Catch A startup has emerged out of the shadows with seed funding worth $5 million to develop an AI agent assistant that will take on administrative duties, which consume a substantial portion of a busy executive’s workday.
The session was jointly led by Entrée Capital You can also find out more about the following: PitangoThe event will be attended by Seedcamp You can also find out more about the following: Factorial Capital. Catch aims to do something harder than just build an AI that summarises emails or suggests what the user should next do. It wants to allow executives to give software complete administrative tasks and then have it completed.
Nir Sabato and Yoav Ramon founded Catch, a software that focuses on the administration of executives, such as scheduling, managing emails, tracking travels, calendar clashes, sending reminders to employees, and making outbound telephone calls. Catch claims that its software has been used to schedule more than 12000 meetings, handle more than 200,000 emails, and complete more than 20,000 tasks.
AI from Assistance to Delegation
Agentic AI faces a major challenge in determining the difference between helping with a job and actually finishing it.
The large language models are already able to summarize an inbox or draft an email. They can also suggest flights and determine the availability of two people. These individual abilities can be turned into systems that reliably communicate with others, coordinate, and update calendars, book reservations, or follow-up when situations change.
This gap is the basis of Catch.
Instead of asking users to create their own agent or configure workflows, the system integrates with the existing tools that executives use. Catch’s website lists support for Gmail, Outlook and Google Calendar as well as Slack, WhatsApp and iMessage. Users can communicate with their assistant in the same way they would with a real executive assistant.
After connecting, the system begins to understand preferences, such as meeting practices, calendar priorities, and travel needs. Catch claims it will then be able to act in a proactive manner. Re-scheduling a meeting is one example. A trip coming up can also prompt the assistant to make hotel reservations.
It is a major shift in the way AI products are designed. Catch tries to blend into existing software and communication channels, rather than creating a new interface for executives to maintain.
The freedom to act without boundaries is not autonomy
It isn’t whether AI models can comprehend a request that poses the biggest challenge for administrative agents. The problem is not whether an AI model can understand a request, but when to let software act independently and when to keep a human involved.
It appears that Catch is taking a multi-leveled approach.
In routine administrative activities, it is important to limit unnecessary approvals. Actions with significant financial or other consequences may need confirmation. The company’s travel workflowCatch is able to interpret travel briefs, make comparisons of flights, hotels and constraints and build an itinerary. But the user has to give their explicit consent before any booking can be made.
Similarly, Catch’s inbox product Can triage messages and create drafts based upon a previous user’s writing style. The current workflow also keeps users involved prior to sending messages.
This boundary could prove to be crucial. agentic AI adoption. The idea of full autonomy may sound appealing, but businesses tend to gradually adopt the agents, as long as their systems prove that they are capable of making reliable decisions in clearly defined areas.
Administrative tasks could prove to be an ideal proving ground for AI because they combine repetitive tasks and contextual decision making, allowing them to see if AI can go beyond automation.
Catch builds specialized AI beneath the assistant
Catch’s conversational interface is based on its AI architecture, rather than solely relying on an all-purpose foundation.
The system is described by the company as a narrowly focused “splinter agents”The agents are only allowed to access the data they need for their specific tasks. It is hoped that this will reduce the sensitive information exposed to agents during execution, while improving reliability and latency.
Catch states that it can use anonymized data to enhance its models by using reinforcement and supervised learning. Its security documentation Outside AI providers have a contractual obligation to not retain customer data, or use it for their models.
As an assistant to an executive may be able access travel plans, calendars and emails as well as sensitive company data, the importance of security increases when AI systems can do more than generate text. Catch is SOC 2 Type II certified, along with CASA Tier-2 certification and Google verification. It encrypts data stored and sent, while using fine-grained permissions to connect services.
AI and Venture Investing Founders
Two senior executives, who had different approaches to the problem of administration when founding their company.
The following are some of the most effective ways to increase your ROI: Nir Sabato previously spent several years investing at Entrée Capital, where he focused on areas including artificial intelligence and enterprise technology. Prior to entering venture capital, he worked in strategy positions at Fiverr and BLEND. Catch claims Sabato’s work with founders exposed how much valuable executive time is still consumed by routine coordination.
CTO Yoav RamonThe, on the other hand, has been working with machine learning for over a decade. He has worked in speech and language technologies at Hi Auto, and as the CTO for healthcare AI company Nym. Teams deployed large language model within regulated environments.
This combination is important because building a chatbot that can be used as an administrative assistant requires much more than just a good one. The system must understand users’ preferences, keep context throughout interactions, work with external software, and decide when to take action.
The next battlefield for AI may be the ordinary workplace
The fact that Catch has narrowed down the problem may be more significant than its entry into a market with burgeoning agentic AI.
AI firms are trying to develop agents that can perform dozens, if not hundreds of different types of tasks. Catch takes the opposite approach, focusing on a limited domain while trying to build a system that is reliable enough for users to stop doing the tasks.
This could be a key pattern in the future of AI for enterprises.
It is possible that the most useful commercial agents are not digital workers who can perform any task. It may be that they are highly specialized, able to perform a specific type of task with enough depth for the system to gain increasing trust.
A test in executive administration can be a very revealing one. While scheduling meetings, managing travel, and coordinating emails may seem mundane in comparison to coding and scientific research, these workflows demand AI that can interact with people, understand incomplete directions, remember preferences, and make real-world decisions.
The significance of systems like Catch, which can handle these workflows reliably, goes beyond the fact that executives will save a few minutes. The AI agent would be able to demonstrate how they are starting to go beyond just generating valuable information, and taking on the responsibility of getting things done.
New $5 Million Seed Round gives Catch Addition of capital to see if this transition is scaleable.

