AI Usage & Billing¶
Billing → AI usage (also in the Admin group of the sidebar) shows what the AI features cost your organization, priced per call at the rates of the model that served it, and broken down by feature, by model and by day.
Owners and admins can open it. Periods are Day, Week (Monday to Sunday, UTC) and Month — the same windows as the bot-flow AI Usage tab, so the two pages reconcile.
What is counted¶
Every AI call made on your organization's behalf is recorded with the feature that spent it:
| Feature | What it covers |
|---|---|
| Bot flows | AI nodes in published flows: responses, intent classification, knowledge-base lookups, transcription of voice notes, smart queries |
| Call Intelligence | Call transcription, analysis, and analysis repair |
| Meetings | Meeting transcription and Minutes of Meeting generation, including regenerations |
| Voice agent | Each completed live voice call, plus its transcript analysis |
| Inbox AI | Auto-replies, intent classification, the internal assistant, the web assistant, conversation analysis and summaries, reply suggestions |
| Re-engagement | Generated re-engagement messages, sequence content and upsell drafts |
| Other | Product-catalogue embeddings and document ingestion for the AI knowledge base |
Not counted, on purpose: message translation (not an AI model), rule-based auto-tagging, contact enrichment (data vendors, not a model), documentation search (runs on ConnectGain's own key, at ConnectGain's cost), and the Test button on the AI provider settings (a fixed handful of tokens).
Embeddings are the one estimate: the provider does not report token counts for them, so they are estimated from the text length and marked as estimated.
What you see¶
- KPI cards — spend, tokens, calls, audio minutes and cost per minute for the period
- Per-feature table with share bars
- Daily trend, stacked by feature
- Model mix, overall or for one feature
Call Intelligence shows the same figures where the spend happens: this month's real spend on the dashboard card, and a per-stage cost breakdown (transcription / analysis / repair) on each call. Meetings show the transcription-versus-minutes split under an AI cost tab.
Exporting the report¶
Click Export at the top of the page to save the report for the period you are viewing — for example to attach to a renewal or share with finance. Choose:
- Export PDF — your browser's print dialog opens: choose Save as PDF as the destination. The PDF carries the ConnectGain logo, your organization name, the period and when it was generated, followed by the KPI figures, spend by feature (with a total), the model mix and spend per day. Arabic and other right-to-left languages print correctly.
- Export CSV — downloads the same figures as a spreadsheet file that opens in Excel or Google Sheets. Numbers are plain values you can sum and chart: costs in US dollars, token counts in full and audio in seconds.
How costs are priced¶
Each call is priced at the published list price of the model that served it — Gemini by default, or the vendor Call Intelligence is configured for (OpenAI, Anthropic, Z.ai/GLM, or a self-hosted model at $0). These are estimates for attribution, not invoices: with your own Gemini key, Google's invoice is the source of truth.
Transcription is priced as audio input plus text output, not as a single input figure.
Calls and meetings processed before AI usage tracking was introduced have token counts but no priced records. Those screens say so rather than show a guessed figure.
See also¶
ConnectGain — omnichannel inbox, CRM & automation for WhatsApp, Messenger, Instagram, Telegram and more. Open the app · Docs home