Step 1 — Gather context
If references/learnings.md exists, read it first for accumulated platform knowledge.
- What do you need help with?
- A) Setting up recording, transcription, or AI summaries
- B) CRM integration (Salesforce, HubSpot, Pipedrive)
- C) AI coaching scorecards and performance dashboards
- D) AI Crew autonomous agents (deal rescue, follow-ups)
- E) API integration (transcripts, recordings, webhooks)
- F) Comparing Demodesk to alternatives
- G) Troubleshooting (transcription accuracy, speaker ID, language detection)
-
Which Demodesk modules are you using?
- A) AI Assistant only (recording + transcription + summaries)
- B) AI Coach (scorecards, feedback, dashboards)
- C) AI Analyst (pipeline insights, "Ask AI Analyst")
- D) AI CRM Concierge (zero-setup CRM field detection)
- E) AI Crew (autonomous workflow agents)
- F) Multiple / all of the above
-
Which CRM?
- A) Salesforce
- B) HubSpot
- C) Pipedrive
- D) None / not connected yet
Skip-ahead rule: if the user's prompt already contains enough context, skip to Step 2.
Step 2 — Route or answer directly
| Problem domain | Route to |
|---|---|
| Choosing between Demodesk and other note-takers | /sales-note-taker {user's question} |
| Building a coaching program or framework | /sales-coaching {user's question} |
| Reviewing a specific sales call | /sales-call-review {user's question} |
| CRM integration strategy across tools | /sales-integration {user's question} |
| Meeting scheduling optimization | /sales-meeting-scheduler {user's question} |
When routing to another skill, provide the exact command.
Step 3 — Demodesk platform reference
Read references/platform-guide.md for the full platform reference — modules, pricing, integrations, data model, API endpoints, workflows.
Answer the user's question using only the relevant section. Don't dump the full reference.
For API-specific questions, also read references/demodesk-api-reference.md for endpoint details, authentication, rate limits, and webhook events.
Step 4 — Actionable guidance
Focus on the user's specific situation:
- CRM sync issues: Check which CRM is connected, verify field mapping in Settings > Integrations, confirm the meeting was recorded (not just scheduled). AI CRM Concierge detects fields automatically but needs human approval before first sync.
- Transcription problems: Verify the meeting had audio (not screen-share only). Language detection is automatic but can fail with code-switching or strong accents — check if manual language override is available.
- Scorecard setup: Start with built-in templates (BANT, MEDDIC, SPIN), then customize. Scores are 1-5 per question. Export via API
/recordings/{token}/scorecards. - AI Crew pricing: Usage-based by runs (each agent execution = 1 run). Starter (1,000 runs/mo) is free and included with every plan; Growth (5,000 runs) €299/mo; Scale (25,000 runs) €999/mo; Enterprise unlimited custom. Start on the free Starter tier to gauge run volume before upgrading.
If you discover a gotcha, workaround, or tip not covered in references/learnings.md, append it there.
Gotchas
Best-effort from research — review these, especially items about plan-gated features and integration gotchas that may be outdated.
- Full Google account access required — Demodesk requests broad Google permissions for calendar/recording. No granular scope options.
- No bot announcement by default — Demodesk records silently (no bot joins call). Users must handle consent/disclosure themselves. Check local recording laws.
- AI Crew has a free Starter tier — 1,000 runs/mo are included free with every plan. Higher capacity is usage-based by runs: Growth 5,000 runs €299/mo, Scale 25,000 runs €999/mo, Enterprise unlimited custom (priced in "runs," not "events").
- Webhook setup requires emailing support — you can't self-serve webhook endpoints. Email support@demodesk.com with your endpoint URLs.
- V1 API is legacy — use V2 (
/api/v2/) for recordings, transcripts, summaries, scorecards. V1 still works for meetings/users but isn't getting new features. - Rate limits are per API key, not per user — 120 req/min global, 60 req/min on the recordings endpoint, 20 req/min on listing endpoints. Batch transcript endpoint handles up to 100 tokens per request.
- Auth uses Bearer tokens — current V2 scheme is
Authorization: Bearer YOUR_API_KEY; the legacyapi-key:header is deprecated but still accepted. - Pricing is in EUR — Coaching & AI plan is €49/user/mo monthly, €39.20/user/mo annual (20% off); API access included. Free viewer seats; meeting invitees don't need a seat.
Related skills
/sales-note-taker— Choosing between AI note-takers (Fathom, Fireflies, Avoma, Gong, etc.) or wiring their APIs into downstream systems/sales-coaching— Building coaching programs, onboarding, role-plays, measuring coaching impact/sales-call-review— Reviewing a specific sales call for coaching feedback/sales-integration— Connecting sales tools via webhooks, Zapier, Make, or API/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: CRM fields not updating
User says: "Demodesk isn't syncing call notes to Salesforce after meetings" Skill does: Checks CRM connection status, verifies AI CRM Concierge field mapping, confirms meeting was recorded and transcription completed, walks through Salesforce event object mapping and duplicate prevention settings.
Example 2: Custom coaching scorecards
User says: "I want to score my team's discovery calls on MEDDIC criteria" Skill does: Walks through creating a custom scorecard template with MEDDIC dimensions (Metrics, Economic Buyer, Decision Criteria, Decision Process, Identify Pain, Champion), configuring 1-5 scoring per question, setting up team performance dashboards, and exporting scores via the API.
Example 3: Transcript pipeline
User says: "How do I download all team transcripts into our data warehouse?"
Skill does: Covers V2 API authentication, listing recordings with filters, cursor-based pagination, batch transcript download (POST /transcripts/batch with up to 100 tokens), rate limit handling (120 req/min, 429 with Retry-After), and webhook-driven pipeline using recording.transcription_postprocessed event.
Troubleshooting
Transcription in wrong language or speakers mixed up
Cause: Automatic language detection fails with code-switching (e.g., switching between English and German mid-call) or strong accents. Speaker diarization struggles with similar-sounding voices. Solution: Check if the meeting had clear audio (no background noise). Demodesk supports 98 languages but auto-detection can misfire — look for manual language override in meeting settings. For speaker attribution, ensure all participants joined with distinct audio sources (not sharing a mic).
AI summaries contradict what was actually said
Cause: AI notes are generated from transcripts — if the transcript has errors, the summary inherits them. Notes can also be placed into irrelevant templates. Solution: Always verify summaries against the raw transcript. Check that the correct note template was applied. Report persistent inaccuracies to Demodesk support for model improvement.
Webhook events not arriving
Cause: Webhook setup requires manual activation by Demodesk support. Self-service webhook configuration is not available.
Solution: Email support@demodesk.com with your endpoint URLs and which events you need (e.g., recording.transcription_postprocessed, demo.ended). Verify your endpoint returns 200 within the timeout window. Check for 429 rate limiting if processing is slow.