ThunderPhone v2

ThunderPhone builds and runs AI phone agents for inbound front-desk coverage, outbound campaigns, scheduling, pre-sales intake, and regulated call handling. The core workflow runs through a no-code operator dashboard where you configure agents, upload knowledge base documents, set outbound campaign pacing and retry logic, and monitor live calls — or you reach for the API when the platform UI runs out of room. Calls route over PSTN or SIP using existing telephony, or embed directly into a web or mobile app. The agent handles interruptions, cross-talk, and background noise without losing its place, captures DTMF keypad input for IVR flows, detects voicemail on outbound, and preserves context through warm and cold transfers to human agents.

The standout differentiator is audio accuracy under real-world conditions. ThunderPhone claims the top score on Big Bench Audio — a public dataset and benchmark measuring voice model responses to challenging inputs — at 99.4% as of July 2026. That benchmark gap matters in practice: it is the difference between an agent that confirms ‘EX-350’ and one that ships parts to a site for the wrong model number. The vendor publishes the dataset on Hugging Face for independent verification.

ThunderPhone fits teams with high call volume, noisy or accented callers, multilingual caller bases, or compliance requirements (HIPAA BAA and GDPR DPA are available at no extra charge, the vendor states). The no-code platform handles most production agent configurations, but call flows that require complex conditional branching — routing based on what a caller said three turns ago, or dynamic logic across multiple agent states — push teams to the API layer. At that point you are running a hybrid setup: the platform for operations and monitoring, custom code for flow logic. Teams that need that level of branching from day one should evaluate whether starting API-first is cleaner than migrating later.

TCPA-compliant cold outbound is available through a separate product called ThunderHuman, not the core platform — so teams planning cold outreach campaigns need to confirm that path before committing to the stack. Live supervision lets operators watch active calls and whisper guidance to the AI mid-conversation, which is useful for QA and agent tuning without pulling the call to a human. A/B experiment tooling and AI-graded test calls are included in the operator dashboard.