Encore AI Raises $30M Series A to Clone Your Best Salespeople
Encore AI announced a $30 million Series A on August 7, led by Team8 with Planven, Lukatz, and Garage participating — plus several banks and insurers who used the product first and then wrote checks, which is the kind of investor list you cannot fake (coverage: theaiinsider.tech). Valuation undisclosed. The Israeli company, founded in 2022 as Insait IO and led by CEO Dvir Ginzburg, says ARR has grown more than fivefold since its seed round about eighteen months ago, with over 40 enterprise customers, mostly financial institutions.
The approach is what earns the coverage here. Encore does not sell a generic voice bot. It runs what it calls interaction mining: ingest a company's own call recordings, emails, and texts alongside CRM outcome data, identify which strategies its best human performers actually use in which situations, then train voice agents that replicate those specific patterns — or whisper them to human employees mid-call. The training corpus is the company's own institutional memory, not the internet.
That framing is quietly the interesting bet. Every enterprise sitting on years of recorded customer interactions owns a dataset that describes exactly how its business gets done, and almost none of them have extracted it. Voice agents trained on generic data plateau at generic; agents distilled from your own top decile start where your best people are. It is on-policy distillation, applied to sales floors instead of language models — and the fact that regulated banks are both buying it and investing in it says the compliance story survives contact with reality. U.S. expansion is where the money goes.
← Back to all articles
The approach is what earns the coverage here. Encore does not sell a generic voice bot. It runs what it calls interaction mining: ingest a company's own call recordings, emails, and texts alongside CRM outcome data, identify which strategies its best human performers actually use in which situations, then train voice agents that replicate those specific patterns — or whisper them to human employees mid-call. The training corpus is the company's own institutional memory, not the internet.
That framing is quietly the interesting bet. Every enterprise sitting on years of recorded customer interactions owns a dataset that describes exactly how its business gets done, and almost none of them have extracted it. Voice agents trained on generic data plateau at generic; agents distilled from your own top decile start where your best people are. It is on-policy distillation, applied to sales floors instead of language models — and the fact that regulated banks are both buying it and investing in it says the compliance story survives contact with reality. U.S. expansion is where the money goes.
Comments