Fabr drives enterprises to evolve from "using AI tools" to "AI-native", and promotes the implementation of enterprise-level AI-native collaboration platforms.
When many enterprises are still adding AI tools and focusing on single-point efficiency, the team led by Fabr's founder Fang Bo has noticed that the increase in the number of tools and Agents will not bring about improvement in organizational efficiency, a feeling that many companies undergoing AI transformation can deeply relate to. If the demand background, historical decisions and verification experiences are still scattered across conversations, documents and different systems, the speedup of a single task can hardly be converted into the efficiency improvement of the enterprise or organization. Fabr hopes to continuously precipitate the information of these work processes, and make them effective context for subsequent personnel and Agents.
Starting from the R&D pain points of the team
The origin of the project came from the team's own R&D practice. Previously, the Fabr team had a large technical team and introduced AI Coding tools relatively early, but soon found an abnormal phenomenon: the speed of individual developers writing code became faster, but the overall project delivery efficiency did not increase proportionally.
From the proposal to the launch of a requirement, the intermediate process also involves requirement clarification, design decision-making, task decomposition, code implementation, testing and acceptance. Relevant information is scattered among different personnel, tools and AI conversations, and the team still needs to repeatedly explain the background and confirm the results. The more people and Agents participate, the more obvious the collaboration cost brought by context transmission becomes.
Fabr was originally a system developed to solve this internal problem. The team hopes to let personnel and Agents share the same background and collaborate around the same task, reduce repeated communication and handover costs, and improve the team's delivery efficiency. In this way, AI not only helps individuals write code faster, but also participates in the complete R&D collaboration process, continuously mastering the full context of the same task from requirement background, decisions to code, testing and acceptance. As the effect of internal use gradually emerged, the team further realized that the facts formed in the R&D process and their correlations are the foundation on which personnel and Agents can continue to work in relay. Fabr has thus gradually developed from an internal R&D collaboration system to an enterprise-level AI-native R&D collaboration platform based on FDM.
In the practice of Fabr's own team, the size of the R&D team was adjusted from about 300 people to 67 people, the meeting time was reduced by about 60%, and the time for new employees to get familiar with the project and work independently was shortened by more than 50%. After veteran employees leave, the verified project experience can still be accessed and reused by the team. The team believes that these changes occur along with the adjustment of FDM and the collaboration mode between people and Agents: part of the background that originally required meetings for synchronization can be accessed along the task, and the division of labor between personnel and Agents as well as result verification are also clearer.
Fabr is currently in the field of enterprise-level AI applications and enterprise services. It first targets the R&D teams of medium and large enterprises that plan to introduce AI Coding and Agents, and plans to extend to more high-intelligence work scenarios that rely on professional knowledge, complex collaboration and continuous judgment in the future.
The core team of Fabr has experience spanning corporate strategy, investment, basic software R&D and fintech. CEO Kenneth has been engaged in work related to corporate strategy, strategic investment and digital transformation, with work experience in Huawei, Bain, Accenture and TAL. CTO Ken once participated in the early core R&D of Google Android and the design and development of components such as JIT, and has experience in quantitative technology, entrepreneurship and large-scale R&D team management. The team hopes to combine technical R&D capabilities with practical enterprise implementation experience to solve the collaboration and management problems after Agents are introduced into organizations.
Fabr started productization in early 2026 and was officially launched to the public in July. It is currently in the stage of commercial verification and large-scale expansion.
During the project advancement process, the team found that the biggest challenge is no longer product R&D and product promotion, but the rapidly growing demand of enterprises for AI-native consulting and implementation. Many enterprises initially believed that introducing models, Agents or AI tools means completing the AI-native transformation; after entering the actual implementation stage, they found that the division of labor and collaboration between people and Agents, the connection of task processes, the continuous update of knowledge, and the verification of results all require corresponding methods and organizational mechanisms.
At present, Fabr has integrated relevant practical methods and configuration schemes applicable to different industries and scenarios into the product supporting system. The next problem to be solved is the scaling of delivery capabilities: the team is adopting methods such as customer stratification, unified implementation and acceptance standards, and cultivating partners and internal implementation personnel of customers to reduce the dependence on core experts accompanying each customer one by one, while ensuring the implementation quality of different customers.
Make the real work process become the FDM enterprise full-volume model
The core technology of Fabr is FDM (Fabric Domain Memory), which is currently implemented in the form of a full-data model of enterprise R&D. FDM is not a new general large model, but receives tasks, human-machine conversations, requirement changes, code changes, tool calls, testing and delivery results generated in the real work of the enterprise, and retains the identity, time and source of information. On this basis, FDM identifies the dependency, verification, conflict and substitution relationships between information through semantic annotation, partial order relationship and deterministic verification, continuously updates the currently valid project status, so that the subsequent person or Agent taking over the task can know what to follow at present, why the adjustment is made, and whether the result has been verified. When new personnel or Agents enter the task, the system can provide relevant, effective and source-based background information according to the task and permissions; new analysis, code, testing and human feedback will continue to be input into FDM.
Compared with the knowledge base dominated by document storage and retrieval, FDM presents the relationships between work processes and information, so that the background and judgments formed by one task can be continuously used by subsequent work.
FDM also does not bind enterprise experience to a certain model or Agent. The model can be upgraded or replaced, and personnel will also change, but the formed business background, engineering rules, historical decisions and verification experience of the enterprise can still be used. New personnel and Agents do not need to re-understand the entire project, but can inherit the judgments previously formed by the organization.
In addition to context supply, Fabr also associates tasks, Token consumption, the execution process of personnel and Agents, verification results and final delivery. Managers can not only view how many resources AI has used, but also further understand which task the relevant investment occurs in, what results are formed, and whether the results have been verified.
Software R&D is the first scenario where FDM is applied, because the relationships among requirements, decisions, code, testing and delivery in R&D tasks are clear, and the results are easier to verify. However, the application of the enterprise full-volume model is not limited to R&D.
The long-term direction of FDM is to extend these capabilities to more high-intelligence work. Scenarios such as investment research, consulting, product design, financial analysis, and sales tracking also need to retain the facts, judgments and verification processes when conclusions are formed. AI Coding is only the first scenario verified by Fabr. FDM aims to serve more work completed jointly by people and Agents.
Starting from R&D, expand more enterprise collaboration scenarios
At present, less than 3 months after Fabr's launch, Fabr has carried out cooperation, pilot and scheme promotion in more than 50 large and medium-sized enterprises, covering financial services, software and digital services, advanced manufacturing, medical technology, educational technology and other fields. The practical issues that customers pay attention to include: 1. Reduce context breakage during project handover; 2. Enable different teams to collaborate around the same set of effective background; 3. Map AI input to specific tasks and delivery results.
The initial feedback from customers focuses on project handover and cross-team collaboration. The taking-over members can continue to work along the task link, reducing the time to sort out group chats and documents from scratch. Enterprise experience will not be lost with personnel turnover, and new employees can also understand the historical background of enterprise projects faster; different departments collaborate around the same task and its context, which can also reduce repeated confirmation and greatly cut down meeting time. Fabr also associates Token consumption with tasks and delivery results, helping enterprises see clearly where AI resource investment occurs. Many enterprises spend high AI costs on AI transformation but cannot evaluate the team's use effect. Specific customer effects and named cases will be further disclosed after obtaining customer authorization.
In terms of business model, Fabr does not adopt the traditional SaaS seat-based subscription charging method. The platform does not charge seat fees, but collects platform service fees based on the actual Token usage. When personnel enter the platform for collaboration, corresponding platform service fees are generated only when the model is actually called and Tokens are consumed. This model lowers the threshold for enterprises to use AI on a large scale. The more people participate, the better the collaborative work capability of Fabr can be demonstrated. Of course, as the number of teams, Agents and tasks accessed by the enterprise increases, the actual usage of the platform will naturally expand, generating continuous revenue.
The Fabr team believes that AI-native is not just deploying more models and Agents, but allowing enterprises to continuously retain the facts, judgments and experiences formed in real work, forming a set of continuously updated full-volume models. Every collaboration between people and Agents, every decision and every verification result will become the data assets and experience assets of the enterprise, and provide a basis for the next execution.
On the one hand, Fabr will continue to promote existing enterprise cooperation to enter a deeper usage stage, forming verified implementation cases; on the other hand, after the software R&D scenario is fully verified, it plans to gradually extend FDM to more high-intelligence collaboration scenarios. AI Coding is the first verified entry point completed by Fabr. Its long-term question to answer is whether enterprises can move from scattered use of AI tools to an AI-native working mode that can be jointly inherited and continuously updated by personnel and Agents.