Sales Note-Taker & Conversation-Intelligence Strategy

Help the user (a) pick the right AI meeting note-taker for their sales team, and/or (b) wire its API into downstream systems (CRM, warehouse, Slack, internal tools). This skill is tool-agnostic — it references specific platforms for selection and integration patterns but focuses on cross-vendor strategy.

Step 1 — Gather context

If references/learnings.md exists, read it first for accumulated knowledge.

Ask the user:

1. What's your primary goal?

   • A) Pick a note-taker for my team (selection/comparison)
   • B) Integrate an existing note-taker's API (transcripts → somewhere)
   • C) Both — evaluating a platform AND planning the integration
   • D) Debugging a specific API/webhook issue

2. What's your use case?

   • A) Sales org — discovery/demo calls, pipeline intelligence, coaching
   • B) Customer success — onboarding, QBRs, renewal health
   • C) Cross-functional — sales + CS + internal meetings
   • D) Specific workflow — e.g., auto-log transcripts to HubSpot, push action items to Slack, pipe everything to Snowflake

3. If integrating, where are transcripts going?

   • A) CRM (HubSpot, Salesforce, Attio, Pipedrive)
   • B) Data warehouse (Snowflake, BigQuery, Redshift, Databricks)
   • C) Internal app / database
   • D) Slack / notifications only
   • E) Multiple destinations (fan-out)

4. Scale?

   • A) <50 meetings/week (single team)
   • B) 50-500/week (mid-market sales org)
   • C) 500+/week (enterprise, multi-team)

Skip-ahead rule: if the user's prompt already has enough context, skip to Step 2.

Step 2 — Route or answer directly

Problem domain	Route to
Reviewing a specific call for coaching or scoring	/sales-call-review {user's question}
Building a coaching program or training cadence	/sales-coaching {user's question}
General CRM-to-tool integration patterns (Zapier, webhooks, iPaaS)	/sales-integration {user's question}
Salesloft Conversations config	/sales-salesloft {user's question}
ZoomInfo Chorus config	/sales-zoominfo {user's question}
Seismic Meeting Intelligence config	/sales-seismic {user's question}
Meeting scheduling (not note-taking)	/sales-meeting-scheduler {user's question}

Otherwise, answer directly below.

Step 3 — Platform reference

Start with references/platforms.md for the quick comparison table across all platforms.

Then read the detailed file matching the user's segment:

Segment	File	When to read
SMB / budget	references/platforms-smb.md	Budget < $50/user/mo, small teams, free-tier shoppers
Mid-market / sales	references/platforms-midmarket.md	Sales-focused teams, methodology scorecards, coaching, CRM enrichment
Enterprise / revenue	references/platforms-enterprise.md	$1K+/user/yr budget, revenue intelligence, bolt-on CI
Field / in-person	references/platforms-field.md	In-person sales, field reps, mobile recording

For comparison questions spanning segments (e.g., "Fathom vs Gong"), read both relevant files.

Answer using only the relevant section(s) — don't dump the full reference.

Step 4 — Actionable guidance

You no longer need the platform reference — focus on the user's specific situation.

Selection framework (if goal = selection)

Rank these in order for the user's context:

1. Budget per seat — Tactiq (free tier + $8-$12/mo) < Fathom (free tier + $16-$20/mo) < MeetGeek (free tier + $9.99-$17) < Fireflies ($10-$19) < Otter ($17-$30) < tl;dv (free tier + $18-$98/mo) < Avoma ($19-$79) < Claap ($24-$48/user/mo) < Gong ($1,200-$1,600/user/yr). Free tier only exists on Fathom, Fireflies, Otter, Read.ai, MeetGeek, tl;dv, Claap (very limited), Tactiq (very limited — 10 transcripts/mo).
2. CRM depth — Gong and Avoma have the deepest native CRM field-mapping and deal intelligence; Fathom Business, Fireflies Business, Fellow Business ($15/seat, HubSpot/Salesforce) are solid; Otter is thinnest.
3. Coaching/QA features — Gong > Avoma > tl;dv Business (MEDDIC/BANT/SPIN playbooks, aggregated insights) > Salesloft Conversations / Chorus > Fathom/Fireflies (basic scorecards only).
4. API completeness — Gong (richest, webhooks + REST) > Fireflies (GraphQL + webhooks) > Avoma (REST + webhooks) > Fathom (REST + webhooks) > Wave (REST + webhooks + MCP + semantic search, 60/min + 1K/day) > MeetGeek (REST + webhooks + MCP, all plans) > Jamie (REST + webhooks + MCP, Pro+) > Grain (REST, Business+) > Sembly (webhooks-first) > Read.ai (REST beta + MCP + webhooks) > tl;dv (REST v1alpha1 + webhooks + MCP, Pro+) > Granola (REST read-only + MCP beta, Business+, 5 req/sec, no webhooks) > Jamy (REST partial, 2 endpoints + webhooks). Otter is Enterprise-only API (beta).
5. Compliance — Gong, Avoma, Fireflies Business, Fathom Business all have SOC 2 + GDPR. For HIPAA/regulated industries, short-list drops to Gong, Avoma, Fireflies Enterprise.

Quick picks:

• Solo rep / tiny team on a budget → Fathom free tier
• SMB sales team, HubSpot/Salesforce-centric → Fathom Business or Fireflies Business
• Mid-market sales org wanting methodology + coaching → Avoma
• Enterprise revenue intelligence, high budget → Gong
• Enterprise CI + forecasting in one platform, Clari Core user → Clari Copilot ($60-110/user/mo)
• Non-sales general meeting transcription → Otter or Fellow
• Mid-market team wanting CRM auto-enrichment + deal scoring without Gong prices, especially in the lemlist ecosystem → Claap
• Mid-size team wanting customizable templates + analytics + affordable CRM sync → MeetGeek
• GTM team wanting shareable call clips + MCP AI workflows → Grain
• Real-time AI coaching during calls + cross-session intelligence + Apple Watch → Hedy
• Privacy-first EU team, in-person meetings, or bot-free non-negotiable → Jamie
• Noisy environments where call quality is the primary pain + AI notes as a bonus → Krisp
• Real-time live transcription during calls + Sales Copilot with cue cards and objection handling → Colibri
• Real-time AI coaching overlay with knowledge base RAG and pre-call briefs → Cluely
• Multilingual team needing real-time translation in 100+ languages + cross-language search → Jamy
• In-person meetings, phone calls, cross-device recording, semantic search across history → Wave
• Free unlimited recordings + sales coaching with playbook monitoring (MEDDIC/BANT/SPIN) → tl;dv
• Zero-setup Chrome extension, bot-free, Google Meet-first, no API needed → Tactiq
• Bot-free AI notepad for VCs/product teams, privacy-sensitive contexts, HubSpot CRM → Granola
• Open-source wearable for all-day in-person capture, developer extensibility, lowest price ($89) → Omi
• Bot-free video recording with screen capture, Chrome-first teams, CRM auto-sync on Business plan → Bluedot
• Multilingual global team needing 58-language transcription + cross-meeting AI intelligence (Notta Brain) → Notta
• All-in-one AI assistant that handles email + meetings + calendar + custom workflows in a single tool → Lindy
• In-person field sales coaching with virtual ridealongs, home services industry → Rilla
• Home services field + call center coaching with active AI feedback, ServiceTitan/Jobber/Housecall Pro → Sales Ask
• macOS sales team wanting auto-CRM update from calls without manual logging, domain templates (legal, medical, recruiting) → Sonnet
• Agency wanting AI-generated deliverables (pitch decks, briefs, emails) from client meetings, bot-free → Supernormal
• Ultra-cheap personal voice notes with HIPAA compliance, no meeting bot needed → VoiceToNotes
• Mid-market team wanting AI roleplay + real-call scoring + CRM auto-fill without Gong pricing → Outdoo
• Mid-market team wanting coaching-focused conversation intelligence with wide CRM support (8 CRMs) without Gong pricing → Jiminny
• Small-to-mid contact center needing automated QA scoring on 100% of calls without Gong pricing → Enthu.AI
• Enterprise team wanting conversation intelligence tightly integrated with revenue forecasting (Clari Core) → Clari Copilot
• Real-time in-call coaching + QA automation + revenue intelligence in one platform, APAC/multilingual focus → Salesken
• Enterprise contact center (100+ agents) needing 100% auto QA with path to AI virtual agents → Observe.AI

Integration patterns (if goal = API integration)

Webhook-first vs polling — pick webhook whenever available.

Pattern	When to use	Trade-off
Webhook push (event-driven)	Default — when vendor offers transcription.completed or meeting.ended events	Must expose a public HTTPS endpoint; must handle retries, idempotency, signature verification
Polling (list + fetch)	Backup for vendors without webhooks (Otter API beta), or backfilling history	Wastes API calls, higher latency, hits rate limits faster
Hybrid	Webhook for real-time + periodic reconcile-poll to catch missed events	Best for production — covers webhook outages

Standard integration architecture (works across vendors):

1. Ingest — Webhook receiver (Lambda/Cloud Run/serverless) validates HMAC signature, enqueues raw payload
2. Fetch — Worker pulls full transcript from API using meeting_id / recording_id from payload
3. Normalize — Map to a common schema: {meeting_id, external_id, vendor, started_at, duration_s, participants[], transcript[{speaker, ts, text}], summary, action_items[], recording_url}
4. Persist — Raw JSON to S3/GCS (audit trail) + normalized rows to warehouse + lightweight record to CRM
5. Fan-out — From the normalized layer, push to Slack / Salesforce opportunity / HubSpot deal / Notion / wherever

Auth patterns:

• API key in header — Fireflies, Fathom, Sembly, Read.ai, MeetGeek (simplest; rotate per-user or per-workspace)
• OAuth 2.0 — Avoma, Gong (required for multi-tenant apps)
• Workspace-scoped keys — Otter (Enterprise only)

Rate limits to design around:

• Gong: 3 req/sec, 10k/day, ~1k/hr per key — batch transcript fetches, cache call metadata
• Fireflies: GraphQL complexity limits — don't request full transcript + audio URL in one query for 100 meetings
• Fathom: document per-key limits; paginate with cursor
• Avoma: standard REST pagination; webhooks preferred for volume
• Otter: Enterprise-only, undocumented beta limits — assume conservative

Transcript format normalization:

Every vendor returns a slightly different shape. Build a normalize_transcript(vendor, payload) function early. Common fields: speaker (email or display name), timestamp_ms (offset from start), text, confidence (optional). Fathom uses markdown export natively; Gong/Avoma return structured JSON with monologues; Fireflies returns sentences; Otter returns speaker-labeled paragraphs.

If you discover a gotcha, rate-limit ceiling, auth quirk, or vendor-specific payload oddity not covered in references/learnings.md, append it there with today's date.

Gotchas

• Webhooks without signature verification are dangerous. Every vendor offers HMAC or similar — Fireflies uses x-hub-signature (SHA-256), Gong signs with its own header, Avoma provides a secret. Always verify before trusting the payload.
• "Transcript ready" ≠ "summary ready." Vendors often fire the transcript webhook before AI summary/action-items are complete. Either wait for a separate summary.completed event (Fathom, Fireflies) or re-fetch a few minutes later.
• Rate limits are the silent killer at scale. Gong's 3 req/sec hits fast when you have 500 calls/day and need metadata + transcript + participants for each. Cache aggressively, use batch endpoints where available, and queue API calls instead of fanning out in parallel.
• Otter's public API is Enterprise-only and in beta. Do not plan a production integration on Otter's API unless you've confirmed Enterprise access. Zapier is the safer fallback for Otter.
• Don't skip reconcile-polling. Webhooks fail — network blips, deploy downtime, vendor-side bugs. Run a nightly job that lists meetings from the last 48h and fills gaps. This has saved every production integration we've seen.
• PII and storage compliance. Transcripts contain names, pricing, contract terms, sometimes PHI. If you're piping them to a warehouse, plan retention + access controls before the pipeline ships — not after Legal gets involved.
• "Cheapest per seat" is misleading. Fathom free looks great until you need Business-tier CRM sync; Fireflies Business limits monthly meetings per user. Price the plan you'd actually run, not the teaser.
• Platform-specific conversation intelligence (Salesloft, Chorus, Seismic) ≠ replacement for a dedicated note-taker. They're bolt-ons to existing tools. If you don't already use Salesloft/ZoomInfo/Seismic, don't buy them just for meeting recording — use a dedicated note-taker.

Related skills

• /sales-fathom — Fathom platform help (REST API, webhook HMAC, OAuth, SDKs, MCP, pricing tiers)
• /sales-fireflies — Fireflies.ai platform help (GraphQL API, Webhooks V1 + V2, HMAC, AskFred, Real-time, MCP, rate limits per tier)
• /sales-avoma — Avoma platform help (REST API, webhooks, MEDDIC/BANT scorecards, Lead Router, pricing tiers)
• /sales-gong — Gong platform help (Revenue AI OS — call recording, Smart Trackers, coaching scorecards, deal boards, REST API, webhook automation, Forecast, Engage, MCP)
• /sales-otter — Otter.ai platform help (OtterPilot, live transcription, AI Chat, Zapier automation, CRM sync, pricing tiers)
• /sales-fellow — Fellow platform help (REST API, bot-based/bot-free recording, CRM sync, templates, pricing tiers)
• /sales-grain — Grain platform help (REST API, MCP Server for Claude/Cursor, video clips, Zapier automation, CRM sync, pricing tiers)
• /sales-sembly — Sembly platform help (agentic meeting intelligence, AI-generated deliverables, webhook automations, 10 native CRM connectors, pricing tiers)
• /sales-read-ai — Read.ai platform help (engagement/sentiment analytics, Search Copilot, REST API + MCP Server, webhook automations, CRM sync, pricing tiers)
• /sales-meetgeek — MeetGeek platform help (bot + no-bot recording, customizable templates, REST API with regional endpoints, MCP Server, 7 CRM connectors, conversation analytics)
• /sales-jamie — Jamie platform help (bot-free AI note-taker, REST API with personal/workspace keys, webhook automations, CRM sync to HubSpot/Salesforce/Attio, MCP server, EU privacy-first)
• /sales-krisp — Krisp platform help (noise cancellation + AI meeting notes, accent conversion, Voice AI SDK, webhook integrations, Call Center AI, pricing tiers)
• /sales-colibri — Colibri.ai platform help (real-time transcription, AI summaries, Sales Copilot with live guidance, conversation analytics, legal transcription, pricing tiers)
• /sales-cluely — Cluely platform help (real-time AI coaching overlay, pre-call briefs, knowledge base RAG, CRM sync via Merge.dev, coaching analytics, pricing tiers)
• /sales-jamy — Jamy.ai platform help (AI meeting assistant with real-time translation in 100+ languages, cross-language search, HubSpot CRM sync, API/webhooks, pricing tiers)
• /sales-wave — Wave platform help (mobile-first AI note-taker, REST API with semantic search and webhooks, MCP server, Voice ID, 76-language transcription, pricing tiers)
• /sales-tldv — tl;dv platform help (AI note-taker with unlimited free recordings, REST API v1alpha1, sales coaching with MEDDIC/BANT/SPIN playbooks, CRM sync, MCP server, pricing tiers)
• /sales-claap — Claap platform help (AI sales call assistant with no-bot recording, REST API + OAuth, CRM auto-enrichment, deal scoring, MCP connector, pricing tiers)
• /sales-tactiq — Tactiq platform help (Chrome extension bot-free note-taker, Zapier automation, native HubSpot/Salesforce/Pipedrive CRM sync, AI credits, pricing tiers)
• /sales-hedy — Hedy AI platform help (real-time AI meeting coach, REST API, webhooks, MCP server, Zapier/n8n automation, pricing tiers)
• /sales-granola — Granola platform help (bot-free AI notepad, REST API, MCP beta, HubSpot/Attio/Affinity CRM sync, Zapier automation, pricing tiers)
• /sales-omi — Omi AI platform help (open-source AI wearable necklace, Developer API, app marketplace with webhook integrations, in-person + online capture, pricing tiers)
• /sales-bluedot — Bluedot platform help (bot-free AI note-taker with video recording, Chrome extension + desktop/mobile apps, Svix webhooks, HubSpot/Salesforce CRM sync, screen recording, pricing tiers)
• /sales-notta — Notta platform help (AI meeting transcription in 58 languages, Notta Bot, Notta Brain cross-meeting AI, Zapier automation, CRM sync to Salesforce/HubSpot/Zoho/Freshsales/Salesflare)
• /sales-lindy — Lindy platform help (no-code AI agent builder — meeting notes are one module alongside email triage, calendar management, custom workflows, chatbots, AI phone agents)
• /sales-rilla — Rilla platform help (AI coaching for in-person/field sales, virtual ridealongs, mobile recording, CRM sync via Merge, home services focus)
• /sales-salesask — Sales Ask platform help (AI coaching for in-home + call center sales, Coach Dean active feedback, ServiceTitan/Jobber/Housecall Pro, home services focus)
• /sales-sonnet — Sonnet AI platform help (bot-free recording, auto-CRM update to HubSpot/Salesforce, domain templates for sales/legal/medical/recruiting, relationship tracker, concept maps)
• /sales-supernormal — Supernormal platform help (AI agent for agencies, bot-free recording, AI-generated deliverables from meetings, HubSpot/Salesforce/Pipedrive CRM sync, credit-based pricing)
• /sales-voicetonotes — VoiceToNotes platform help (ultra-cheap personal voice-to-text, HIPAA compliance, AI summaries/grammar/OCR, web + iOS + Android, no integrations)
• /sales-outdoo — Outdoo platform help (AI roleplay coaching, conversation intelligence, revenue intelligence, CRM automation, Open API, mid-market Gong alternative)
• /sales-jiminny — Jiminny platform help (conversation intelligence, revenue intelligence, coaching, 8 native CRM integrations, partial REST API, mid-market Gong alternative)
• /sales-convin — Convin platform help (contact center AI with 100% automated QA, Real-Time Assist, AI Phone Call agent, LMS, 70+ languages, no public API)
• /sales-uniphore — Uniphore platform help (enterprise Business AI Cloud — conversation intelligence, U-Assist real-time guidance, automated QA, CSATai, Emotion AI, ActionIQ CDP)
• /sales-observe-ai — Observe.AI platform help (enterprise contact center intelligence, 100% Auto QA, Agent Copilot, VoiceAI/ChatAI agents, 250+ integrations)
• /sales-nice-cxone — NICE CXone platform help (full CCaaS with built-in recording, QM, Interaction Analytics — evaluate whether built-in CI is sufficient or layer a dedicated tool on top)
• /sales-enthu — Enthu.AI platform help (contact center QA with auto-scoring on 100% of calls, agent coaching, compliance monitoring, affordable Gong/CallMiner alternative)
• /sales-demodesk — Demodesk platform help (AI conversation intelligence + coaching scorecards + autonomous AI agents, REST API V2, GDPR-native EU)
• /sales-clari-copilot — Clari Copilot platform help (real-time battlecards, live coaching, coaching scorecards, gametapes, Clari Core API, enterprise pricing)
• /sales-salesken — Salesken platform help (real-time in-call coaching, QA automation on 100% of calls, revenue intelligence, field sales analytics, multilingual, Salesforce/HubSpot/Zoho)
• /sales-revenue-io — Revenue.io platform help (Salesforce-native revenue orchestration, RingDNA dialer, Moments real-time coaching, CI on Orchestrate tier, no public API)
• /sales-salesloft — Salesloft Conversations (bolt-on conversation intelligence inside Salesloft)
• /sales-zoominfo — ZoomInfo Chorus (bolt-on conversation intelligence inside ZoomInfo)
• /sales-seismic — Seismic Meeting Intelligence (bolt-on inside Seismic)
• /sales-momentum — Momentum platform help — AI revenue orchestration with optional CI add-on ($30/user/mo), automated CRM updates from calls, MEDDIC Autopilot (acquired by Salesforce Feb 2026)
• /sales-call-review — Review a specific call for coaching
• /sales-coaching — Build coaching programs that consume call data
• /sales-integration — General CRM/tool integration patterns (Zapier, webhooks, iPaaS)
• /sales-do — Not sure which skill to use? The router matches any sales objective to the right skill. Install: npx skills add sales-skills/sales --skill sales-do

Examples

Example 1: SMB team picking a note-taker on a budget

User says: "I run a 6-person SaaS sales team on HubSpot. Which AI note-taker should we pick — we don't want to pay Gong prices." Skill does:

1. Rules out Gong (budget), Otter (thin CRM), Fellow (not sales-first)
2. Short-lists Fathom Business ($20/seat) and Fireflies Business ($19/seat) — both have HubSpot native sync
3. Recommends Fathom if the team values a generous free tier for individual reps and simpler setup; Fireflies if the team wants conversation search + topic analytics
4. Notes Avoma ($29/seat) as the upgrade path if they later want methodology scorecards (MEDDPICC/BANT) Result: Clear 2-option shortlist with a decision tiebreaker, and an upgrade path.

Example 2: Building a transcript pipeline to Snowflake

User says: "I need to pipe every Fireflies transcript into Snowflake so the data team can build dashboards. What's the architecture?" Skill does:

1. Recommends webhook-first architecture (Fireflies Transcription completed event)
2. Sketches the flow: Fireflies webhook → API Gateway/Lambda → validate x-hub-signature → enqueue → worker calls https://api.fireflies.ai/graphql for full transcript → write raw JSON to S3 + normalized rows to Snowflake via Snowpipe
3. Flags GraphQL complexity limits — fetch transcript sentences in one query, fetch audio/summary separately
4. Adds nightly reconcile job that lists last 48h of meetings to catch missed webhooks
5. Points to /sales-integration if they want iPaaS (Workato, Boomi) instead of custom Result: A production-shaped architecture with specific endpoints, known rate-limit pitfalls, and a reconcile safety net.

Example 3: Webhook vs polling for Otter

User says: "We're on Otter and want transcripts in Slack within a minute of a call ending. Webhook or polling?" Skill does:

1. Confirms Otter public API is Enterprise-only beta — asks if they have that tier
2. If yes: use Workspace Webhooks (they exist) — point directly at a Slack-forwarding service
3. If no Enterprise access: fall back to Zapier's Otter integration (pre-built Slack action)
4. Warns that Otter's beta API limits are undocumented — test end-to-end before committing
5. Suggests evaluating Fathom free tier or Fireflies free tier if the real requirement is "cheap + real-time + Slack" Result: Grounded answer specific to Otter's reality, with a fallback if their tier doesn't support it.

Troubleshooting

Webhook fires but transcript is empty or truncated

Symptom: Payload arrives on transcription.completed but the transcript body is partial or blank Cause: Most vendors fire the event as soon as ASR completes — AI summary, speaker diarization, and action items are still being generated on the vendor side Solution: Either subscribe to a separate summary.ready / processing.completed event (Fathom, Fireflies offer one) or re-fetch from the meeting-detail endpoint 60-180 seconds after the initial webhook. Add a small queued retry with exponential backoff.

Rate limits hitting 429 in production

Symptom: Gong or Fireflies returning 429 during transcript backfill or high-volume days Cause: Backfilling in parallel or fanning out transcript fetches for every new webhook without queuing Solution: Queue outbound API calls with a token-bucket rate limiter tuned below the published limit (e.g., 2 req/sec for Gong's 3/sec limit). Honor the Retry-After header, not a fixed sleep. Cache call metadata so you don't re-fetch it on every downstream consumer.

Transcript format differs across vendors — downstream queries break

Symptom: A change from Fathom to Fireflies (or adding a second vendor) breaks every analytics query Cause: No common schema — each vendor's payload is piped into the warehouse verbatim Solution: Add a normalize_transcript(vendor, payload) step between ingest and warehouse. Store raw JSON in S3/GCS (audit/reprocessing), but write only the normalized schema to the warehouse. Common fields: meeting_id, vendor, external_id, started_at, duration_s, participants[], transcript[{speaker, ts_ms, text}], summary_md, action_items[], recording_url. When adding a new vendor, all you touch is the normalizer.