Eigen RadarAI
Analysis

Genspark demonstrates agents working from shared company context

Genspark presented its enterprise-agent architecture at a Tokyo roundtable, alongside Japanese customers’ accounts of earlier deployments. A shared context layer connects workplace information across tools, while smaller and advisory models form part of its spending controls. A live request to fix a translation-tool shortcut gave the presentation a concrete software task; business results remain attributed to the companies.

Artificial Intelligence··Morning
An audience watches a presenter and computer user from behind.

Tokyo presentation centers on shared enterprise context

Genspark held a media roundtable in Tokyo on October 14, presenting its enterprise-agent architecture alongside Japanese customers’ experiences. Its technology leadership outlined the company’s current approach to agents working across organizational tools.[1], [2]

Kaihua Zhu, Genspark’s co-founder and technology chief, described Second Brain, a shared context layer combining Outlook mail, Notion content, meeting notes and project history. Genspark says the layer lets agents working through different tools draw on common organizational information. It also described US inference providers and contracts specifying that data is not retained.[1]

A shortcut problem became a live agent task

During the presentation, a reporter raised a keyboard-shortcut issue in Speakly, Genspark’s real-time translation tool. The Windows shortcut used the right Alt key, which some recent computers replace with a Copilot key. A representative demonstrated asking a product-manager agent to correct it. For inference spending, Genspark described routing simple requests to less expensive models and pairing small execution models with stronger advisory models.[1]

Customers describe recruitment and property-management work

Leopalace21 described an earlier deployment for property managers, supported by training, shared templates and rules governing data entry. Nisso described recruitment agents comparing local wages, rewriting job advertisements and matching candidates. These deployments preceded the roundtable. Nisso emphasized retaining Microsoft 365 access permissions, while Genspark outlined controls over enterprise users’ access to agents. Customer results remain the companies’ accounts of their use.[1]

References

  1. News sourceAI WatchGenspark demonstrates enterprise agents and customer workflows in Tokyo↩1↩2↩3↩4
  2. News sourceDream SeedGenspark presents enterprise-agent architecture at Tokyo roundtable↩