Maria Zhang 的 Palona 在融资达到2000万美元之际推出餐厅运营层

Why it matters

Palona is testing whether restaurant AI can advance from taking calls to coordinating workflows across demand, POS data and camera feeds. The financing gives Zhang room to prove that broader claim.

Maria Zhang 于 8 月 17 日推出了更广义版本的 Palona AI，将其餐厅电话代理扩展为一种软件，结合客户对话、业务系统和安全摄像头视频流以触发运营工作。Palona 还在宣布称，其 A 轮融资使总融资（包括已转换的 SAFEs）达到 2,000 万美元。

该融资公告并未明确说明 A 轮的具体规模。Palona 先前在 2025 年 1 月以1,000 万美元种子轮推出，新公布的总额包含早期融资和已转换的 SAFEs。Palona 未披露领投方或估值。

结构很重要，因为 Palona 做出的产品押注比其最初销售的餐厅语音代理更大。Zhang 希望 Palona 能接住一个未接来电，判断这是一个高价值的餐饮询单，将机会分配给经理，并最终利用实时运营信号帮助餐厅完成订单。那将使 Palona 更接近实体业务的日常控制系统，在那里错误会影响厨房、员工和等待的顾客。

“Physical businesses need AI that can understand what is happening and act in real time,” Zhang said in the announcement.

Zhang's return to restaurant software

Restaurants have followed Zhang through two startup cycles. She founded Alike, a mobile app that recommended restaurants and other places based on a user's existing preferences. Yahoo acquired Alike for an undisclosed amount in February 2013 and moved its five-person team into Yahoo's mobile group.

Zhang later held engineering and AI leadership roles at Google, Meta, LinkedIn, Yahoo and Tinder. When she started Proactive AI Lab in June 2024, the project behind Palona, she described a system of specialized agents and multimodal models designed to help businesses act on customer needs. Palona's initial 2025 launch focused on emotionally aware sales agents for consumer-facing businesses, with early deployments outside restaurants as well as at Pizza My Heart.

The current strategy is narrower by industry and wider by workflow. Restaurants give Palona a demanding test environment: calls arrive during service, menus change, orders need to reach a point-of-sale system accurately and operational problems require action within minutes. The data is fragmented across conversations, reservation systems, ordering software and camera feeds.

Zhang built Palona with Tim Howes, the co-creator of LDAP and a co-founder of LoudCloud, Opsware and Rockmelt, and Xue "Steve" Liu, a McGill University professor who previously served as vice president of R&D, chief scientist and co-director of the Samsung AI Center Montreal, and earlier led research and innovation as chief scientist at Tinder. Zhang and Howes first worked together at Yahoo after the Alike acquisition. She later recruited her former manager to become Palona's CTO, describing him in a September 2025 post as a longtime mentor.

That history gives Palona a founding team with experience spanning consumer personalization, AI infrastructure and real-time systems. Palona's challenge is converting that pedigree into reliable software for operators whose tolerance for invented menu items, lost orders or noisy alerts is close to zero.

From answering calls to watching the floor

Palona organizes the new offering into Revenue Expansion, Revenue Intelligence and Operations Excellence. Its Ordering Agent handles calls and messages, answers questions and sends orders to a restaurant's point-of-sale system. Its Catering Agent gathers details such as party size, date, dietary needs and budget before routing an opportunity to a manager.

Revenue Intelligence analyzes conversations for demand, stalled inquiries and conversion problems. Operations Intelligence connects to 现有的安全摄像头 and is designed to flag issues involving food preparation, cleanliness, guest experience and staff compliance.

Palona calls the underlying loop "Capture -> Understand -> Act -> Learn." The important step is "Act." Restaurant software already produces dashboards and reports. Palona is trying to turn detected intent or an operational event into an assigned workflow while the order can still be won or the problem can still be fixed.

Palona says its Interaction Model represents how people, objects, locations and processes relate over time, using spatial, temporal and semantic context. The announcement provides no technical evaluation of that model. A separate, verifiable piece of Palona's intellectual property is U.S. Patent No. 12,481,517, issued in November 2025. The patent covers routing requests among specialized AI agents using factors including user intent, latency, cost, availability and historical performance. It does not by itself validate Palona's restaurant outcomes or the broader Interaction Model claim.

Palona says its software supports systems including Toast, Olo, Square, OpenTable, Resy and Yelp Reservations. Palona's public materials do not establish which integrations are active at each named restaurant operator.

The operating results need a larger denominator

Palona says a multi-brand production study spanning Cali BBQ, Rooted Hospitality and Giordano's recorded 481 orders over 194 location-days and identified 305 large-order and catering inquiries across seven restaurants. At Cali BBQ, the company says Father's Day revenue rose 20% year over year and Palona became the restaurant's highest-average-order-value channel after adding catering and large orders.

Those figures show the type of return Zhang is selling: recovered demand rather than labor reduction alone. An unanswered phone call has an immediate revenue value when the caller is trying to place a large order.

The announcement does not identify the Father's Day comparison periods, baseline revenue, control locations or attribution method. It also does not report how many of the 305 inquiries became paid orders. The 481-order study is production data supplied by Palona, without an independent audit. Palona has not published revenue, annual recurring revenue, pricing or a total customer count.

The customer wording also deserves precision. Palona says it works with Din Tai Fung, Mountain Mike's Pizza, Giordano's, Rooted Hospitality and Cali BBQ. The announcement does not describe the commercial scope of each relationship or distinguish paying deployments from studies and other operating partnerships.

Palona is moving beyond the voice-AI fight

Restaurant ordering is already a competitive market. SoundHound says its voice technology operates in more than 10,000 restaurant locations and covers phone, drive-thru, text, scan-to-order and in-car channels. ConverseNow sells phone ordering, drive-thru automation, text conversations and point-of-sale integrations.

Palona 的做法是将价值单元扩展到已完成电话订单之外。Zhang 正在把下单、餐饮资格审核、对话分析和基于摄像头的运营结合到同一产品系列中。如果 Palona 能可靠地将这些信号连接起来，它就可以针对更大的运营预算进行销售，并构建比独立电话代理更深的集成。

这种扩展也增加了实施负担。摄像头分析引入了隐私、精确性和告警质量的问题。餐饮集团需要证明 Palona 能够识别有意义的事件，而不会用误报淹没管理者。销售点和预订集成必须在不同品牌、不同地点以及菜单变化的情况下保持准确。

Palona 的已披露的投资者团体 包括 Ardenwood Ventures、CrimsonOx、UpHonest Capital、Turbo、Llama Ventures、Neo、Fusion Fund、Defy 和 Maynard Webb。

这 2000 万美元的总额为 Palona 提供了加深部署并测试其餐厅系统能否扩展到其他实体业务的资源。餐厅仍然是决定性的试验场。Zhang 已经证明 Palona 能够接听电话并发现需求。下一阶段需要证明其代理能够在软件与餐厅现场之间以足够的精确性协调工作，从而让运营方将业务更大的一部分托付给 Palona。