |
| 10 | A/B testing & optimization | Drift/ElevenLabs (live-traffic split); most competitors lack it | Honest A/B (no fabricated significance) — rare; most SMB rivals have none | 3.5 | Live-traffic split + a real significance gate (clearly labelled) |
| 11 | Integrations / API / webhooks | Zendesk 1,900+ marketplace; Intercom 300+ | Webhooks (now SSRF-guarded + send-test) + internal API | 2.5 | A public API + a few first-class integrations; marketplace is later |
| 12 | Security / compliance | Auth0/Zendesk (SOC2, ISO27001, FedRAMP); Sierra PCI L1 | EU-sovereign self-host, GDPR, PII-redaction, Postgres RLS, ownership-gated APIs (IDOR just closed), no-train | 3.5 | Pursue SOC2/ISO certification (we have the posture, not the paper) — a real enterprise unlock |
| 13 | Proactive / triggers | Sendbird (behavior/system triggers), Zendesk Outbound | Proactive triggers (time/scroll/exit/URL) + honest frequency cap | 3.5 | Outbound campaigns (email/WhatsApp) once channels land |
| 14 | Multilingual | Ada 63, Chatbase/Intercom 45–95 langs, auto-detect | i18n EN/DE/FR (3) + widget multilingual tables | 3 | Auto-detect + expand language coverage (biggest single quick credibility win) |
| 15 | Deployment / install | All ship a JS snippet; some no-code builders | Per-platform copy snippet (just fixed), SDK URL unified, test-drive, preview | 3.5 | Framework SDKs (React/Vue) + a no-code first-run (stepper just shipped) |
| 16 | Pricing model | Outcome-based (Sierra/Intercom $0.99/res), seat+usage | Prepaid balance / Mollie (EU), PAYG | 3 | An outcome/resolution metric to enable value-based pricing later |
Product takeaway: strongest where it's hardest to copy — embodied avatar (4.5) on a real platform (analytics 4, HITL 4, persona 4) with EU-sovereign self-host. Catch-up priorities: omnichannel (2), multilingual breadth (3), CRM/integrations (2.5–3), compliance certs.
Part 2 — BACKEND / PLATFORM matrix (vs Laravel/Next/NestJS/Supabase/Firebase/Appwrite/Convex/Hasura/Encore)
Our backend = a self-hosted Rust microservice stack (openalice-tenants, -auth, -agent-lite, -gateway, -rt/-realtime, -voice/-stt/-tts, -rag, -passport, -ml-gateway) on EU infra (Hetzner/blal.de), per-service Postgres/Redis (org convention).
| Category | Bar / leaders | OpenAlice backend | Score | Improvement |
|---|---|---|---|---|
| Database | Supabase/Hasura (RLS-native), per-service DBs | Postgres + RLS + sqlx migrations (tenants 54), per-service DBs (tenants / visitor-db) | 4 | Backups + PITR + branching for staging (see Atlas runbook) |
| API layer | Hasura auto-GraphQL, Supabase auto-REST+types | Hand-written axum REST, typed; internal-secret + Bearer/tid | 3.5 | Consider typed client codegen; per-route rate-limit |
| Real-time | Supabase Realtime, Laravel Reverb | openalice-rt / -realtime / -rt-client (3-layer WS stack) | 3.5 | Presence + scale story (Redis pub/sub) documented |
| Storage | Supabase/Appwrite (S3 + transforms + CDN) | VRM/asset serving, file uploads (KB/VRM) | 3 | S3-compatible + on-the-fly image transforms + CDN |
| Serverless / jobs | Encore (infra-from-code), Supabase Queues | Rust services + cron (Codex/digest crons) | 3.5 | A durable queue primitive for webhooks/retries |
| Vector / AI | Supabase pgvector, Convex hybrid search | openalice-rag + pgvector; ml-gateway | 3.5 | Hybrid (vector+keyword) + reranker |
| Observability | Encore (auto-tracing), Grafana stacks | Prometheus + Grafana (grafana.blal.pro), structured logs | 3.5 | Distributed tracing (OTel) across services |
| Multi-tenancy | Zitadel/Supabase (best), RLS | Tenant model + Postgres RLS + FORCE; ownership-gated reads (IDOR closed) | 4 | Keep extending the pool-isolation pattern (cf. Atlas ext-db) |
| EU-sovereign self-host | Only Supabase/Keycloak/Ory/Zitadel/Appwrite/Hasura/Encore qualify | Native — Rust, self-hosted, EU (Hetzner), own keys, no SaaS dependency | 5 ★ | This is the moat. Keep it pure (no US-cloud auth dependency) |
| DX | Encore/Supabase (local stack, codegen) | cargo + docker-compose per env; throwaway gates | 3.5 | A one-command local stack + type codegen would lift velocity |
| Deployment safety | — | Hardened this session: --no-deps recreate, health-gate, pg_dump-before-deploy (Atlas runbook) | 3.5 | Staging + zero-downtime (blue-green) next |
Part 3 — AUTH deep-dive (vs WorkOS/Clerk/Auth0/Stytch/Supabase/Keycloak/Ory/FusionAuth/Logto/Zitadel)
NAO emphasized this. Our auth = openalice-auth + openalice-tenants + openalice-gateway (jwt-auth-enforcement) + openalice-passport ("sovereign identity custodian").
| Auth capability | Bar | OpenAlice | Score | Fix |
|---|---|---|---|---|
| Core authN (pw/magic/OTP/social) | All | JWT-based; provider/social via auth service | 3.5 | Confirm social-provider breadth; add passkeys/WebAuthn |
| Enterprise SSO (SAML/OIDC) | WorkOS 60+ IdP, Keycloak | SAML 2.0 SSO shipped (tenants) | 3.5 | Self-serve IdP config portal; OIDC breadth |
| Directory sync / SCIM | WorkOS 18+, Zitadel | SCIM 2.0 shipped (tenants) — ahead of Logto/Supabase | 3.5 | Provider coverage (Okta/Entra/Google) |
| MFA / 2FA | Auth0 (push/WebAuthn), adaptive | Custom MFA | 3 | Passkeys + adaptive/risk-based later |
| Orgs / multi-tenancy | Zitadel (instance/org/project), Stytch | Tenant + team (owner/admin/member RBAC) | 4 | Member identity (add email to /me/team/members — flagged) |
| RBAC / FGA/ReBAC | Ory Keto (Zanzibar), Auth0 FGA | RBAC (owner/admin/member); no FGA/ReBAC | 1 ⚠ | Add Ory Keto (Apache-2.0, self-hostable, Zanzibar) without changing our token issuance |
| Sessions / tokens / M2M | All | HS256 JWT (Bearer + tid claim), X-Internal-Secret M2M | 3.5 | Refresh-token rotation + revocation review; agent/M2M tokens |
| Compliance | Auth0 (ISO/FedRAMP/FIPS) | GDPR + EU-sovereign; no SOC2/ISO cert | 3 | Certification is the enterprise unlock |
| Self-host / EU sovereignty | Keycloak/Ory/Zitadel | Native (our moat) | 5 ★ | — |
Auth verdict: we're stronger than the SMB pack (SAML SSO + SCIM shipped, which Logto/Supabase lack) and sovereign (which WorkOS/Clerk/Auth0/Stytch are not). The one strategic gap is FGA/ReBAC → adopt Ory Keto as a self-hostable side-car (keep our HS256 issuance, borrow only fine-grained authz). Passkeys + cert-track are the next two.
Part 4 — Prioritized improvement roadmap (where to invest)
Moat-deepening (keep the lead nobody else has):
- Avatar quality (lip-sync/perception) + keep the avatar↔platform integration tight.
- Stay EU-sovereign-pure (never take a US-cloud auth dependency).
Highest-ROI catch-up (credibility gaps):
- Omnichannel: WhatsApp + email (web-only is the #1 product gap; on the embed roadmap).
- Multilingual breadth + auto-detect (3→many langs — cheap credibility win; the i18n substrate is now in place).
- FGA/ReBAC via Ory Keto (self-hostable, fills the one real authz gap).
- CRM sync (HubSpot/Salesforce) + Shopify actions (commerce/lead value).
Enterprise unlock (slower, high value):
- SOC2 / ISO 27001 certification (we have the posture; the paper unlocks enterprise deals).
- Public API + a few first-class integrations (marketplace later).
- Deployment: staging + zero-downtime (Atlas runbook is step 1; backups + health-gate now in place).
Honest scorecard (means): Product ≈ 3.4/5 (with a 4.5 on the differentiator), Backend ≈ 3.7/5, Auth ≈ 3.3/5 — with two 5/5 structural moats (embodied-avatar-on-a-platform, EU-sovereign self-host) that the entire competitive field lacks. The path to a 4+ overall is omnichannel + multilingual + FGA + certs, not a rebuild.
Source data: the five research waves (CS-AI incumbents, SMB widgets, voice, avatar/builders, backend frameworks/BaaS, auth platforms) — 39 platforms, official docs/pricing/security pages cited inline in each wave's report.