祝融说。

In Chapter 9, we focused on how to activate the value of the computing platform within the enterprise through application promotion and knowledge precipitation. However, in the frontier field of AI, where technology iterates rapidly, knowledge spans vast domains, and application scenarios run deep, any enterprise — even a giant like State Grid — if it attempts to "work behind closed doors" relying solely on its own strength, will inevitably encounter technology bottlenecks, innovation silos, and talent ceilings. For the platform's value to achieve exponential growth, organizational boundaries must be broken, moving from "owned by me" to "usable by me," upgrading from building a powerful "internal platform" to cultivating a thriving "industrial ecosystem."

Computing ecosystem construction marks the transition of computing platform construction from "tactical execution" to "strategic layout." It is the necessary path for the platform to leap from a mere technical support tool to an "industrial router" and "innovation reactor" that brings together wisdom, links resources, and incubates innovation. Its core idea is to proactively and strategically introduce external wisdom, technology, products, and services, combining them with internal needs, scenarios, and data to form a value co-creation network where "1+1 > 2."

This chapter will explore in depth the four key pillars of building a computing ecosystem, which together form a complete closed loop from "bringing in" to "going out," from "value creation" to "value measurement":

1. "Borrowing Brains" — Introducing Wisdom: How to build and manage a high-level external expert team to provide us with the most cutting-edge strategic insights and technical guidance? (10.1 Expert Think Tank: Building and Managing an External Expert Team)
2. "Linking" — Introducing Capability: How to build a multi-level, complementary computing ecosystem partner system, leveraging the best products, technologies, and services from the industry? (10.2 Partners in Progress: Building a Computing Ecosystem Partner System)
3. "Measuring" — Evaluating Value: How to establish a scientific evaluation system for ecosystem construction effectiveness, quantifying the real value brought by the ecosystem and guiding its continuous optimization? (10.3 Value Measurement: Ecosystem Construction Effectiveness Evaluation System)
4. "Iterating" — Verification and Promotion: How to use a pilot-first approach to verify the results of ecosystem cooperation on a small scale, and based on successful experience, carry out promotion and iteration, achieving large-scale implementation of ecosystem value? (10.4 Pilot First: Promotion and Iteration of Ecosystem Construction Results)

These four links together depict a grand ecosystem blueprint from "single-point cooperation" to "system construction," from "value input" to "value output."

Expert Think Tank: Building and Managing an External Expert Team

At the crest of the AI wave, the scarcest resource is not computing power but top-tier wisdom and insight. Building a high-level external expert think tank is like equipping our computing aircraft carrier with the world's top "navigators" and "radar systems," helping us see the route ahead, avoid potential reefs, and seize fleeting strategic opportunities.

Positioning and Value of the Expert Think Tank

• Positioning: The company's "super external brain" and "strategic advisory group" in the field of AI. It is an informal, flexible, top-tier intellectual resource pool bound by project cooperation.
• Core value:
  • Strategic foresight and technology prediction: Provide the most authoritative, independent third-party perspectives and recommendations for the formulation of the company's AI technology roadmap, major technology selection, and cutting-edge technology deployment (such as quantum computing, brain-inspired intelligence).
  • Key technical problem solving: Provide "point-to-point" expert diagnosis and solutions for "bottleneck" technical problems encountered in areas such as large model training, algorithm optimization, and platform architecture that are difficult for internal teams to solve.
  • High-end talent referral and cultivation: Help us attract and identify top talent in the industry through the influence of experts. At the same time, through exchanges and cooperation with experts, rapidly improve the vision and level of our internal core technical team.
  • Industry influence endorsement: The recognition and participation of top experts are themselves a powerful endorsement of our computing platform's technical strength and industry position, helping to enhance the brand image.

Construction Strategy of the Expert Think Tank: Systematic Search, Precise Recruitment

The construction of the expert think tank must never be "all in the basket is vegetables"; it must be targeted and built on demand.

• Map the expert capability spectrum:
  • We first identify the several core capabilities that the company's AI development currently needs and will need most in the future, forming an "expert capability demand map." For example:
    • Algorithm theory layer: Top theoretical scientists in the fields of natural language processing, computer vision, reinforcement learning, etc.
    • Platform architecture layer: Senior architects in the fields of large-scale distributed computing, high-performance networking, AI chip design, etc.
    • Industry application layer: Industry experts with rich AI implementation experience in the energy, finance, manufacturing, and other sectors.
    • Strategy and ethics layer: Thought leaders in AI strategy planning, data security, algorithm ethics, and other areas.
• Diversified expert sourcing channels:
  • Top universities and research institutes: Focus on professors and researchers from relevant departments and laboratories of top institutions such as Tsinghua University, Peking University, and the Chinese Academy of Sciences. They are the sources of cutting-edge theoretical innovation.
  • Leading technology companies: Chief scientists and senior technical experts from leading domestic and international AI companies (such as Huawei, Alibaba, Tencent, NVIDIA). They possess the richest experience in large-scale engineering practice.
  • Open source community leaders: Core developers (Committers/Maintainers) with significant contributions and influence in core open source communities such as PyTorch, Kubernetes, and TensorFlow.
  • Industry organizations and alliances: Connect with and link authoritative experts in the industry by participating in industry technical standard organizations and industry alliances.
• Flexible cooperation models:
  • Technical advisor/Strategic advisor: Sign an annual advisory agreement, regularly participating in our strategic seminars and technical review meetings.
  • Project-based cooperation: For specific technical problems or R&D projects, hire experts as project consultants for short-term, focused guidance.
  • Joint laboratory/joint research project: Establish a joint laboratory with the expert's lab or team, jointly apply for national or provincial/ministerial research projects, engaging in deep industry-academia-research binding.
  • Invited lectures and presentations: Invite experts to give internal technical lectures and training at the company.

Efficient Management and Operations of the Expert Think Tank

Inviting experts is only the first step. How to efficiently "utilize" the wisdom of experts and establish long-term, trusting cooperative relationships requires refined operations.

• Establish expert management files: For every expert in the database, establish detailed files including their research fields, technical expertise, cooperation history, and output results.
• Build normalized communication and service platforms:
  • Assign an expert liaison officer: Assign a dedicated internal liaison officer for each core expert, responsible for daily communication, demand coordination, and service assurance.
  • Regularly hold "Think Tank Roundtables": Once or twice a year, invite all think tank experts to gather around a major strategic or technical issue for focused "brainstorming."
  • Provide "two-way value": We not only need to "extract" wisdom from experts but also provide value to them. For example, open our unique, large-scale real business scenarios and anonymized data to them as "test fields" for their academic research; provide them with an industrialization implementation and verification platform for their research results.
• Establish results transformation and incentive mechanisms:
  • Clearly define intellectual property ownership: In cooperation agreements, clearly define the ownership and sharing mechanisms for intellectual property resulting from the cooperation.
  • Results evaluation and incentives: Provide market-competitive compensation or project rewards for significant expert suggestions that are adopted, or key technical problems solved with their help.
  • Honorary incentives: Grant experts honorary titles such as "State Grid Distinguished Expert" and conduct internal publicity to enhance their sense of honor.

Through the systematic construction and operation of the expert think tank, we have successfully installed a "super CPU" that brings together top global wisdom for the development of the computing platform, ensuring that we can always stand at the commanding heights of technology and strategy.

Partners in Progress: Building a Computing Ecosystem Partner System

If the expert think tank solves the introduction of "wisdom," then the partner system solves the introduction of "capability." In the complex industrial chain of AI, we need to work closely with various types of partners — hardware manufacturers, software companies, algorithm companies, solution providers — to complement each other's strengths and jointly provide end users with a complete, powerful, and easy-to-use computing service.

Partner System Layering and Positioning

We divide partners into three levels based on the type of capabilities they provide and the depth of cooperation:

Level 1: Infrastructure and Platform Partners (IaaS & PaaS Partners)

• Positioning: The "cornerstone" of the computing platform. They provide the hardware equipment and core platform software that form the foundation of our platform.
• Main partner types:
  • AI chip and server manufacturers: Such as NVIDIA, Huawei, Intel, etc.
  • High-performance network and storage manufacturers.
  • Cloud platform and container technology manufacturers: Such as Huawei Cloud, Alibaba Cloud, Rancher, etc.
• Cooperation model: Primarily product procurement, joint technology innovation, and deep performance tuning. We need to establish "strategic partner" relationships with them to obtain the latest technology, the most competitive pricing, and the highest level of technical support.

Level 2: Algorithm and Model Partners (Model & Algorithm Partners)

• Positioning: The "soul" of the computing platform. They provide high-quality, pre-trained foundation models or specialized models for specific domains.
• Main partner types:
  • General large model vendors: Such as Zhipu AI, Baichuan Intelligence, Moonshot AI, etc.
  • Industry model providers: AI algorithm companies focused on vertical domains such as energy, finance, and healthcare.
  • Open source model communities: Such as Hugging Face, ModelScope, etc.
• Cooperation model: Model introduction, joint fine-tuning, accuracy and performance evaluation. We will build a "Model Zoo," integrating and adapting excellent models from partners for internal users to "use out of the box."

Level 3: Application and Solution Partners (ISV & SI Partners)

• Positioning: The "amplifier" of computing value. They are the Independent Software Vendors (ISVs) and System Integrators (SIs) closest to business scenarios, capable of transforming our computing power into final industry solutions.
• Main partner types:
  • Power industry intelligent solution providers.
  • General AI application developers (such as intelligent customer service, RPA, data analysis tools, etc.).
  • Large consulting and integration companies.
• Cooperation model: Joint solution development, market co-expansion, project delivery. We will launch "Computing Platform Certified Solutions," conducting technical certification for solutions developed by partners on our platform, and jointly promote them to end users.

Full Lifecycle Partner Management

We have established a full lifecycle partner management process from "recruitment-empowerment-co-creation-incentive."

"Recruitment" — Strict Admission, Quality First

• Establish partner certification standards: Establish clear partner admission and rating standards from the dimensions of technical strength, industry reputation, service capability, and team size.
• Proactive recruitment and mutual selection: In addition to accepting applications, we also proactively seek and invite partners with outstanding advantages in specific fields to join.

"Empowerment" — Teach a Man to Fish, Grow Together

• Technical empowerment: Provide partners with free development and testing computing resources, detailed technical documentation and API interfaces, and professional technical training and certification.
• Market empowerment: Include partners and their solutions in our marketing activities, sharing sales leads and brand resources.

"Co-Creation" — Deep Binding, Joint Innovation

• Establish joint innovation laboratories: Set up joint laboratories with core strategic partners, jointly investing R&D resources to conduct forward-looking technology R&D and product incubation for major needs in the power industry.
• Create joint solutions: Deeply integrate and optimize our computing platform with partners' applications or models, creating "1+1 > 2" competitive joint solutions.

"Incentive" — Win-Win Cooperation, Sharing Results

• Establish a clear business model: Design reasonable revenue-sharing mechanisms (such as project commission, product resale discounts).
• Annual partner conference: Hold an annual partner conference to commend and reward the year's excellent partners, share successful cases, announce the latest cooperation policies, and build the ecosystem's "centripetal force."

Through the construction of this multi-layered, standardized, and win-win partner system, we have successfully "linked" the best capabilities in the industry to our platform, greatly expanding the breadth and depth of our service to internal users.

Value Measurement: Ecosystem Construction Effectiveness Evaluation System

Ecosystem construction invests a lot of resources and energy. How effective is it? What specific value does it bring to the company? These questions must be answered through a scientific, quantitative evaluation system. This is not only to "prove" the value of the ecosystem to management but also to "guide" its healthy development.

Core Principles of the Evaluation System

• Combination of results-oriented and process-oriented: Both measure the ultimate business results brought by the ecosystem (results) and assess the health and activity of the ecosystem itself (process).
• Combination of quantitative indicators and qualitative evaluation: Both have objective data indicators and subjective evaluations from internal users and external partners.
• Combination of short-term benefits and long-term value: Both look at short-term project returns and assess the ecosystem's improvement to the company's long-term technical capabilities, innovation capability, and brand influence.

Ecosystem Construction Effectiveness Evaluation Indicators (KPIs)

We divide the evaluation indicators into four dimensions: "Ecosystem Prosperity," "Technology Contribution," "Business Value," and "Brand Influence."

Ecosystem Prosperity (Measuring the Scale and Activity of the Ecosystem)

• Expert think tank:
  • Number and level of experts in the database.
  • Frequency of expert participation in activities (meetings, reviews, lectures).
  • Number of expert suggestions adopted.
• Partners:
  • Total number of certified partners and the number at each level.
  • Partner activity (such as API call frequency, number of joint solutions submitted).
  • Partner satisfaction (obtained through annual surveys).
• Ecosystem activities:
  • Number of ecosystem events (conferences, salons, trainings) held and participants.

Technology Contribution (Measuring the Ecosystem's Improvement to Internal Technical Capabilities)

• Number of key technical problems solved: Problems successfully solved using expert or partner technology that were previously "bottlenecks" for internal teams.
• Number of new technologies/new models introduced: Number of new technologies or models first introduced and successfully applied within the company through ecosystem cooperation.
• Joint R&D results: Number of jointly applied patents, published high-level papers, and contributed open source code.
• Platform capability improvement: Number of cases where the core performance indicators of our computing platform (such as throughput, latency) were improved through deep cooperation with partners.

Business Value (Measuring the Ecosystem's Direct Contribution to Business)

• Number of incubated innovative applications: Number of AI applications successfully incubated and launched through ecosystem cooperation (such as innovation contests, joint solutions).
• Direct/indirect economic benefits created: Total amount of quantifiable cost reduction, efficiency improvement, revenue increase, and expenditure savings brought by these applications.
• Shortened project delivery cycle: Average reduced project delivery time compared to full in-house development through the introduction of mature partner solutions.
• Internal user satisfaction: Satisfaction scores from internal users on models, tools, and solutions introduced through the ecosystem.

Brand Influence (Measuring the Ecosystem's Enhancement of the Company's Brand)

• Industry voice: The degree of leadership or participation in industry standard setting and industry alliances.
• Media exposure: Number of positive reports by mainstream and industry media on our ecosystem construction.
• Awards received: Number of provincial/ministerial-level or above awards won by ecosystem projects or joint solutions.
• Talent attraction: Number of high-end talent resumes attracted through ecosystem activities.

Operating Mechanism of the Evaluation System

• Data collection and reporting: We will establish an "Ecosystem Operations Dashboard," automatically collecting and aggregating the above quantitative indicators. Compile an "Computing Ecosystem Construction Effectiveness Evaluation Report" quarterly and annually for reporting to management.
• Regular review and adjustment: The operations team regularly (e.g., every six months) conducts in-depth analysis of the evaluation report, identifying highlights and shortcomings in ecosystem construction, and dynamically adjusting our ecosystem strategy and resource allocation direction accordingly.

Pilot First: Promotion and Iteration of Ecosystem Construction Results

The ultimate value realization of ecosystem cooperation results, whether introducing a new model or creating a joint solution, depends on their implementation in real business scenarios. We adopt a "pilot first, points driving the whole" strategy to steadily and efficiently promote the large-scale rollout of ecosystem results.

Selection and Implementation of Pilots

• Select "showrooms": We carefully select pilot units and pilot scenarios. A good pilot should have:
  • Urgent business needs: The business side has a strong willingness to use new technologies to solve real problems.
  • Representativeness: The success of this scenario has strong demonstration and replication value for other units.
  • Strong partner willingness: The pilot unit's leadership values it, and the technical team has a high degree of cooperation.
• Form a "Iron Triangle" team: Each pilot project forms a joint project team composed of "business experts + ecosystem partner experts + platform technical experts," ensuring close coordination between needs, technology, and platform.
• Agile iteration, small steps fast: Pilot projects adopt an agile development model, quickly developing a Minimum Viable Product (MVP), verifying it in real scenarios, and rapidly iterating and optimizing based on user feedback.

Evaluation and Promotion of Pilot Results

• Quantitatively evaluate pilot effectiveness: After the pilot ends, conduct a comprehensive review and evaluation, using data to prove the value of the pilot (e.g., efficiency improved by XX%, cost reduced by XX% compared to traditional methods).
• Create a "promotion toolkit": Standardize and productize the successful experience of the pilot into a "Promotion Toolkit," containing:
  • Standardized solution whitepaper.
  • Reusable deployment scripts and configuration templates.
  • Detailed user manual and training videos.
  • Clear ROI (Return on Investment) analysis report.
• Large-scale promotion:
  • Incorporate the pilot results into the "benchmark cases" mentioned in Chapter 9 for vigorous promotion.
  • Organize "Pilot Results Promotion Meetings," inviting relevant units from across the company to learn and exchange.
  • List mature solutions on the company's application marketplace for "one-click" deployment and use by various units.

Continuous Iteration of the Ecosystem

The ecosystem is not static; it requires continuous iteration and evolution based on technological development and changes in business needs.

• Establish a feedback closed loop: Establish smooth channels to collect user feedback from the pilot and promotion process, and input it to our ecosystem partners, jointly continuously optimizing and upgrading the solutions.
• Dynamic partner management: Based on the annual evaluation results, dynamically adjust partners — introduce new, more vibrant partners and eliminate those performing poorly or no longer meeting our development needs — maintaining the "metabolism" and competitiveness of the entire ecosystem.

Chapter Summary

In Chapter 10, we have painted the most grand and exciting blueprint for the value realization of the computing platform — building an open, win-win, and prosperous computing ecosystem. Through the construction of an expert think tank, we introduced the top-tier "wisdom"; through the construction of a partner system, we linked the most powerful "capabilities"; through the establishment of a value measurement system, we scientifically measure and guide the growth of the ecosystem; and finally, through pilot-first and promotion iteration, we steadily transform the results of the ecosystem into business value across the entire company.

At this point, the core content of Part 4, "Value Realization and Ecosystem Expansion," has been fully presented. We have moved from internal application promotion and knowledge precipitation to external ecosystem construction and value co-creation. This marks that our computing platform has transcended the scope of a mere technical infrastructure. It is evolving into a "tropical rainforest" that empowers innovation, a powerful "new engine" driving the company's digital and intelligent transformation. Here, internal needs meet external wisdom, mature technologies collide with emerging scenarios — a vibrant, endlessly thriving intelligent innovation ecosystem is growing vigorously on the fertile soil of State Grid.