Encore AI raises $30M to turn customer interactions into autonomous agents
The company formerly known as Insait says its agents learn from an organization's own top performers rather than generic playbooks. Its headline ROI claim, though, comes from the company itself.
Encore AI, the Israeli startup formerly known as Insait, has raised a $30 million Series A led by Team8, with participation from investors including customer financial institutions. TechCrunch reported the round Wednesday morning, and the company confirmed it in coverage from Calcalist.
Mining the call logs
Encore's platform analyzes millions of customer interactions, spanning phone calls, chats, emails, and CRM records, to identify how an organization's top-performing employees actually handle customers. It then converts those patterns into autonomous agents or real-time copilots for the rest of the workforce. The company calls the patented approach Interaction Mining, and says it combines structured business data with voice and text analysis in a single model of how work gets done. The same underlying system can either act autonomously or sit beside a human employee as a live coach, depending on the risk tolerance of the workflow.
Founded in 2022 by Dr. Dvir Ginzburg, Encore serves banks, insurers, and healthcare organizations across Israel, Australia, Europe, and the United States. The customer list skews heavily regulated, which is not an accident: those are the industries with the deepest archives of recorded interactions and the strongest compliance reasons to keep agent behavior anchored to precedent.
The claims, and who is making them
Encore says its agents can deploy in weeks rather than months because they ingest interaction data a company already has, skipping the long requirements-gathering phase that stalls most enterprise agent projects. The company also cites a tenfold return on investment within months for a large lending client. Both figures are company-provided. No independent benchmark or named customer accompanies the ROI claim, so it should be read as a sales number until a customer says it out loud.
The lead investor is worth a note as well. Team8 is a venture group with deep roots in Israeli enterprise software and security, and the presence of customer financial institutions in the syndicate suggests at least some buyers liked the product enough to want equity. That is a stronger signal than the ROI slide, though still not an audit.
Copying your best people, at scale
The pitch inverts the standard agent playbook. Most vendors ship a generic agent and ask the customer to prompt it into competence. Encore starts from the org's own best humans and works backward, betting that a bank's top loan officer already encodes more institutional judgment than any prompt library. It is the same precedent-over-generality thesis showing up across the category, from security triage to sales, and it arrives as platform players race to put agents in front of customers, as GaaS News covered in our report on the HubSpot Agent Hub public beta.
The money keeps arriving too. Agent infrastructure and distribution deals have dominated the funding cycle this month, up to and including Stripe's reported interest in OpenRouter, which GaaS News covered last week.
For the agentic AI as a service market, Encore's round sharpens a question buyers should be asking every vendor: where does the agent's judgment come from? Agents trained on an organization's own interaction history promise faster deployment and fewer off-script surprises, but they also inherit whatever bad habits the top performers had. If the approach works in lending and insurance, expect every horizontal agent platform to bolt on an interaction-mining story within a year, and expect the proof, as always, to lag the pitch.