Vision session with Nadav Greenberg (2026-04-15)
Summary
- Strategy crystallization with nadav-greenberg (advisor). Sharpened SDR-first wedge with brain layer + agent orchestration as the actual core product.
- Voice clarified as a feature, not the core — most replaceable roles happen to be phone-based, so voice is the natural starting capability, but the brain is the real product.
- Brain layer must work for human workers too, not only AI agents — validated by plan to test with monday.com’s own human SDRs.
- API-first architecture decided as a hard constraint — every capability accessible via API, enabling agents in Claude Code / OpenAI / etc.
- Second vector resolved: lightweight marketing data connectors (NOT a full marketing agent) — driven by Nadav’s capacity challenge (14 engineers, 5-month window).
- Competitive deep-dives: Sierra (per-resolution pricing, CX-focused), OneAI (per-lead pricing), Leaping AI (“self-learning” is actually manual prompt updates), Wonderfull (services-heavy, non-English play).
Decisions raised
- DEC-015 — Voice is a feature, not the core. Brain + agent orchestration is the platform.
- DEC-016 — Second vector = lightweight marketing data connectors for source attribution (NOT a standalone marketing agent).
- DEC-017 — Support both pre-built opinionated agents AND custom / bring-your-own agents.
- DEC-018 — Product is API-first: every capability accessible via API.
- DEC-019 — Brain layer works for human workers too; validate cross-funnel value on humans before requiring agents.
Action items raised
- AI-016 — Vitali/Ayala — Productize SDR voice agent (UX, onboarding, playbook builder, CRM polish, dashboards).
- AI-017 — Tomer/Niv — Build lightweight marketing data connectors (Google Ads, Facebook Ads source attribution).
- AI-018 — Saar/Vitali — Design and implement API-first architecture.
- AI-019 — Saar/Tomer — Validate Harmony brain on human workflows: test cross-funnel insights with monday’s human SDRs and marketing campaigns first.
Open questions raised
- OQ-011 — What is the minimum viable second agent for cross-funnel value? Connector enough, or full marketing agent?
- OQ-012 — Can 14 engineers become a legitimate SDR player in 5 months against $100M+ funded competitors?
- OQ-013 — Pursue non-English languages as differentiation (Spanish, Arabic, Hebrew)?
- OQ-014 — How to price the brain / orchestration layer standalone (for bring-your-own-agent customers)?
Entities
- saar-arbel, vitali, nizan-shifman, tomer-friedman, niv, ben, ayala — Harmony core team
- nadav-greenberg — strategy advisor running this session
- sierra — per-resolution pricing competitor (CX-focused)
- oneai — per-lead pricing competitor
- leaping-ai — competitor whose “self-learning” is manual prompt editing
- wonderfull — services-heavy competitor with non-English strategy (Gordon Ramsay backed)
- elevenlabs — voice infra competitor; vulnerable to open-source disruption
- anthropic — possible distribution partner via Claude Code plugin
Concepts
- brain-layer — knowledge/memory layer + agent orchestration; THE core product
- voice-as-feature — voice is a premium capability, not the core positioning
- api-first-architecture — hard constraint, enables external-platform distribution
- bring-your-own-agent — platform supports custom and external agents
- global-maximum — long-term framing reaffirmed
- self-learning-loop — true differentiator vs Leaping AI’s manual approach
- wedge-strategy — SDR-first wedge; brain as platform expansion
- pricing-models — per-lead vs per-resolution vs hybrid debated
- marketing-connectors — second vector (data-only, not a full agent)
Notable quotes
“Voice doesn’t appear as a core function in the vision document.” — Nadav, prompting DEC-015 “Most companies are not all-agent yet — proving Harmony on human workflows de-risks adoption.” — Nadav, prompting DEC-019