Encore AI Raises $30 Million to Train Agents From Real Customer Conversations

Encore AI has raised a $30 million Series A led by Team8 for a platform that learns from customer interactions and deploys voice or text agents based on the methods that produce successful outcomes. The company was founded in 2022 as Insait IO and originally built recommendation tools for financial advisers.

The rebranded product collects call recordings, emails and messages, then connects them with customer-relationship-management data. It divides interactions into stages and identifies which explanations, examples or tactics helped move a sale or support case forward.

Encore calls the process interaction mining. The resulting playbooks can guide human employees in real time or operate through autonomous agents. CEO Dvir Ginzburg says the systems may reproduce useful anecdotes or conversational techniques used by a company's strongest relationship managers.

The startup reports more than 40 enterprise customers, mainly financial institutions, and says annual recurring revenue has increased more than fivefold since its seed round. It did not disclose revenue or valuation. Financial services offers valuable structured outcomes, but it also requires strict supervision of communications and advice.

Conversation data includes personal information about customers who did not choose to train an AI agent. Encore's clients must address consent, retention and regional recording laws before transferring calls or messages. The platform should support redaction and purpose limits so that learning from successful interactions does not become unlimited reuse of every customer statement.

Learning from customer calls can shorten the time required to configure an agent for sales or support. It can also reproduce the habits of a top performer that are persuasive but inconsistent with policy. Encore should separate effective language from unauthorized promises, discrimination or workarounds that employees developed informally. Customers need evaluation sets representing difficult and unusual cases, not only successful conversations. If the system can show which examples shaped an answer and let managers remove them, it may become a controlled training platform rather than an opaque imitation of historical behavior.

Learning from successful employees can improve consistency, but it may also reproduce bad habits, bias or statements that compliance teams would reject. Encore must show which source interactions influenced an agent and let customers approve playbooks before deployment. The technology becomes more valuable when it captures institutional knowledge without turning every accidental pattern into company policy.