Sherlock Calls is a voice AI observability platform that helps engineering and operations teams diagnose call failures, correlate incidents across multiple telephony and AI speech providers, and conduct root-cause investigations directly from Slack. In production voice AI systems, the evidence behind a failed call is often distributed across multiple independent vendors. Telephony logs may exist in platforms such as Twilio or Genesys, speech synthesis records may reside in tools like ElevenLabs or Vapi, and additional information may appear in webhook payloads or CRM interaction histories. Each provider uses different identifiers, timestamps, and definitions of completion. A call that one provider records as successfully completed may have delivered silence to the caller because upstream latency prevented audio from being transmitted in time. In many cases, no single system records this as a failure. As a result, diagnosing incidents requires engineers to manually correlate logs across several dashboards, a process that can take hours per incident. Sherlock Calls addresses this problem by connecting to a team’s existing voice and communication infrastructure using OAuth authentication. The platform aggregates and correlates call-level data from multiple providers without requiring SDK installation, code changes, or additional agents. When a failure pattern is detected, or when a team member submits a natural-language query in Slack, the system generates a structured case file containing a unified cross-provider timeline, a root-cause hypothesis supported by layered evidence, and prioritized triage steps. The resulting report is delivered directly to the Slack thread associated with the alert or investigation. Key capabilities include: * Cross-provider correlation across telephony and voice AI systems including Twilio, ElevenLabs, Vapi, Retell AI, Genesys, Amazon Connect, Aircall, and HubSpot, with additional CRM integrations such as Salesforce and Dynamics 365 under development * Natural-language investigation queries in Slack that allow engineers to request incident analysis and receive structured case files within seconds * Automated anomaly detection that alerts teams to spikes in call failures or degradations in speech synthesis latency * Scheduled operational digests that summarize call volumes, failure rates, and usage trends on a daily or weekly basis * Per-call cost attribution across providers, breaking down expenses by telephony usage, speech synthesis, and large language model inference Sherlock Calls is designed for voice AI engineering teams, customer experience operations groups, and product teams responsible for AI-powered phone agents, IVR systems, and outbound voice campaigns. It is particularly relevant for organizations that operate multiple voice AI providers simultaneously and require a single investigation surface instead of navigating several independent dashboards. A free tier provides 100 investigation credits per Slack workspace. Paid plans start at $50 per month.