TL;DR
Bars Juhasz, CEO of no-code automation firm WorkBeaver, argues that AI agents are being rolled out too quickly and that the demand side — office workers — should drive how automations are adopted. WorkBeaver offers menu-driven, no-code agent creation for non-technical users and plans a release that will let agents learn by observing user interactions.
What happened
Bars Juhasz, founder and CEO of no-code automation startup WorkBeaver, told The Register he worries companies are rushing AI adoption and that premature deployment could harm workers and the broader socio-economic landscape. After selling most of his shares in Undetectable.ai around August 2024, Juhasz focused on building tools that let everyday staff create agentic automations without code. WorkBeaver launched a private alpha in January and moved to open beta in September; the company reports nearly 4,000 users, mainly among small and medium enterprises. Its current interface uses a menu-driven prompt flow to translate task descriptions into agent behavior for tasks like form-filling, appointment setting and data entry. Juhasz said the next release, due imminently, will add an interaction-monitoring mode so agents can learn by watching mouse and keyboard activity. Technically, WorkBeaver was developed with Google Cloud Vertex AI, but the product is model-agnostic, encrypts data locally, stores only basic account info on its servers and uses allow-listing to limit agent access.
Why it matters
- Shifts control of automation toward workers could change who benefits from productivity gains and how layoffs are justified.
- Lowering the technical barrier may let non-technical staff use agents, but also raises training and education demands.
- Embedding agents with device control introduces new security and privacy risks that enterprises will have to manage.
- Faster, top-down AI mandates could prompt operational backtracking and create roles focused on cleaning up automation errors.
Key facts
- Bars Juhasz is CEO of WorkBeaver and previously co-founded Undetectable.ai; he sold most of his shares in that company around August 2024.
- WorkBeaver aims to let non-technical users create agentic automations without writing code or using APIs.
- Private alpha launched in January; open beta began in September; the company reports roughly 4,000 users, mostly from small and medium-sized businesses.
- Common use cases presented by the company include automated form-filling, reminder and appointment setting, email sending, data gathering, and data entry.
- The current interface is menu-driven prompting; the upcoming release will let an agent learn tasks by monitoring mouse and keyboard interactions during a collaborative session.
- WorkBeaver was built using Google Cloud Vertex AI but the system is model-agnostic and can be trained to work with other foundation models or run in private clouds.
- Juhasz says the company stores only an email address and account balance on its servers; most user data is kept locally, encrypted.
- WorkBeaver has a stated zero data retention agreement with Google Cloud and uses application and folder allow-listing to limit agent access.
What to watch next
- Whether the new interaction-monitoring release increases adoption among non-technical workers and reduces the need for prompting guidance.
- If the laptop preinstall deal WorkBeaver is pursuing for business machines in South America materializes and expands distribution.
- Potential security incidents or regulatory scrutiny tied to agents controlling user devices (not confirmed in the source).
Quick glossary
- Agentic automation (AI agent): Software that combines AI models with access to tools or device controls to perform tasks autonomously or semi-autonomously on behalf of a user.
- No-code: A development approach that allows users to create software or automations through graphical interfaces or guided flows rather than writing code.
- Model-agnostic: Design that allows a system to work with multiple underlying AI models rather than being tied to a single provider or architecture.
- Prompting: Providing text or structured input to an AI model to guide its responses or task behavior.
- Zero data retention: A policy or technical approach in which data is not stored persistently by a service provider and is kept only briefly in memory as needed for processing.
Reader FAQ
What is WorkBeaver trying to do?
WorkBeaver provides no-code, agentic automation tools aimed at letting non-technical workers create automations through menus and, soon, by having agents learn from observing user interactions.
Does WorkBeaver store user data centrally?
According to the company, only an email address and account balance are stored on its servers; other data is kept locally and encrypted, and it has a stated zero data retention arrangement with Google Cloud.
What security steps does the company take for agents that control devices?
WorkBeaver uses application and folder allow-listing and other unspecified measures to limit what agents can access.
Has WorkBeaver been commercially successful?
The CEO describes the company as a commercial success and reports nearly 4,000 users, primarily in small and medium enterprises.
Will WorkBeaver replace enterprise AI platforms?
not confirmed in the source
AI + ML 12 Workers should control the means of agentic production, suggests WorkBeaver boss What if AI vendors focused on the demand side? Thomas Claburn Sun 21 Dec 2025 // 12:21 UTC INTERVIEW…
Sources
- Workers should control the means of agentic production, suggests WorkBeaver boss
- WorkBeaver CEO: Workers should control agents, not …
- Workers should control the means of agentic production …
Related posts
- Oracle’s AI-powered MOS support portal overhaul sparks customer backlash
- Virgin Atlantic launches web-based AI travel agent built with OpenAI
- Microsoft aims to replace its entire C/C++ codebase with Rust by 2030