By Dr. Yuan Xiaohui

Vice President, Tencent Research Institute | July 24, 2026

Keynote Address delivered at the World Artificial Intelligence Conference (WAIC 2026) Tencent Forum, adapting the research report "From Super-Individuals to Super-Teams."

Over the past two to three years, the proliferation of artificial intelligence has expanded individual worker capability, establishing the rise of the "Super-Individual." Single operators now direct software agent fleets across market research, document drafting, and software engineering, yielding sharp multiples in personal throughput.

Yet an enterprise paradox has emerged: while discrete functional nodes register marked gains, enterprise-level performance metrics show little acceleration.

By SOLOMOAT Editorial Team

Corporate executives are reassessing the unit economics of foundation model token expenditures. The fundamental question confronting leadership is whether individual worker acceleration translates into aggregate operational efficiency and top-line revenue expansion.

The Mechanism of Bottleneck Migration: Amdahl’s Law in Enterprise Systems

The failure of individual efficiency to scale linearly across an entire enterprise is explained by Amdahl’s Law, which defines the theoretical maximum speedup of a system as strictly constrained by its non-parallelizable, serial components:

$$S_{\text{latency}} = \frac{1}{(1 - p) + \frac{p}{s}}$$

Where:

$p$ represents the proportion of the workflow that can be parallelized or automated by AI (execution tasks).

$(1 - p)$ represents the strictly serial, un-automatable portion (judgment, cross-functional alignment, strategic review, final physical delivery, and enterprise sales).

$s$ represents the acceleration coefficient applied to the parallelizable component.

Assuming execution historically consumed 80% of project cycles while qualitative review, verification, and sales accounted for 20%, deploying a multi-agent harness to accelerate execution tenfold reduces that 80% block down to 8%.

However, because the 20% serial review baseline remains fixed, total project duration contracts from 100% to 28%. The aggregate speedup of the organization is strictly capped at ~3.57x, despite individual contributors reporting a 10x acceleration in task turnaround.

When accounting for real-world institutional friction—budget approvals, conflicting priorities, compliance reviews, and governance meetings—the realized speedup contracts further. The organizational constraint has not vanished; it has migrated to the serial checkpoints of the firm.

Case Analysis: Enterprise-Wide Restructuring at Yiran

To resolve this bottleneck migration, organizations must treat artificial intelligence as a structural transformation rather than an ad-hoc desktop utility.

Shenzhen Yiran, a cross-border e-commerce enterprise focused on the South Korean retail market with 150 employees, deployed a dedicated 10-person task force to deconstruct every functional role across the corporate hierarchy.

The enterprise audited 1,400 discrete operating tasks, segmenting daily and weekly cadences to transition routine execution to software agents while centralizing human judgment at critical decision gates.

The firm recorded an aggregate organizational efficiency gain of 76% across its operating units:

The Super-Team: Human Judgment and Agent Orchestration

Scaling from individual leverage to a high-performing enterprise requires moving beyond an uncoordinated collection of solo operators.

A "Super-Team" balances the complementary capabilities of human operators and synthetic agents:

Enterprise competitiveness is shifting from manual execution to systematic discovery. As generative models commoditize routine production, competitive advantage concentrates in problem formulation, cross-disciplinary synthesis, and rapid empirical validation.

True institutional innovation rarely emerges from isolated talent; it is generated when disparate domain frameworks, field observations, and strategic perspectives intersect within a shared operational environment.

Modern Enterprise Infrastructure: From Static Files to Dynamic Agents

Over the past four decades, desktop software suites centered work on static files—documents, spreadsheets, and presentation decks.

The agent-driven era requires a new operating infrastructure built to manage human-agent collaboration, dynamic context routing, centralized memory, and self-improving workflows.

Platforms like WorkBuddy illustrate this infrastructure transition. Their utility lies not merely in accelerating personal tasks, but in serving as a shared operating environment linking human judgment, autonomous agents, and institutional memory:

The Three-Phase Organizational Transformation

The transition to an agent-integrated enterprise unfolds across three distinct phases:

Market leadership will not belong to the enterprise with the largest inventory of software tools, nor to the firm with the greatest collection of uncoordinated solo operators.

Enduring advantage belongs to organizations that build unified operating environments where human judgment and autonomous software agents continuously learn, evaluate, and build together.

Frequently Asked Questions

Why can AI raise personal productivity without improving company performance?

Enterprise outcomes are constrained by serial work such as judgment, cross-functional alignment, review, and customer delivery. Automating parallel tasks does not remove these bottlenecks.

What should leaders measure after deploying AI?

Measure end-to-end cycle time and constraint movement, not only the output of the individual function that adopted a tool.