AI Rudder has launched CALL-E, a product that allows AI agents to make real phone calls on their users' behalf. The Singapore-headquartered company, which has handled more than 80 billion AI calls, is positioning CALL-E as a solution to a growing user demand: handing off everyday errands like moving a dinner booking or chasing the dry cleaner to personal AI agents. CALL-E works inside ChatGPT, personal AI agents such as Muse, OpenClaw and Hermes Agent, and AI coding tools including Claude Code, Codex and Cursor.
The announcement matters because personal AI agents are proliferating rapidly. Meta's Muse passed five million US downloads in its first three weeks, and Grok Bot reached 400,000 weekly users in its first month. Both are piloting phone calls, but as AI Rudder CEO Teng Ren noted, even the smartest AI model can fail a phone call. A single 'mm-hmm' from a restaurant host can derail an agent that misreads agreement as a question, leading to a booked time that was never actually offered. Such unnatural behavior can prompt a human on the other end to hang up, undermining the entire premise of delegation.
CALL-E addresses this by building an approval step into every call: a call only proceeds after the user approves the exact plan, so no unapproved call ever rings. After the call, CALL-E reports what was agreed, how confident it is that the goal was met, and the exact lines from the call that support the conclusion, giving users verifiable proof rather than a simple 'Booked for 7pm' that could be wrong. The system also introduces itself by default as CALL-E, a voice agent, with a pre-check to prevent impersonation of authorities like the FBI or 911.
The technical challenge of real phone conversations is significant, and AI Rudder claims its Helm Swift model, which detects interruptions, has the joint-lowest false-positive rate on the TurnBench Interruption leaderboard as of October 2026. That matters because false positives cause an agent to stop talking unnecessarily, breaking the flow of a call. By handling pauses, interruptions and overlapping speech more naturally, CALL-E aims to make AI agents indistinguishable from competent human callers in the moments that count.
Regulatory implications are equally important. The FCC's 2024 Declaratory Ruling (FCC 24-17) clarified that AI-generated voices fall under the Telephone Consumer Protection Act, subjecting such calls to consent and caller identification obligations. Separately, Article 50 of the EU AI Act requires that people be told they are talking to an AI. CALL-E's default disclosure and approval workflow align with these emerging rules, potentially setting a compliance benchmark for the sector. Security is another concern, since an agent making calls carries the user's phone number and errand details. CALL-E runs on AI Rudder's enterprise technology, built to meet SOC 2 Type 2 and ISO 27001 alongside GDPR and PDPA requirements, and keeps a record from plan to result with approval and evidence attached so disputed calls can be traced.
CALL-E is available at https://www.heycall-e.com/, where new users start with 100 free credits. The launch signals that the next phase of AI agent competition will be fought not just on model intelligence but on trust, verifiability and regulatory readiness in the messy, high-stakes world of real-time voice communication.


