[Portus AI] "Humans Decide, AI Acts." The Future of White-Collar Work Transformed by Self-Initiating AI

指示待ちの「自動的なAI」から、自ら提案し実行する「自発的なAI」へ

Hello everyone. I am Kazuki Watanabe, CEO of Portus AI. Thank you for having me today.

We are an AI startup spun out of the University of Tokyo's Matsuo Lab. Guided by our vision of building "human-like AI," we are developing self-initiating AI systems. Both my co-founder, Miyamoto, and I come from academic backgrounds in computer science, specializing in artificial intelligence. We have just entered our third year since founding.

Our goal is to build a world where white-collar tasks are divided between humans and AI: "Humans retain decision-making power, while AI handles the execution." Think of the stereotypical corporate manager who sits at their desk and simply stamps approvals. We want to reduce human tasks to just that final "stamp" of approval, letting AI handle all the heavy lifting before and after.

For example, when an email arrives from a job candidate, the AI reviews past interactions and requirements, drafting the reply automatically. The human only needs to swipe (judge) the draft on their smartphone to send it. Whether it is code reviews or administrative tasks, we want to streamline all white-collar work so that humans only have to make judgments on a single screen.

Current popular tools like ChatGPT are what we call "passive AI." They only respond when we prompt them. This still requires humans to manually copy, paste, and issue commands. In contrast, our "self-initiating AI" instantly grasps the context when an email is received, drafts the response, and presents it to the user just one step before sending, asking, "Is this okay to send?"

AIとブラウザを直結し、24時間Webの変化を監視する「スレア」の革新

As the first step toward this self-initiating AI, we have launched "Sleear," an application that directly connects AI with Web browsers. Sleear automates "Web-based information gathering," which is a core driver of competitiveness for many companies today.

Instead of employees manually checking competitor pricing, regulatory updates, or public bidding portals every day, our autonomous system powered by Large Language Models (LLMs) monitors these websites 24/7 at a fraction of the cost.

Setup is extremely simple. Users only need to provide two inputs: the target URL and the specific information they want to retrieve (e.g., "Extract the latest news from the top-left section of the Yahoo News page"). This initial setup takes only 30 seconds. From then on, whenever a change occurs, the AI automatically extracts, analyzes, and sends notifications via Slack, Teams, or email.

メガバンクも採用。金融リスク監視や入札情報のパーソナライズに威力

Within three months of its beta release, Sleear was adopted by a major Japanese megabank. Megabanks must constantly monitor websites of global financial regulators to track blacklist updates and financial policy changes. While this was previously done manually at a high labor cost, Sleear has dramatically improved operational efficiency.

Sleear is also highly effective for monitoring public procurement and bidding opportunities. Instead of just dumping raw search results, the AI filters and personalizes the information to match the specific needs of each client. For instance, it filters out irrelevant HR-related bids for construction companies, notifying them only of relevant projects.

Currently, we have nearly 70 users who have registered and are monitoring close to 400 targets.

30秒の初期設定だけで、あとはAIがすべてを完結する月額サブスクモデル

Our business model is a straightforward monthly subscription. Pricing starts at 9,000 yen (approx. $60 USD) per month, with tiered plans based on the number of monitored URLs and update frequencies.

The biggest advantage of Sleear is that human intervention is required only during the first 30 seconds of setup. Because users do not need to log in to a dashboard unless they receive an alert, tracking traditional SaaS metrics like daily active users (DAU) is not applicable. Instead, our customer success team actively engages with clients to verify and report the value delivered.

Although it has not been long since the official launch, our customer churn rate is currently at zero, and we are receiving highly positive feedback regarding future integrations.

シードラウンドで1.6億円を調達中。日本の企業の半分が「自発的AI」を使う未来へ

With Sleear, we aim to establish immediate cash flow. Looking ahead, our goal is to have 1 in 20 startups in Tokyo using self-initiating AI in two years, and 50% of all companies in Japan adopting it within five years.

Currently, we are raising a seed round of 160 million yen (approx. $1 million USD) at a pre-money valuation of 620 million yen. We are in the final stages of closing this round. I am currently in Jakarta meeting with local venture capitalists to discuss partnerships.

We are looking for new use cases, client introductions, and collaborative partners. We welcome opportunities to connect with investors who share our vision of University of Tokyo-backed technology transforming the future of work with self-initiating AI. Thank you.

質疑応答・フィードバック

島澤氏（コメンテーター）: Thank you, Mr. Watanabe. This is a very interesting and high-demand area. When was Sleear released?

渡邉氏: We launched the beta version last September and officially released the production version at the beginning of this month.

島澤氏: Is it already generating revenue? And out of the nearly 70 users, how many are paying subscribers?

渡邉氏: Yes, it is already a paid service. Currently, about 20% of our active users are on paid plans.

島澤氏: I see. Since Sleear automatically sends alerts to Slack or email after the initial setup, users don't need to log in to the app, meaning traditional activity metrics might look low. How are you managing customer retention and satisfaction?

渡邉氏: Exactly. Because the value is delivered directly through notifications, users don't need to interact with our dashboard. That is why our customer success team manually validates the utility for our clients. Fortunately, our churn rate is currently zero, and we are getting enthusiastic requests for more integrations.

島澤氏: The AI agent and browser automation space is incredibly hot right now. Is your ultimate goal to integrate with various apps so that users just swipe "OK" or "NO" to get things done?

渡邉氏: Yes, precisely. We want the AI to learn the user's specific preferences and style, creating a world where AI agents proactively handle operations on the user's behalf.

島澤氏: As a startup backed by the prestigious Matsuo Lab, we expect great things from your technology. Thank you very much.

渡邉氏: Thank you. We will do our best.