Make moves faster and more precise.

Editor: Gu Dao

Writer: Li Li

Source: Shoucai – Shoutiao Finance Research Institute

The winds of the times never cease, only their direction changes. Those adept at riding these winds and seizing first-mover advantages often reap the earliest rewards of cyclical booms.

In today's burgeoning yet fiercely competitive AI industry, the race for Agent supremacy increasingly tests comprehensive operational capabilities. The market is not short of models or application gateways; what's scarce is a systemic ecosystem that stably integrates model capabilities into business workflows.

On July 10, Baidu hosted 'Baidu AI Day – Baidu Dazi Special Event,' showcasing intelligent routing, multi-terminal shared memory, Browser Use, professional Skill ecosystems, and the first fully automated Skill deployment process, while announcing the enterprise version and the Dazi Alliance Plan.

In today's AI landscape, the gap between top-tier models has narrowed significantly. The true differentiator lies in system upgrades beyond the model itself—this is the significance of Baidu Dazi's latest overhaul.

Having secured an early ticket to the next round of competition, what has Baidu done right? Can this lead overcome past lags and achieve a overtaking on a bend (overtaking on a curve)? Amidst rapid advancement, are there hidden vulnerabilities?

1

Practicality Over Flashiness: Four Key Strengths

First, it's certain that 2026 marks explosive growth for intelligent agents, dubbed the 'Year of AI Agent Applications.'

Yet, implementation bottlenecks persist: scattered answers fail to handle complete project tasks; fragmented content struggles to span full workflows; individual efficiency gains don't translate to team-wide improvements.

This 'implementation gap' pushes practitioners to a critical inflection point. The focus shifts from smooth dialogues or stunning outputs to closing work loops end-to-end, turning completed tasks into reusable knowledge. In short, the industry pivot from 'showmanship' to 'practicality.'

Against this backdrop, Baidu Dazi undergoes a transformative update: a major personal version upgrade, an enterprise version debut, and the full launch of the Dazi Alliance. Aiming to resolve real user pain points, it enhances capabilities across four dimensions—'brain, hands, skills, professional suites'—transforming from an 'occasional novelty tool' into an indispensable daily work companion and life assistant.

As Shen Dou, Baidu's Executive Vice President and President of Baidu Intelligent Cloud, stated on-site, 'In the future, 90% of work may involve intelligent agents deeply participating and assisting. Baidu Dazi's mission is to dive into office scenarios and deliver real results.'

Specifically, the upgraded Baidu Dazi offers these strengths:

1. **Enhanced Proactivity**: Intelligent agents are no longer passive chatbots but can autonomously assess task types and select optimal execution paths. For example, in the personal version, when asked to compile ICML 2026 award paper information, it automatically browses the web, retrieves scattered abstracts, core arguments, and award rationales, and organizes them into structured documents—all without tool-switching.

2. **Multi-Terminal Synergy**: Shared memory and system calls break down barriers across devices and apps. A project idea discussed on a phone seamlessly continues on a computer, with full context synchronization. It can also directly operate browsers for information filtering and form filling, eliminating manual copying and pasting for true cross-device, cross-system continuity.

3. **Ecosystem Scalability**: Baidu Dazi establishes a fully automated Skill deployment pipeline, managing the entire chain—from development standards and skill submission to security evaluation, effect verification, and operational launch—on a unified platform. Its capabilities now span search research, life services, content creation, office collaboration, and more. As third-party integrations grow, its task-handling scope will expand further.

4. **Vertical Productivity**: The 'Professional Suites' elevate Baidu Dazi from a general assistant to a vertical productivity tool. For instance, it strengthens PPT generation for high-usage office scenarios, optimizing template libraries and Exclusive illustration model (dedicated image models) for frequent tasks like project reports and research briefings. Combined with intelligent layout and auto-chart generation, outputs nearly meet deliverable standards.

Overall, Baidu Dazi's upgrade elevates its combat effectiveness significantly, resembling a 'six-sided warrior.' Its evolution aligns with AI Agent industry's core value logic: prioritizing user-centric practicality over flashy demos or parameter races, translating Baidu's accumulated chip, cloud, large model, and engineering strengths into real productivity.

While the industry debates Agent futures, Baidu Dazi has already become a tangible work secretary and life assistant for everyone, accelerating large-scale Agent adoption.

2

Self-Developed, Self-Controlled: Deepening the Full-Stack Moat

Beneath these functional innovations and experience optimizations lies the self-developed, self-controlled full-stack system.

Take intelligent routing: users perceive Baidu Dazi as an experienced project manager, instantly assigning tasks to the right model or tool without inefficient detours. This seemingly simple feat represents a significant industry leap.

Industry analyst Sun Yewen notes that such scheduling is no easy feat—it requires real-time bidirectional communication and dynamic collaboration between models and clouds. Even leaders like OpenAI struggle with full-stack synergy, lacking access to chip and cloud foundations. Without control over compute supply rhythms and costs, product iteration and large-scale deployment face significant hurdles during industry-wide compute crunches.

Baidu, however, has connect through (connected) this collaborative logic, building a complete four-layer full-stack system—'chip, cloud, model, agent'—with interlocking components. Progress in one layer amplifies others, while weaknesses are compensated across layers.

At the base, 'chips' anchor costs and supply stability. Fully autonomous control eliminates reliance on overseas chip vendors, enabling self-determined inference costs and compute expansion rhythms.

Above, 'cloud' is the core carrier for large-scale Agent operations. The upgraded 'New Full-Stack AI Cloud' adapts bidirectionally for the Agent era: Agent Infra maximizes intelligence per Token, reducing costs and speeding task completion; AI Infra enhances model training/inference, high-performance clusters, and green data centers, boosting compute cost-efficiency. The cloud layer evolves from a passive 'container' to an active 'engine' for Agent operations.

Next, 'models' form the capability core. Fully self-developed large models free Baidu from waiting for third-party updates or external pricing strategies, enabling direct customization of Agent execution logic at the model layer.

At the top, 'agents'—led by Baidu Dazi—comprise a product matrix for general office, video generation (Baidu Yijing), low-threshold agent development (Miaoda), and enterprise strategic decision-making (Famou 2.0). Each product validates delivery capabilities in its scenario, translating technological strengths from the lower three layers into tangible user value.

This forms a virtuous cycle: chip acceleration reduces inference costs, model iteration broadens application scenarios, cloud expansion supports larger user bases, and agent experience improvements drive further upgrades across chips, clouds, and models.

Baidu's true achievement isn't merely assembling chip, cloud, model, and agent components but making them 'rotate' synergistically in real-world scenarios, forming a collaborative flywheel. By lowering barriers and costs for large-scale Agent adoption while enhancing user experience, it distances itself from peers, quietly fortifying Baidu Dazi's core barriers.

3

Scenario Revolution: Topping Monthly Access Growth Charts

Superior products never lack markets or growth potential.

According to Baidu Intelligent Cloud, Baidu Dazi's daily queries surged 20-fold within three months of its March 22, 2026, launch. Ranked #1 on the AI Product Chart's 'AI Lobster Growth List' with 114.72% monthly access growth on May 8, it also placed third overall with 1.16 million monthly visits—the sole product among the top ten growth leaders to achieve triple-digit growth. Such explosive momentum is rare across the AI sector.

With robust functionality and user experience, Baidu Dazi was selected as a 'Museum Piece' at the 2026 World Artificial Intelligence Conference (WAIC), the only general intelligent agent product among the ten honored.

A McKinsey November 2025 report suggests that ~57% of current U.S. work hours could theoretically be automated by intelligent agents. At Create 2026, Baidu founder Robin Li proposed DAA (Daily Active Agents) as a new metric, shifting focus from Token consumption to measuring how many agents actually complete real work and deliver tangible results.

Viewed through this lens, Baidu Dazi's upgrade carries deeper significance: the personal version becomes an indispensable daily productivity partner, the enterprise version amplifies organizational capabilities, and the Dazi Alliance extends these mature capabilities to partners across industries. These three layers advance toward a unified goal—evolving agents from occasional helpers into holistic productivity partners.

As Shen Dou shared at AIDAY, after integrating with Baidu's internal business systems, Baidu Dazi has transformed from an individual assistant to a high-frequency team collaboration tool.

In short, with a validated model, Baidu Dazi faces a virtually limitless growth market. As user scale expands and scenario data accumulates, the full-stack collaborative flywheel will spin faster, accelerating the release of 'chip-cloud-model-agent' system advantages. This scenario revolution holds immense growth potential.

Bai Wenxi, Chief Economist of the China Enterprise Capital Union, told Shoucai that Baidu Dazi's upgrade reflects Baidu AI's strategic shift from 'model muscle-flexing' to 'Agent consolidation,' with AIDAY and WAIC Museum Piece serving as key pillars of Baidu's 2026 AI layout (AI strategy). Overall, Wenxin is the 'foundation,' while Dazi is the 'facade + cashier'—Baidu AI aims to transition from 'selling model APIs' to 'selling Agent execution + organizational subscriptions,' a critical driver for valuation reassessment.

4

Five Internal and External Pressures: Ambitions of Overtaking on a Curve

Yet, complacency is unwarranted. A large market breeds fierce competition, and despite its lead, Baidu Dazi must tread carefully. Several challenges demand prompt attention and continuous evolution.

First, **commercialization hurdles**. While Baidu Dazi disclosed a 20-fold increase in daily queries, core metrics like active users, paid conversion rates, and retention rates remain undisclosed. The enterprise version, in particular, must extend compute revenue to subscription models, requiring alignment of product iteration, cloud sales delivery, and security controls to avoid prolonged trial phases and actual revenue conversion.

Second, **enterprise adaptability challenges**. Directly migrating personal-side execution capabilities to enterprise scenarios may expand business scope, but inadequate permission isolation, audit trails, or error recovery could lead to AI overstepping boundaries or untraceable operations, disrupting normal business. Enterprise services demand higher reliability and stability than personal use.

Third, **service stability risks**. For instance, on July 14, Baidu Dazi's mobile app experienced lag, with some users reporting disrupted usage. Baidu attributed this to a sudden surge in active users and concurrent task submissions, causing short-term service capacity strain.

While this reflects market enthusiasm, Baidu initiated emergency capacity expansion and database optimizations. However, the incident underscores the need for vigilance. If enterprise-scale deployments later encounter peak-load capacity mismatches, user trust and business continuity could suffer.

Fourth, **ecosystem collaboration risks**. The Dazi Alliance spans channel partners, business opportunity (business opportunities), and delivery partners, fostering ecosystem synergy. However, varying service capabilities and compliance standards among partners may lead to inconsistent delivery quality or unclear accountability, undermining collaboration effectiveness and brand reputation.

Finally, **intensifying market competition**. The track ( track , 'track') features not only Tencent WorkBuddy, Alibaba AgentTeams, and ByteDance Coze but also vertical players like Mingtu Technology's WorkBrain, creating a crowded landscape.

Among them, Tencent's WorkBuddy, leveraging the strong connections within the WeChat ecosystem and WeCom, focuses on developer ecosystem incubation and has established the nation's first OPC cooperative community in Chengdu. Alibaba's AgentTeams, on the other hand, deeply integrates e-commerce and enterprise service resources within the Alibaba ecosystem, boasting mature implementation experiences in e-commerce intelligent customer service and supply chain collaboration scenarios. Amidst fierce competition, Baidu Dazi does not lead in every aspect and has its lagging weaknesses, with competitive pressures that cannot be underestimated.

In response, Bo Wenxi pointed out that Baidu Dazi represents Baidu AI's most 'product-oriented' move in this round, with greater strategic value than ERNIE's iterations. However, it has shortcomings in long-task stability, cross-application permissions, B-end sales, and C-end brand perception, necessitating improvements in Skill ecosystem openness, end-to-cloud collaboration, and scenario completeness. By 2026, the true test will be whether enterprise version sign-ups and DAA can truly take off.

Indeed, Baidu Dazi's breakthrough, from 'showcasing skills' to 'getting things done,' essentially reflects the systematized advancement of China's AI industry. As the flywheel of full-stack collaboration begins to turn and the demands of real-world scenarios continue to accumulate, the most significant chapter of this long-distance race in the Agent sector is just beginning.

To continue leading the pack, Baidu Dazi still has much groundwork to lay and fundamental skills to hone. Its future growth potential hinges on the successful implementation of its business, the full release of ecosystem synergies, and the sustained realization of commercial value.

Nevertheless, a good start is half the battle won. By steadily embedding technology into real business processes, Baidu Dazi's recent upgrade undoubtedly dispels doubts about the overall lag in AI transformation and secures an early advantage in the new cycle.

However, as the saying goes, 'the last leg of a journey marks the halfway point.' With the market evolving rapidly and competitors not standing still, can Baidu truly achieve a overtaking on a bend (overtake on a curve) and sustain its lead? Baidu has time to celebrate but no time to rest. How many more evolutionary surprises await remains to be seen!

This article is original content from Shoucai.

If you wish to reprint it, please leave a message.