The Path of Silicon-Carbon Co-governance for a Crypto Company — Cobo's Internal AI Transformation
- Core Viewpoint: This article elaborates on Cobo's practical journey from exploring the integration of AI and blockchain to focusing on an internal "Silicon-Carbon Co-governance" transformation. It emphasizes the deep integration of AI Agents into company management and business processes through top-down mandatory implementation, ultimately achieving a shift in the organizational work model from "people-driven processes" to "goal-driven systems."
- Key Elements:
- Initial exploration of the MCP App Store failed due to high AI barriers, insufficient MCP standardization leading to high integration costs, and poor implementation results.
- Transitioned to internal applications, building a "brain" based on Claude/Gemini (adopting zero data retention clauses), and self-developed a knowledge base and Agent framework containing strict data permissions and security isolation mechanisms.
- Using the mandatory implementation of the "OKR Agent" as a breakthrough point, systematizing goal setting, tracking, and review, allowing AI to directly impact performance and driving company-wide familiarity.
- Through evaluation and bonus mechanisms, forcing various departments to develop business Agents, ultimately launching over 100, changing the decision-making inertia of "thinking about hiring first when encountering a problem."
- Key prerequisites for successful transformation include the company having healthy cash flow to support long-term investment, strong top-down push from management, and mandatory use rather than encouragement.
- The spillover of internal transformation capabilities gave rise to the new product "Cobo WaaS Skill," aiming to upgrade wallet APIs into financial capability modules that can be directly invoked by AI Agents.
Since late 2024, Cobo has been exploring the integration of AI and blockchain alongside its core crypto custody and stablecoin payment businesses.
Our initial focus was on the potential for standardized skills brought by MCP. Theoretically, if skills are sufficiently standardized, AI could invoke capabilities like plugins, and blockchain would become the most natural financial infrastructure for AI.
Consequently, we internally incubated an MCP app store. However, it was quickly invalidated.
At that time, the barrier to entry for AI was still so high that only seasoned engineers could proficiently invoke it. MCP also lacked sufficient standardization; each integration was time-consuming and labor-intensive, with high costs and slow progress. The implementation results fell far short of expectations.
Nevertheless, the AI team was already established. It was expensive, difficult to recruit for, and impossible to disband easily.
So, we decided to pivot. Since we couldn't yet transform the client world, we would start by transforming ourselves.
First Problem: Security
As an asset custody company, Cobo's data, internal technical processes, and frameworks are extremely sensitive. Internally, we also have strict data tiering. However, without data and real business inputs, it's impossible to train the company's own Agent.
Our initial thought was local model deployment. But the reality is that the intelligence level of local models couldn't meet the requirements. They could run, but weren't user-friendly; they could answer, but weren't smart enough.
Ultimately, we chose Claude and Gemini as the primary models (ZDR—Zero Data Retention clauses can be applied for, achieving the highest level of isolation).
But large language models are just the underlying "brain" of the business. What's truly complex is data and permissions.
We later built a complete internal knowledge base and Agent framework.
Internal Knowledge Base + Cobo's Self-Developed Agent System
The knowledge base is responsible for tiering the company's internal data. It allocates readable scopes based on employee permissions.
When Agents invoke the knowledge base, they inherit the employee's permissions, rather than having an "omniscient view."
The details here include:
- How to isolate network environments
- How to restrict cross-tier data flow
- How to control log retention for auditability
- How to prevent sensitive information leakage
These aspects aren't glamorous, but they determine whether this initiative can run long-term. AI must not become a security vulnerability.
Problem After Architecture Was Built: No One Used It
Even today, the company still faces a practical problem: many front-line business units are dismissive of AI.
If we merely encourage usage, AI-driven workflow transformation won't happen.
We later realized we had to start with company management.
First Breakthrough: OKR Agent
Our first strongly promoted use case wasn't customer service or code writing.
It was OKR.
We used AI to break down company strategy, to help set OKRs, to track progress, and to review bottlenecks.
In other words, we gradually transformed company management from human-led governance to silicon-carbon co-governance. This process was extremely uncomfortable for employees.
Previously, goals could be written more vaguely, and processes could be explained more loosely. Now, weekly data is right there, leaving fewer excuses.
From that moment on, goals were no longer just discussions in meetings; they became continuous records in the system.
Strategy OKR Weekly Business Progress Tracking
But it was also starting with performance management that everyone truly became familiar with AI. Because if you don't participate, it directly impacts your compensation.
From Performance to Business: Comprehensive Agent-ification
Once OKRs were running, we began promoting the agent-ification of internal services. Through evaluations and bonuses, we mandated each department to establish Agents related to their own business.
Customer service built a customer service Agent. Legal built a contract assistance Agent. Sales built a CRM Agent.
Searching for the Most Passive-Aggressive Customer Agent
Ultimately, over 100 Agents were launched.
We cannot precisely quantify the results of "silicon-carbon co-governance."
But at least one change is clear:
Previously, when encountering a problem, the first reaction was, "Should we hire another person?" Now, the first reaction is, "Can we get the system involved first?"
This is precisely our understanding of silicon-carbon co-governance. It's not about AI replacing humans. It's about humans getting accustomed to working alongside systems.
Practical Lessons Learned from This Year's Journey
First, have healthy cash flow.
If a company's cash flow isn't healthy, this kind of transformation won't reach the finish line. AI is not a cost-saving tool; it's an upfront investment for long-term structural upgrades. We are grateful that Cobo's main business still generates healthy cash flow.
Second, it must be top-down.
Organizations don't change spontaneously. If management doesn't strongly push, this initiative will naturally fail.
As is well known, Cobo's founders are heavy AI enthusiasts. CTO Dr. Jiang started some AI research as a postdoc at CMU in the early 2000s.
Third, usage must be mandatory.
If it's only encouraged, AI will forever remain at writing emails. Real process-changing integration inevitably requires a degree of "compulsion."
Fourth, solve your own business problems first.
Many companies talk about AI + Web3. But if they haven't completed their own internal AI transformation, what they preach externally are just concepts.
Looking Back
We also cannot fully quantify this transformation. The company has begun shifting from "people-driven processes" to "goal-driven systems."
If "intelligent organizations" truly emerge in the future, they certainly won't evolve naturally. They will be forged through rounds of discomfort.
Because of the full team's participation, the company can also better understand the real needs in the AI era.
This is also a byproduct of our internal transformation.
Recently, we launched Cobo WaaS Skill. Cobo WaaS Skill is an integration and operational capability layer specifically designed for AI Coding Agents. Through structured knowledge, executable examples, and scenario orchestration, it enables Agents to accurately invoke WaaS APIs. We are upgrading wallet APIs into financial capability modules that can be directly invoked by AI Agents. The development cycle is shortened from weeks to the duration of a conversation.
This isn't the result of a single product inspiration. It's the natural spillover of capabilities after our internal round of silicon-carbon co-governance.
We are still exploring.
But at least, today's Cobo is no longer the company it was in 2024.
