Decisions Log
Last updated: 2026-04-23
| ID | Decision | Date | Context | Owner | Source | Related Domain |
|---|---|---|---|---|---|---|
| DEC-021 | Amplemarket will not implement active voice agent calling for outbound prospecting due to legal restrictions (e.g., US FTC regulations); voice agents limited to inbound support scenarios only. | 2026-04-23 | Regulatory/compliance constraint | elad-flishko | amplemarket_overview_2026-04-23_08-30.json | sales-automation |
| DEC-020 | AI assistant should be positioned as pocket tool (private to AM) rather than on-call voice for customer meetings to avoid perception of deprioritizing customer time | 2026-04-23 | Jonathan Buchen cautions that enterprise customers perceive AI pre-calls as wasteful of their time; advocates for private in-call chat instead (Win.ai model) | harmony-core | ai_for_am_2026-04-23 | go-to-market |
| DEC-008 | Operational separation for Harmony/Charm: independent domain, Slack, G-Suite, AWS, GitHub, Okta, but single Monday legal entity. Teams get fast-track IT/Security approval path for pre-approved vendors. | 2026-04-20 | Balancing startup autonomy with Monday’s compliance obligations (SOC 2, SEC/SOX, GDPR, ISO). Will hire external SOC 2 advisor from day 1. May 1, 2026 launch; early-summer alpha to users. | Vitali / Nizan | new_startup_workspaces_2026-04-20_09-15.json | domains/infrastructure.md |
| DEC-007 | Harmony and Charm will seek SOC 2 certification by September 2026; will receive separate SOC 2 report (not bundled into Monday’s audit). Monday retains monitoring/control; teams operate semi-autonomously within compliance boundaries. | 2026-04-20 | Both products are “powered by Monday” from legal perspective, so they inherit Monday’s regulatory obligations for data breach reporting, fines, and SEC/SOX compliance. External advisor to ensure gaps don’t emerge post-launch. | Vitali / Nizan / Gil Lan | new_startup_workspaces_2026-04-20_09-15.json | domains/infrastructure.md |
| DEC-006 | Monday legal entity remains; brand/operations separated. Teams manage own bank accounts, budgets (SaaS, vendors), and can add third-party tools via a “fast-track” process (pre-approved list + security review for new vendors). | 2026-04-20 | Creates autonomy (can spend budget, move fast) without creating separate legal entity (avoids investor/stakeholder comms complexity and enables Monday to retain compliance ownership). | Vitali / Nizan | new_startup_workspaces_2026-04-20_09-15.json | domains/infrastructure.md |
| DEC-019 | Harmony “brain” (cross-department intelligence layer) must work with human workers too, not only AI agents. Initial proof of cross-funnel value should be tested on human SDRs and human-created campaigns before requiring agents everywhere. | 2026-04-15 | Nadav raised that most companies are not all-agent yet. Proving Harmony on human workflows de-risks adoption and broadens the addressable market. | Nadav / Saar | vision_2026-04-15_10-00.json | domains/product.md |
| DEC-018 | Product must be API-first: every capability accessible via UI should also be available via API. This enables agents to live outside Harmony (in Claude Code, OpenAI workflows, etc.) and opens distribution through AI platforms. | 2026-04-15 | Discussion about making Harmony the default sales plugin for Claude/OpenAI ecosystems. API-first enables partnership plays (e.g., Anthropic collaboration) and removes dependency on own UI. | Saar / Vitali | vision_2026-04-15_10-00.json | domains/infrastructure.md |
| DEC-017 | Product should support both pre-built opinionated agents AND custom agent creation (“bring your own” or “create custom”). Some customers need agents for non-sales functions that Harmony doesn’t natively cover. | 2026-04-15 | Customer conversations revealed demand for agents beyond SDR/support (e.g., legal, finance). Framework of role-specific agents enables adding new agent types easily. | Saar / Vitali | vision_2026-04-15_10-00.json | domains/product.md |
| DEC-016 | Near-term investment in the “second vector” (beyond SDR) should focus on marketing data connectors that feed cross-funnel insights into the SDR agent — not a standalone marketing agent product. This proves Harmony’s value without overextending the team. | 2026-04-15 | Extensive debate about focus vs breadth. Consensus: build lightweight marketing connectors (data-only, not actioning) to enrich SDR context and demonstrate cross-funnel learning. The marketing “agent” is initially just a connector, not a product. | Saar / Nizan / Nadav | vision_2026-04-15_10-00.json | domains/product.md |
| DEC-015 | Voice is a premium tool / feature of agents, not the core product. The core is the platform that powers agents with shared context, learning, and orchestration. Voice happens to be the first capability because most replaceable roles are phone-based. | 2026-04-15 | Nadav pointed out voice doesn’t appear as a core function in the vision document. Team agreed: voice is best-in-class capability that enables the first wedge, but the platform (brain + agent orchestration) is the actual product. | Nadav / Saar | vision_2026-04-15_10-00.json | domains/product.md |
| DEC-014 | First ICP for the SDR wedge is mid-market B2B teams with inbound + outbound motion, meaningful call volume, existing Salesforce/HubSpot, and ideally Gong. Avoid SMB self-serve and enterprise-heavy first deployments. | 2026-04-14 | Approved strategy narrows the wedge to customers where touch onboarding is viable and downstream learning signals exist. | Saar / Nizan | competitor_strategy_9e25dcc6.plan.md | domains/product.md |
| DEC-013 | Core market message is “self-learning revenue voice agent” rather than generic AI builder / voice platform. Differentiation rests on downstream CRM + Gong feedback loops and cross-funnel memory. | 2026-04-14 | Competitor research showed the crowded market is agent creation and surface-level automation; Harmony should frame around measurable learning from what closes. | Saar / Nizan | competitor_strategy_9e25dcc6.plan.md | domains/product.md |
| DEC-012 | Product packaging will be staged as Execution layer first, Learning layer second, Revenue memory layer third. Global Maximum remains the long-term narrative, not equal-weight near-term scope. | 2026-04-14 | This sequencing keeps the July wedge concrete while preserving the broader revenue-platform expansion path. | Saar / Nizan | competitor_strategy_9e25dcc6.plan.md | domains/product.md |
| DEC-008 | Product will be positioned as a “revenue platform” (like Gong), not claiming to replace CRM. Can stand alone for SMBs but integrates on top of existing CRM (Salesforce/HubSpot) for mid-market. | 2026-04-14 | Offsite consensus: CRM replacement is too risky and distracting; better to integrate via connectors and let native agents grow over time. | Nizan / Saar | offsite_2026-04-14_07-00.json | domains/product.md |
| DEC-007 | Go-to-market is Touch (not No-Touch). Sales reps + solution engineers will onboard customers. Not heavy implementation (months), but meaningful onboarding support (hours/days). | 2026-04-14 | Team agreed enterprise/mid-market sales requires touch; call center setup needs configuration; SDR agent needs tuning per customer. | Nizan / Saar | offsite_2026-04-14_07-00.json | domains/product.md |
| DEC-006 | SDR/call-center voice agent is the first product to ship, targeting July 1 production with paying customers. | 2026-04-14 | SDR is closest to ready; team has voice tech that is faster (70ms vs 800ms response) and cheaper than competitors using 11Labs wrappers. | Vitali / Saar | offsite_2026-04-14_07-00.json | domains/product.md |
| DEC-005 | SDR agent must be “self-learning” — improves automatically based on closed-loop feedback from CRM outcomes (closed-won/closed-lost) and Gong call recordings. This is the key differentiator. | 2026-04-14 | Vitali proposed; team agreed this distinguishes from competitors who pass optimization burden to customer via prompt editing. | Vitali | offsite_2026-04-14_07-00.json | domains/infrastructure.md |
| DEC-004 | Global Maximum vision is the long-term north star. SDR is the entry point, with expansion to marketing agent and cross-department intelligence over time. Two parallel prototype tracks created at offsite. | 2026-04-14 | Team debated whether to start with Global Maximum or SDR. Consensus: SDR first as wedge product, Global Maximum as vision/messaging, with cross-department insights growing organically. | Saar / Nizan | offsite_2026-04-14_07-00.json | domains/product.md |
| DEC-003 | Campaign analysis reports should be generated monthly (not daily). Need to cover all campaigns (not just CRM Main) and include cost + revenue (ARR d14, ARR d28) data. | 2026-04-13 | Tamar confirmed monthly cadence is minimum meaningful interval due to conversion cycle length. Individual keyword-level analysis is too granular for regular reports. | Tamar / Nizan | harmony_analyzing_campaigns_2026-04-13_08-00.json | domains/product.md |
| DEC-002 | Prefer direct API access to Salesforce and Gong over pre-aggregated Snowflake/BigBrain data, because future external customers won’t have those internal tools. | 2026-04-12 | Building for generalizability: need to prove the system works with raw API data, not monday.com-specific aggregations. | Saar | harmony_bizapps_sync_2026-04-12_13-00.json | domains/infrastructure.md |
| DEC-001 | Guy Kori (BizApps) will provide Salesforce and Gong API access. He will add a team member to help provision credentials. Contact Almog Ross at BigBrain for Account Timeline / Snowflake data as supplementary source. | 2026-04-12 | Guy showed Account Timeline tool in HubSpot which aggregates activity feed, emails, Gong calls. Useful as reference but raw API preferred. | Guy / Saar | harmony_bizapps_sync_2026-04-12_13-00.json | domains/infrastructure.md |