AI agents can now place calls to real businesses through CALL-E, a tool launched by Singapore-based AI Rudder. The useful twist is that users must approve the proposed call plan before it dials, and get a transcript and summary afterwards.
Watch Desk analysis
What happened
DANAMIC reports that CALL-E launched on 13 November 2026 and works with agents including ChatGPT, Muse and OpenClaw. The agent works out the errand, CALL-E prepares the conversation, and a user approves the plan before the call is made. Users then receive a summary, full transcript, confidence assessment and evidence of what was agreed.
The service can connect through ChatGPT, an OpenClaw integration, developer tools and an API. For now, calls are limited to US and Canadian phone numbers, so the Singapore launch does not yet mean an agent can ring your local clinic.
Why it matters
This takes an AI assistant from drafting a call script to handling a real conversation. That could be handy for routine errands, but it also means the agent is speaking to another person and potentially making commitments on the user’s behalf. Approval before each call and a written record afterwards are practical safeguards, not decorative extras.
The geographic limit matters too: the feature may be available to users elsewhere, but the businesses it can call must be in the US or Canada. And while CALL-E’s launch material says it introduces itself as a voice agent by default, that wording can be changed.
Our read
The best part is the paper trail: a transcript and summary give users something concrete to check, rather than a chirpy “all sorted” and crossed fingers. The approval step also keeps the agent from turning a vague request into an unapproved phone call. If you try it, check the plan before authorising it and read what the business actually said afterwards.
What to watch
- Whether CALL-E expands calling beyond US and Canadian numbers.
- Whether users can keep clear AI disclosure in place when customising the introduction.
- How reliably the summaries and transcripts capture what was agreed.
- Whether OpenClaw and other integrations make the approval step clear before a call begins.
Discussion spark: Should an AI agent be allowed to make a call once you approve its plan, or should you be able to listen in before it can agree to anything?
Sources and evidence
- Your AI Agent Can Now Dial the Phone for You, With One Catch – DANAMIC (21 November 2026, 11:03 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.