Voicing AI adds live voice diagnostics for regulated contact centers

Voicing AI launched a new diagnostics feature that lets live voice agents analyze call performance and apply fixes without engineering help. The company also pointed to a Latin American telecom deployment that scaled call volume 2.5x in under six weeks and processed about 50,000 calls in its first two weeks live.

Why it matters: - Regulated contact centers need faster ways to fix live voice agents without waiting on engineering teams. - Voicing AI is targeting that gap with diagnostics that can surface issues and apply improvements during production use. - The capability is designed for banking, insurance, healthcare, aviation and telecom customers, where audit trails and controlled changes matter.

What happened: - Voicing AI announced a new voice diagnostics capability inside its enterprise agentic voice AI platform. - The feature lets live voice agents analyze their own call performance and apply improvements without engineering intervention. - The company said the capability is available now to customers in banking, insurance, healthcare, aviation and telecom. - In a Latin American telecom deployment, call volume scaled 2.5x in under six weeks. - The same deployment handled about 50,000 calls in its first two weeks live.

The details: - The diagnostics feature focuses on the period after a voice agent is deployed and handling live production calls. - It reads speech-to-text, text-to-speech and language-model traces from individual calls. - A user can ask a plain-language question about a specific call. - The system reports what worked, what could be adjusted and where performance can be tuned. - The platform can then propose or apply the change. - Voicing AI said the diagnostics capability draws on ten specialist AI agents. - Those agents cover planning, prompting, tool-calling, conversation flow, knowledge retrieval, analytics and reporting. - The platform offers two paths to campaign data. - Teams with SQL skills can build custom charts across call, message and post-call datasets and save dashboards per user. - An Analytics Assistant answers questions in plain language without requiring a query. - Asked to compare call categories by volume, the assistant returns a table and chart that update as a campaign runs. - Post-call analysis includes customer-defined extraction fields, intent classification and a deterministic rule engine. - The rule engine can set, compute or backfill data fields by rule rather than model output. - Results can be pushed to customer systems through post-call actions and webhooks. - The platform supports configurable PII redaction and data retention windows. - Voicing AI said those controls are intended for banking, insurance and healthcare use cases that require an audit trail. - The platform supports six tool-handler types: REST webhooks, custom Python functions, call transfer, DTMF capture, knowledge retrieval and native data-store integration. - It also includes about 20 built-in tools. - Custom code runs in a sandboxed environment with a fixed outbound IP address. - That setup lets the code reach customer APIs behind an IP allowlist. - On call transfer to a human agent, the platform passes context through custom SIP headers. - The headers include a language-model-generated summary of the caller's request. - The telecom deployment cited by the company was an outbound collections program migrated from an external codebase. - The deployment included an outbound campaign engine, a customer-data pipeline, segment-based routing across five conversational entry points, a knowledge base and scheduled data exports.

Between the lines: - The product pitch is less about building a voice agent and more about improving one after it is already in production. - That approach fits enterprise buyers that already have legacy workflows, live campaigns and existing codebases. - The specialist-agent design is meant to narrow each task so suggestions are actionable without human rewriting. - Voicing AI is also trying to quantify improvement speed, tracking the time between a flagged issue and a measurable fix across deployments. - The company plans to publish those results later.

What's next: - Voicing AI said it will continue tracking the time from a flagged issue to a measurable improvement across live deployments. - The company plans to publish those results at a later date. - More information is available in the company's announcement and on Voicing AI's LinkedIn page.

The bottom line: - Voicing AI is betting that live diagnostics and self-improving voice agents will help regulated enterprises move faster after deployment, not just at launch.