A New Route to Sales Intelligence: Mining Conversations Instead of Metrics

Listening to What Works

Encore AI has secured $30 million in Series A funding led by Team8, with participation from Planven, Lukatz, and Garage, plus a handful of financial institutions that first used the product before writing checks. The round underscores a shift in how enterprise AI agents are being trained: not by feeding them generic playbooks or policy documents, but by analyzing the actual interactions between employees and customers to identify which behaviors lead to conversions, renewals, or resolved tickets.

Founded in 2022 as Insait IO by CEO Dvir Ginzburg, the company initially built recommendation software for financial advisers and relationship managers. Over the past 18 months, it has pivoted into what Ginzburg calls "interaction mining," a process that ingests call recordings, emails, and text messages, maps them to CRM data, and breaks them into stages to determine which conversational tactics moved deals forward and which stalled them.

The result is an AI agent that doesn't just follow a fixed script. It adapts based on the strategies that have proven effective within a specific organization, sometimes replicating the jokes, anecdotes, or framings that top performers use. The system packages these learnings into agents that can either operate autonomously or serve as real-time assistants, surfacing recommended responses while a human employee is mid-conversation.

The Mechanics of Interaction Mining

Encore's platform collects unstructured conversational data from multiple channels, then connects it to structured CRM records. The system divides each customer journey into discrete stages and correlates conversational behavior with outcomes such as closed deals, contract renewals, or escalated support tickets.

This granular analysis allows the platform to identify not just which employees are strong performers overall, but which individuals excel at particular stages of a process. One sales rep might be especially effective at discovery calls, while another shines during objection handling. Encore's agents synthesize these stage-specific strengths into a single conversational model that can be deployed across the organization.

The platform also surfaces operational insights: where existing processes create friction, which talking points consistently fail to resonate, and which customer segments require different approaches. For enterprises, this doubles as a diagnostic tool, exposing gaps that would otherwise remain invisible in dashboards tracking only closed-won rates or average handle time.

Deployment Models and Customer Base

Encore's agents can communicate directly with customers via voice or text, handling routine inquiries, qualification, or follow-up autonomously. Alternatively, they can operate in co-pilot mode, listening to live conversations and prompting human agents with suggested responses, relevant case studies, or reminders about what has worked with similar customers in the past.

The company now serves more than 40 enterprise customers globally, with the majority coming from financial services. Ginzburg disclosed that annual recurring revenue has grown more than fivefold since the seed round closed less than 18 months ago, though he did not share absolute figures or the Series A valuation. Notably, several of the investors in this round are banks and insurers that adopted the product before deciding to back the company financially.

The CRM Incumbent Question

Encore is early to the interaction-mining category, but it faces a structural challenge: large CRM incumbents like Salesforce, SAP, Zoho, and HubSpot already sit on vast troves of customer data and are rapidly embedding generative AI into their platforms. These vendors could, in theory, build similar conversational analysis features and distribute them to millions of existing users.

Ginzburg argues that access to data alone is insufficient. The major CRM players have historically treated customer interactions as auxiliary information, logged after the fact and rarely ingested as training material for predictive or generative models. To replicate Encore's approach, they would need to overhaul their data pipelines, persuade customers to share call recordings and message transcripts at scale, and rebuild agent architectures around conversational history rather than form fills and status updates.

Whether that technical and organizational inertia will protect Encore's position remains an open question. At DailyTechWire, we've tracked several verticalized AI agent startups that gained early traction by solving workflow problems incumbents ignored, only to see those incumbents eventually ship competitive features once the market validated the use case. Encore's bet is that its head start in conversation-centric architecture and its deep relationships with financial institutions will create enough switching cost to sustain differentiation.

Capital Deployment and Regional Expansion

Encore plans to allocate the Series A proceeds primarily toward expanding its U.S. sales organization and onboarding additional large financial institutions. The company's current customer base skews heavily toward banks, insurers, and wealth managers, sectors where compliance, risk, and relationship continuity make conversational nuance especially valuable.

The funding also positions Encore to explore adjacent verticals. Healthcare, telecommunications, and B2B SaaS all generate high volumes of customer interactions and face similar challenges in translating employee expertise into scalable AI behavior. Whether Encore will pursue horizontal expansion or deepen its vertical focus in financial services will likely depend on how quickly it can demonstrate ROI in its existing accounts and whether enterprise buyers outside finance prove willing to share the conversational data the platform requires.

What This Signals About Agent Training

Encore's approach reflects a broader trend in enterprise AI: the recognition that generic foundation models, even when fine-tuned on industry corpora, often lack the context to handle the idiosyncratic processes and relationship dynamics that define success in specific organizations. Training agents on proprietary conversational data lets companies encode their own institutional knowledge, including the unwritten norms, cultural references, and rhetorical strategies that don't appear in official playbooks.

This method also introduces new operational dependencies. Companies must commit to capturing and storing conversational data at scale, navigate privacy and consent frameworks, and accept that their agents will inherit not just best practices but also the biases and blind spots embedded in historical interactions. If the top-performing sales reps in the dataset consistently upsell certain products or deprioritize certain customer segments, the agents trained on their behavior will likely reproduce those patterns.

The funding round suggests that investors and early customers believe the upside outweighs these risks, at least in sectors like financial services where relationship continuity and personalized advice are core to the business model. Whether interaction mining becomes a standard component of enterprise AI stacks or remains a niche capability for high-touch industries will depend on how well platforms like Encore can demonstrate measurable lift in conversion rates, customer satisfaction, and operational efficiency over the next 12 to 24 months.