Corporate infrastructure in 2030 changed by Gemini Spark and GPT-5.3-Codex-Spark | Cloudization of labor and development

[Summary]

In the AI market in 2026, two "Sparks" are attracting a lot of attention.

One is Gemini Spark, a 24-hour AI agent for individuals that Google announced in conjunction with Google I/O 2026.

The other is GPT-5.3-Codex-Spark, an ultra-high-speed model for real-time coding that OpenAI announced in February 2026.

Although they share the same name, Spark, their roles are different.

• Gemini Spark is a business agent that works across Gmail, Docs, Slides, etc.
• GPT-5.3-Codex-Spark is a development agent that supports code creation and modification with low latency

What they have in common is that some corporate activities will be shifted from ``humans performing tasks on a case-by-case basis'' to ``a digital workforce that operates constantly on the cloud.''

In this article, we will summarize the differences between Gemini Spark and GPT-5.3-Codex-Spark, the impact on Japanese companies, changes in the development field, the future of the organization in 2030, and the areas of benefit that investors should look at.

Conclusions shown by two Sparks

The AI market is shifting from a chat-type to an agent-type.

The generation AI so far is

Human inputs prompts, AI answers

The main focus was on how to use it.

However, Gemini Spark and GPT-5.3-Codex-Spark are changing that premise.

Google has launched an AI agent that runs in the background in the cloud and works across Workspace and external apps.

OpenAI has developed a model that collaborates with humans in Codex in real time to rapidly modify code and change the UI.

In other words, the issue for companies is

Whether to ask the AI a question

rather than

Where should AI be consistently integrated into business and development processes?

is moving to

Gemini Spark: A 24-hour digital organization

Gemini Spark is the core feature that makes Google's Gemini app more agent-like.

According to Google's official blog, Gemini Spark uses Gemini 3.5 and the Antigravity harness and is deeply integrated with Workspace tools such as Gmail, Docs, and Slides.

The biggest feature is that it is cloud-based and runs in the background.

Spark can continue with tasks even when users close their PCs or lock their smartphones.

Furthermore, there are plans to expand collaboration with external apps through MCP connectivity, and collaboration with Canva, OpenTable, Instacart, etc. is also indicated.

Google also explained that it is designed to require user verification during high-risk operations.

This is important.

Gemini Spark should not be seen as an "AI that moves completely on its own," but as an agent that autonomously performs some of its tasks under human direction and permission.

GPT-5.3-Codex-Spark: Lightning-fast development engine

GPT-5.3-Codex-Spark is a model for real-time coding that OpenAI announced in February 2026.

According to OpenAI officials, GPT-5.3-Codex-Spark is a lightweight version of GPT-5.3-Codex and the first model designed for real-time development on Codex.

The characteristic is speed.

OpenAI explains that its partnership with Cerebras will enable output of over 1000 tokens/sec on low-latency hardware.

It is also available as a research preview for ChatGPT Pro users and will be available in the Codex app, CLI, and VS Code extension. Regarding the API, we will start with a limited number of design partners.

The main features are as follows.

Rather than being a model that autonomously performs large tasks over long periods of time, Codex-Spark is a model that allows humans to quickly rewrite the code in front of them and immediately check and modify it.

Gemini and Codex have different roles

Although Gemini Spark and GPT-5.3-Codex-Spark are competitors, the way they are used within companies is quite different.

Gemini Spark is the AI that runs your company.

Codex-Spark is an AI that creates software.

If these two things evolve simultaneously, companies will move toward ``increasing processing capacity without increasing the number of people'' in both business operations and development.

Cloudization of the workforce

What both Sparks have in common is the cloudification of the workforce.

Traditionally, companies needed to hire more people to increase their business volume.

However, in the age of AI agents,

• For back office work, go to a business agent like Gemini Spark
• Development work will shift to a real-time development model like Codex-Spark
• Long-term tasks are sent to higher-level models or subagents

distributed in the form of

This is not a story about immediately reducing labor costs to zero.

Rather, the story is that repeatable, tool-like tasks that are performed by humans will be transferred to AI in the cloud.

For companies, part of intellectual labor, which used to be a fixed cost, can be turned into a variable cost.

For workers, it becomes difficult to maintain market value through simple tasks.

Impact on Japanese companies

Japanese companies have a lot of room to introduce AI agents.

There are three reasons.

1. Labor shortage

In Japan, the working-age population continues to decline.

Hiring difficulties are spreading not only to field positions, but also to administrative, accounting, sales management, and information systems departments.

Therefore, AI agents are easily accepted not only as a ``threat to steal jobs,'' but also as ``infrastructure to supplement the lack of manpower.''

2. Lots of routine work

Japanese companies often have routine tasks such as approvals, reports, meeting materials, Excel tabulation, and email coordination.

These are areas where Gemini Spark type agents excel.

3. Legacy systems remain

On the other hand, many Japanese companies have outdated internal systems.

Just adding an AI agent as is will not work here.

Data preparation, authority design, work flow redesign, and security measures are required.

For this reason, demand for major system integrators, cloud implementation support, and business reform consulting is likely to increase in Japan.

Changes in development sites

High-speed models like Codex-Spark shorten development lead times.

In particular, the following areas will change:

• UI fixes
• Minor refactoring
• Added test
• Type definition and documentation corrections
• Prototyping
• Replacement of existing code

In the past, developers sometimes lost concentration while waiting for AI output.

When output of 1000 tokens/sec becomes practical, code will be generated while humans are thinking.

However, high-speed models are not a panacea.

Complex design decisions, security, performance, data models, and long-term maintainability require review by humans or more accurate models.

In practice,

Draft with Spark and reviewed by top models and humans

It is important to use them properly.

Corporate structure in 2030

Companies in 2030 will no longer be measured solely by the number of employees.

The important thing is

*Sales per person

• Operating profit per person
• Number of AI agents in operation
• Automated work ratio
• Development lead time
• Time available for humans to make decisions

It is.

A typical company in 2030 could have the following structure:

Key point
  ↓
Key point
  ↓
Gemini SparkKey point
  ↓
Codex-SparkKey point
  ↓
Key point

Humans no longer need to do all the work themselves.

Instead, AI will have a heavier role in giving purpose, evaluating results, determining risks, and building consensus with customers and people inside and outside the company.

Human resources who will survive in the AI era

In the age of AI, the important thing is not to work faster than AI.

The important thing is to use AI correctly.

Human resources who are likely to survive have the following characteristics.

The evaluation is not based on the simple amount of work done, but on how big results can be achieved using AI.

Beneficial areas for investors to look at

The market theme represented by the two Sparks is the AI agent economy.

The areas of interest are as follows.

Comparing the performance of individual AI models is not the only thing that is important for investors.

The more AI agents there are, the more cloud, inference semiconductors, SaaS, security, and data infrastructure will be used.

It is important to read this ripple effect.

Risks and precautions

There are high expectations for the AI agent market, but there are also risks.

• Accuracy risk for high-speed models
• Misoperation or wrong sending
• Leaks of confidential information
• Complicated authority management
• Underestimation of implementation costs
• Inadequate on-site work design
• Regulatory and audit compliance

Particularly in corporate implementation, the biggest issue is how much authority should be given to AI agents.

Companies that are unable to design boundaries between operations that should be confirmed by humans and operations that should be left to AI will have limited implementation effects.

Summary

Although Gemini Spark and GPT-5.3-Codex-Spark have the same name, Spark, their roles are different.

Gemini Spark is an AI agent that handles tasks.

GPT-5.3-Codex-Spark is an AI agent that speeds up development.

If these two things become popular at the same time, companies will move their intellectual labor to the cloud, both in terms of white-collar work and software development.

Looking toward 2030, a company's competitiveness may be measured not by the number of employees, but by how safely and efficiently it can operate AI agents.

For investors, it is important to track the entire agent economy, including cloud, semiconductors, SaaS, security, and SIer, rather than just AI models alone.

Source

• Google Blog “The Gemini app becomes more agentic, delivering proactive, 24/7 help” (May 19, 2026)
• Google Blog “Building the agentic future: Developer highlights from I/O 2026” (May 19, 2026)
• OpenAI “Introducing GPT-5.3-Codex-Spark” (February 12, 2026)