Produced by | Bullet Finance

Designed by | Qianqian

Edited by | Songwen

How can enterprises truly integrate AI into their business?

At the recently concluded ninth Huawei Connect conference, a technical leader from a traditional manufacturing company privately quizzed every AI R&D expert they could find on the scene. However, to their surprise, no two experts gave the same answer.

In fact, the chairman of their company had already recognized in 2023 that AI was the next opportunity for enterprise informatization development. By bringing him in as a high-level executive from a major internet company, the chairman hoped to leverage his experience in internet product development to drive the company to embrace AI as soon as possible.

However, once he entered the business realm, the technical leader was at a loss. The entire technical team had no idea how to introduce AI into production operations, let alone integrate it. In the end, they could only upgrade their customer service system using a domestic AI model, far from achieving the chairman's original goal of using AI to improve quality and efficiency. This was the reason he attended the Huawei Connect conference.

The problems faced by this company are common among Chinese enterprises in the AI era.

In February of this year, a HSBC survey of enterprises in eight major global markets revealed that 68% of surveyed enterprises in mainland China expressed interest in investing in AI technology, the highest percentage among all surveyed markets globally. Additionally, 86% of surveyed enterprises stated that the responsibilities of their financial functions were shifting, with a greater focus on providing real-time data and related analyses.

According to a survey by Eastmoney, 68% of surveyed enterprises in mainland China expressed an intention to invest in AI technology to improve operational efficiency. Meanwhile, over 88% of surveyed enterprises in mainland China are increasing their investment in digital talent.

However, despite many enterprises expressing strong interest in AI and eagerly hoping to reshape their business landscape through its introduction, and even elevate it to a strategic level, regrettably, only a handful of enterprises have successfully implemented this vision.

According to a Deloitte survey, the penetration rate of AI technology among Chinese enterprises is less than 23%, and the application rate of large models is even lower at less than 12%. Behind these figures lie the myriad challenges and difficulties faced by enterprises on the path to AI adoption.

1. To embrace AI, enterprises must first cultivate an AI-native mindset

AI, particularly large models, has been a buzzword in enterprise informatization over the past two years.

Even amidst uncertainty, many enterprises are investing heavily in AI R&D to board the high-speed train of AI development, prepared for long-term investment.

It cannot be denied, however, that while enterprises have significant demand for AI, there is also considerable waste. Many enterprises blindly invest in the hype without fully realizing AI's potential.

In the initial stages of industry development, many enterprises viewed AI merely as a "tool" akin to ERP systems, believing that developing a few applications and integrating them into business operations would propel corporate progress.

In reality, however, AI's capabilities extend far beyond this, and so do the challenges. According to an IDC research report, small and medium-sized enterprises (SMEs) primarily face limitations in data quality, computing power, and foundational algorithmic capabilities when developing AI strategies. Sixty percent of SMEs reported encountering data redundancy and management issues in AI projects, leading to resource waste, while 45% stated that data cleaning and preprocessing consumed significant time and resources with limited effectiveness.

Nonetheless, enterprises generally believe that AI technology can significantly enhance business performance, particularly in areas such as customer relationship management, business reengineering, marketing, supply chain management, and customer service.

This creates a paradox between the immense corporate demand for AI and its limited effectiveness in practical applications.

In response to the integration challenges between AI and corporate business, Zhang Ping'an, Huawei's Executive Director and CEO of Huawei Cloud, provided an answer during his speech at this year's Huawei Connect conference.

He believes that to resolve the issue of AI and business integration, enterprises must cultivate an AI-native mindset, rethinking and redesigning their business processes, IT architectures, and business innovations with AI technology and tools as core elements. This approach will fully unleash AI's potential, enhance efficiency, and innovate business models.

The secret to corporate leadership in the AI era lies in this mindset.

This way of thinking requires enterprises to reexamine their business processes, organizational structures, and resource allocations from a strategic perspective. For instance, in process optimization, enterprises can leverage AI algorithms to conduct in-depth analyses of massive amounts of data, identify bottlenecks and redundant steps in processes, and subsequently automate and intelligently transform these processes.

This involves more than simply introducing AI technology and tools into corporate operations; it necessitates a fundamental rethinking and redesign of all aspects of the enterprise, including process optimization, IT architecture reconstruction, and business model innovation, ushering in unprecedented changes.

2. To embrace AI, enterprises must prioritize three areas

By adopting an AI-native mindset, enterprises gain the potential to fully implement AI across all facets. Zhang Ping'an emphasizes the critical importance of focusing on three key areas: computing power, data, and models.

Firstly, computing power is crucial in integrating AI into enterprises, necessitating the construction of AI-native cloud infrastructure tailored to corporate needs. Computing power, akin to a heart, serves as an inexhaustible driving force for AI evolution. In the age of AGI, where large models are central, models form the cornerstone of collaboration between enterprises and AI. To match corporate AI capabilities with business scenarios, extensive data training is indispensable. This not only requires robust computing capabilities to support the processing and analysis of massive amounts of data but also ensures that these capabilities can be flexibly expanded to meet the evolving needs of corporate operations.

Secondly, data quality determines the effectiveness of AI models in corporate applications. As the "fuel" of AI, data influences model performance in corporate operations through training. Therefore, enterprises must establish a knowledge-centric data infrastructure to enable data to serve AI more effectively, transcending traditional business analytics. This entails delving deeply into, organizing, and processing data to transform it into valuable knowledge assets for business decision-making.

Lastly, enterprises should construct appropriate AI models tailored to specific business scenarios. This means avoiding the pitfall of building overly comprehensive models. Models are not necessarily better with larger sizes, nor can a single large model cater to all business scenarios. Instead, enterprises should build tailored AI models that are highly flexible and scalable, optimizing and upgrading them as business evolves.

In this regard, Huawei Cloud, with its expertise in both hardware and software, unveiled targeted solutions at this year's Huawei Connect conference.

In terms of computing power, Huawei Cloud has constructed a diversified, elastic, and efficient AI-native cloud infrastructure. In the era of intelligent technology, model parameters have soared from billions to trillions. To meet the explosive growth in AI computing power and ensure high reliability and efficiency, Huawei Cloud introduced the next-generation AI-native cloud infrastructure, CloudMatrix, featuring a distributed peer-to-peer fully interconnected architecture that evolves from individual computing power to matrix computing power.

According to Zhang Ping'an, CloudMatrix interconnects and pools resources such as CPUs, NPUs, DPUs, storage, and memory, evolving from individual computing power to matrix computing power to construct an AI-native cloud infrastructure where everything is poolable, peer-to-peer, and composable, providing customers with robust AI computing power.

Regarding data, to simplify data usage and enhance model training efficiency, Huawei Cloud has comprehensively upgraded its data governance pipeline, DataArts, offering customers an AI-oriented, knowledge-centric data foundation that significantly boosts resource utilization and data supply efficiency through integrated AI and big data engines, data development governance, knowledge services, and digital intelligence application enablement services.

Concerning models, the diversity of enterprise application scenarios necessitates the construction of multi-modal, multi-sized models on the cloud platform to achieve optimal scenario-model matching and satisfy corporate demands for the economy and professionalism of large models.

In June of this year, Huawei Cloud released the Pangu Large Model 5.0, featuring models with varying parameter specifications ranging from billions to trillions, including NLP, CV, multi-modal, predictive large models, scientific computing foundational large models, industry-specific large models, and out-of-the-box scenario models, catering to enterprises' needs across all business scenarios.

Addressing the opportunities and corporate demands of the AI era, Huawei Cloud continuously iterates its technology and products by building infrastructure, data foundations, and model series tailored to corporate needs, empowering enterprises to embrace AI faster and better.

3. Learn from Huawei on how to leverage AI

In fact, Huawei itself is actively integrating AI into its operations. In its own words, Huawei introduces AI into workflows through "human + AI" and "task + AI," significantly enhancing work efficiency.

At the conference, Tao Jingwen, Huawei's Director and President of Quality Process IT, stated that AI represents a transformation, and enterprises must deeply integrate AI with their processes, organizations, IT, data, and business scenarios to transform it into a service that genuinely adds value to their operations.

For instance, in contract processing scenarios, Huawei leverages digital objects, processes, and rules to achieve high-quality parallel processing of massive contracts. Furthermore, by extracting and comparing key elements of multilingual contracts using intelligent technology, Huawei has shortened risk processing time from two hours to just five minutes.

In R&D scenarios, Huawei has equipped over 110,000 R&D personnel with development assistants that utilize large models to automatically extract job context information, enabling line-level code continuation, function generation, code interpretation, and commenting functions. Through AI technology, Huawei adopts over seven million lines of AI-generated code annually.

In manufacturing scenarios, Huawei employs a multi-model "systems engineering" approach combined with decision-making and generative AI technologies, significantly enhancing overall productivity and shortening order delivery cycles by over 30%.

Drawing from its own practices, Huawei has summarized a "three-layer, five-stage, eight-step" methodology, comprising three layers of redefining intelligent business, AI development and delivery, and continuous operation of intelligent applications. The "five-stage, eight-step" approach guides enterprises in implementing AI step-by-step, starting from business scenarios and proceeding through business processes, organizations, corporate data, and AI applications.

Tao Jingwen emphasized that in implementing intelligence, Huawei adheres to a scenario-driven approach to address enterprises' vast, repetitive, and complex high-energy-consuming issues.

Judging from Huawei's experience, cultivating an AI-native mindset and prioritizing the three key areas are crucial for enterprises to genuinely gain development momentum and advantages in the AI era.

With its comprehensive hardware and software capabilities, Huawei Cloud leverages its rich experience to empower industries, holding significant importance in driving the integration of corporate business and AI.

4. Closing Thoughts

For corporate management teams, understanding the importance of new technologies and strategies for business in the AI era is undeniable. However, embracing an AI-native mindset can establish a holistic and systematic framework, ensuring that no aspect is overlooked and fostering a comprehensive perspective. Only by conducting the most meticulous analysis of one's own business and uncovering the underlying logic can one better serve the business. This is the logic that enterprises should adopt when embracing AI.

*Images in this article are sourced from Shutterstock and are based on the VRF agreement.