Instinct and Meta are testing artificial intelligence agents capable of placing outbound phone calls to businesses on behalf of users. Instinct founder Noah Shinn announced the Instinct Concierge feature on Sept. 16, which allows the company's AI assistant to handle service tasks such as calling restaurants or resolving billing issues. Meta launched its personal AI agent, Muse, on Sept. 8 and is reportedly testing a similar outbound-calling capability for U.S. businesses.
The development marks a shift from AI assistants that provide information to agents that execute actions across connected services. Instinct's assistant can already work across apps and websites, while Muse can browse the web, fill out forms, and complete tasks. Both companies are moving toward systems where users provide a goal, and the agent manages the conversation and follow-up without direct human intervention during the call.
Instinct is rolling out Concierge in early access to some users, with plans to expand availability over time. The company recently secured significant funding to support this expansion. In August, Instinct raised $250 million in a Series B round that valued the company at $2.5 billion. On Sept. 28, it announced a $1 billion Series C round from investors including Sequoia Capital, Benchmark, and Coatue, valuing Instinct at $10 billion. Instinct does not currently publish a monthly subscription price for its service or its phone-calling feature.
Meta's Muse is available in the U.S. and Canada, with a free tier that includes usage limits. Reports indicate Muse reached the top of Apple's U.S. App Store shortly after its launch and accumulated over 3 million downloads by late September. Paid plans have been reported at $20 and $100 per month, though Meta has not announced separate pricing for the phone-calling capability currently in limited testing.
Security and privacy considerations accompany these new capabilities. Instinct's privacy policy states that its assistant can access messages, emails, voice data, account credentials, and payment information from connected services with user permission. The company notes that health-related information may also be processed if users request medical appointments. Users can opt out of having certain information used for future AI model training, though Instinct clarifies that previously trained models may retain prior improvements. Data received directly through Google Workspace APIs is excluded from model training.
Meta states that Muse runs in a dedicated secure virtual machine in the cloud, with connected credentials stored securely so the agent cannot see passwords or payment methods. The system requires user approval for sensitive actions like sending emails or making purchases. For transactions, Muse can generate a one-time card number to protect the user's actual card details. Meta asserts that Muse conversations are not shared with its advertising systems, and users can opt out of AI model training using their interactions.
Both companies warn that AI agents can produce incorrect or incomplete information and may take actions that are not easily reversible. Instinct specifies that authorized actions can create binding agreements as if the user entered them personally. Users are advised to verify actions taken by their agents and to review connected account permissions regularly to limit access to sensitive data.