dealism.ai adds Copilot and Autopilot modes to Inbox

dealism.ai outlined two modes for Inbox: Copilot and Autopilot. In Copilot, users select the Sales Agent in a conversation and click Generate. The agent analyzes the ongoing sales task and the customer’s needs, then generates a recommended reply in the input field. In Autopilot, the system automatically generates and sends an appropriate reply each time a new message arrives, which the company says enables 24/7 automated sales operations.

According to dealism.ai, Autopilot automatically sends replies to new messages, making it possible to design sales response times without relying on staff. Copilot retains human review before sending, while Autopilot can be applied to all contacts, making differences in operating rules a likely factor in whether businesses adopt it.

The key point about dealism.ai’s Inbox is that it distinguishes between Copilot and Autopilot rather than offering only reply assistance. The former is semi-automated, with a person reviewing the reply before sending; the latter automatically sends a response when a message arrives. The two modes therefore have different boundaries of responsibility, even when using the same Sales Agent. The new element is a design that lets users switch from support on a conversation-by-conversation basis to automation across all customers. While it builds on existing chat-assistance features, its operational implications are substantially different. In our view, the focus is less on the label “sales support” than on how much authority over replies and execution timing is delegated to the system. Copilot uses a knowledge base and sales strategy to draft a reply, but a person makes the final send. With Autopilot, responses are sent automatically after messages arrive, making controls against misdirected replies and inappropriate wording central to implementation. In practice, the difference will depend less on the feature names than on the conditions under which automation can be disabled or controlled. Structurally, the input is a new customer message, the processing involves the Sales Agent analyzing the conversation and generating a reply, and the output is a sent message. There is also a pathway for adding supplementary knowledge content, suggesting that the system is designed not as a one-off generation feature but for operations that adjust response quality while updating the knowledge base. However, the company has not specified which knowledge sources are referenced, how much conversation history is used or what exception settings are available when applying the system to all contacts. Unresolved questions include Autopilot’s conditions for stopping, the correction process for incorrect responses, the scope of the knowledge base referenced by the Sales Agent, and the level of detail available for settings that retain human review. Until these points are clarified, evaluating the feature will depend less on the “strength” of its automation than on the granularity of its controls.