Why AI meters from your prepaid wallet (not your own keys)

Live coach, post-call analysis, and transcription run on DialProng’s Anthropic and OpenAI lanes. Usage debits the same prepaid wallet as calling minutes — no API keys to paste, rotate, or leak across clients.

“Why don’t I bring my own OpenAI and Anthropic keys?”

Anthropic lane

Live coach (in-call)
Post-call analysis
Notes & tasks extraction

OpenAI lane

Live chunk transcription
Post-call Whisper
Audio only — never coach LLM

Platform keys · usage meters from your prepaid wallet · Gemini unsupported

Customer-held keys sound transparent — until every floor needs a vault

BYOK AI forces every agency to create, fund, and rotate Anthropic + OpenAI projects per client (or share one risky master key). Rate limits, billing alerts, and key leaks become your ops problem.

DialProng keeps model credentials in the platform vault. Coach and ASR usage meters from your prepaid wallet at a published markup on our provider cost — one balance for minutes and AI, hard-stopped before it goes negative.

Two lanes, two providers — by design

We route tasks to the right model tier instead of one-size-fits-all GPT.

Anthropic — live in-call coach + post-call reasoning (Sonnet / Fable-class analysis)
OpenAI — audio only: live chunk transcription + post-call Whisper
Gemini is not part of the stack — reasoning and ASR never mix lanes

Agencies can still package “managed AI” to clients

Because usage is metered per floor, you can include coaching and analysis in a managed outbound package and re-bill clients from Agency → Billing — without handing them Anthropic or OpenAI consoles.

← Why DialProng runs the carrier (and meters minutes from your wallet)Agency dialer vs single-tenant software →

Customer-held keys sound transparent — until every floor needs a vault

BYOK AI forces every agency to create, fund, and rotate Anthropic + OpenAI projects per client (or share one risky master key). Rate limits, billing alerts, and key leaks become your ops problem.