STT, LLM, and TTS split
Each call breaks cost down by speech-to-text, language model, and text-to-speech, with token usage alongside, so you can see which stage drove the spend.
Product
Voice AI analytics that follow every call.
Analytics is how operators on the voice AI platform see what agents did, what it cost, and where the margin went. Build the dashboards your team needs, drill into any call, and export the numbers your finance team asks for.
Custom dashboards, twelve metrics, eleven ways to cut them
Build custom analytics dashboards across 12 metrics with 11 breakdown dimensions, so the same call data answers questions by agent, campaign, provider, or any other angle your team needs. Every view exports to CSV or PDF, ready for a board deck or a client report.
Campaign operators slice the same numbers against the guardrails they already run on telephony and outbound campaigns, while contact centre teams use the breakdowns to staff and route on evidence, as described on the contact centres page.
Developers get the same depth through the documented REST API covered on the developers page, so dashboards in the product and data pulled into your own tooling tell one story.
Per-call economics, down to the token
Every metered call carries its full cost story. These are product features visible in the dashboard, not numbers you request from support.
COGS, retail, and margin
Every call records cost of goods, retail price, and margin, so agencies and resellers see true unit economics per client instead of estimating them.
Sourcing-mode attribution
Costs attribute to the sourcing mode used: managed key, your own key with BYOK, or your own infrastructure, so mixed setups still reconcile cleanly.
Self-serve spend and credits
Spend, credits, history, and per-call breakdowns are all self-serve in the dashboard, and the figures line up with the plans on the pricing page.
Your own Langfuse traces
Route call and LLM traces to your own Langfuse project per organisation, so deep observability lands in tooling you control.
Telemetry off by default
Product telemetry stays off unless you turn it on. Your call data is yours, and nothing phones home without your say so.
Frequently asked questions
What can I measure with voice AI analytics in PeachDesk?
You build custom dashboards across 12 metrics with 11 breakdown dimensions, so the same call data answers questions by agent, campaign, provider, or any other cut your team needs. Dashboards are yours to configure, and every view exports to CSV or PDF for reporting outside the product.
How do I see what each call costs?
Every metered call carries a per-call cost breakdown with the STT, LLM, and TTS split plus token usage, so you can see exactly where the money went. PeachDesk also records cost of goods, retail price, and margin per call, attributed to the sourcing mode used: managed key, your own key, or your own infrastructure.
Can I check spend and credits without contacting support?
Yes. Spend, credits, history, and per-call breakdowns are all self-serve inside the dashboard. Operators and agencies can reconcile usage against invoices on their own schedule, and export what they need for finance or client reporting.
Does PeachDesk collect telemetry by default?
No. Product telemetry is off unless you turn it on. If you want deeper observability, you route call and LLM traces to your own Langfuse project per organisation, so trace data lands in infrastructure you control rather than ours.