Dvir Ginzburg is the founder and CEO of Encore AI, which helps banks and financial institutions turn customer conversations into measurable revenue using AI-powered digital agents. He tells Laurent Cochet why banks should judge AI agents by the revenue they generate rather than the calls they deflect, how new regulation is helping the industry move, and what he has learned about letting go as the company grows.
From recommendation engines to a digital banker
Ginzburg holds a PhD in computer science and built conversational AI agents at Microsoft and, before that, at Facebook, focusing on recommendation engines that use language models to suggest what a customer might want next. At Encore AI, he says, knowing the right product is no longer enough. The company builds a concierge that guides a client through complex journeys one step at a time, such as opening a deposit account or a new credit account, or buying travel insurance.
The idea started with his mother. She is highly educated and uses an innovative bank, yet every time she opens a deposit account she calls him, because the journey is technical, fixed and linear. As a company, Encore AI has a banker who walks it through the options in a live conversation. Ginzburg says 99.9% of the population does not have a personal banker, and Encore AI aims to give end customers a journey that feels as if they do.
Measuring revenue, not deflection
Ginzburg argues the industry is missing a huge opportunity by judging customer-facing AI agents on deflection, which he describes as the ratio of calls that needed to go to a human representative. Deflection is easy to measure, which he sees as a sign of a young industry. The cost is lost sales from clients who wanted another product, and frustrated customers who leave a bot conversation without asking for a human and are still marked as good conversations.
Encore AI instead takes complex revenue-generating flows in digital channels and turns them into voice or chat interactions, inbound or outbound, which he says convert as well as humans, with agents meant to behave like a bank's top performer. The company has screens showing revenue generated, not deflection rate.
Most organizations separate support and sales conceptually, he says, and many of the clients and prospects his team visits have one room for sales reps and another for customer service. Insurance sales reps may track premium per day, but support reps rarely think to mention extreme sports insurance to a caller flying to the Alps. He describes a bank that launched salary-secured loans that employers can offer their staff. The CEO assumed the product was not good enough, but call analysis showed only 0.1% of reps mentioned it.
Why banks are opening their doors
Encore AI works mainly with retail financial institutions that serve large volumes of retail clients and SMBs. Ginzburg believes founders face an opportunity that will probably not come back: highly regulated institutions, including companies where a single lost lead can cost tens of thousands of dollars, are opening their doors to startups. With earlier analytics tools, traditional banks let innovators fail first; with Gen AI, he says, the biggest incumbents are the first to adopt.
The harder issue is noise. System integrators and banks themselves try to build, and some clients build on top of Encore AI's technology after implementation, which the company endorses. He describes a typical scenario: a bank CEO watches a video of Sam Altman building a full product in 15 minutes, then asks the CIO whether the bank should build it internally.
Regulation, in his view, is helping. Before the EU AI Act, other new regulations and new ISO standards around AI, many banks were a bit afraid to move. Banks once demanded zero hallucinations; he says the new rules set boundaries under which they can be acceptable, provided the bank notifies the client, finds them in real time, mitigates them and makes sure they do not happen again.
Toward a conversational bank
The shift Ginzburg thinks most people underestimate is a move to a fully conversational experience. Instead of opening a mobile banking app, customers would text their bank, on WhatsApp or other channels, for everything from lending and deposits to investments, insurance, balances and money transfers, with two-factor authentication still required but, as he puts it, a technical detail. The most innovative banks and insurers, he says, are working toward one holistic conversational layer.
Customer attitudes are already changing. A year ago, banks could see users move the mouse toward the chatbot bubble and then quickly away, remembering bad bot experiences. Now, especially among younger people, visitors go straight to the bot and treat it like a smart web search. On outbound voice calls, he says, people used to hang up or ask for a representative as soon as they heard an AI agent. Now, in a pre-qualifying system, his team hears clients asking how they can speak with a bot.
Growing the company and letting go
At the time of recording, Ginzburg was living in Europe and preparing to move to the US to grow the business there. Until now he has been a major part of selling, customer success and training. Encore AI recently raised what he calls a very significant round, more than seven times what it had raised before, which lets it hire sales reps from competitors and customer success managers from Fortune 500 companies. Success over the next 12 months, he says, means delighting customers and becoming insignificant in at least 95% of day-to-day decisions.
With more resources, he would accelerate go-to-market, marketing and education, because bank CEOs still see Gen AI mainly as a way to reduce costs, even when, in his meetings, they describe problems such as having 50,000 prospects a day they cannot convert. His advice to founders is that the people around you matter more than strategy or market experience. Coming from a PhD and big-company roles, he also had to learn that trying to do everything yourself means failing.
Watch the full conversation
Watch the full conversation above. Ginzburg suggests LinkedIn as a good starting point for finding Encore AI, and the company's website is gainencore.ai.
AI InterConnect hosts conversations with founders, investors and operators building with AI. Explore more episodes on the AI InterConnect podcast page.