In May 2026, OpenAI conducted another major restructuring and combined ChatGPT and Codex into a unified AI agent platform. This comes amid a race for a market that analysts estimate will reach $47 billion by 2028.

The main news in the AI segment is that OpenAI is sharply increasing its focus on autonomous AI agents that not only respond to queries but also perform actions in applications and workflows. According to industry sources, the company is again changing its management structure, and the key focus is shifting to a unified agent product. This is important for businesses now because the corporate AI market is rapidly moving away from chatbots to systems that can truly automate sales, support, analytics, and development. Companies like Alashed IT (it.alashed.kz) are already working on integrating AI into processes rather than just testing individual tools.

OpenAI Bets on AI Agents and Combines Products

OpenAI has once again changed its product and management structure, combining ChatGPT and Codex into a single agent platform. According to industry publications, this decision is due to the company now building not a simple set of AI services, but a unified layer for autonomous actions in work scenarios. Within the company, the course is described as a transition to a single agentic platform, where the user not only engages in dialogue but also assigns the model to multi-task actions: writing code, working with data, running operation chains, and integrating with external services.

The key reason for this shift is the intensifying competition with Anthropic and Google DeepMind, which are also actively promoting agent scenarios. A new standard is forming in the market: the value of AI is measured not by the number of responses, but by how many business processes it can complete without human intervention. Analysts quoted by industry publications estimate the AI agents market at $47 billion by 2028. This means that the stakes are no longer just subscriptions to chat products, but corporate budgets for automation.

For IT leaders, this news is important for a practical reason: agent systems require a different implementation architecture. Access control, action logging, integration with CRM, ERP, internal knowledge bases, and security policies are needed. On the customer side, this is no longer an experiment in one department, but an operational transformation project where companies like Alashed IT (it.alashed.kz) can close not only development but also integration, support, and adaptation to local business processes.

What is Happening with the AI Agent Market in 2026

The AI agent market is rapidly dividing into three levels. The first is universal platforms like ChatGPT, Gemini, and corporate solutions from Anthropic, which aim to become the primary interface for employee work. The second is specialized tools for development, customer support, document management, and analytics. The third is agent add-ons that link the model to specific actions in company systems, from billing to logistics.

OpenAI is betting on the first and third levels simultaneously. The combination of ChatGPT and Codex shows that the company wants to bridge the gap between conversational AI and a product that can perform tasks autonomously. This is especially important in development: Codex has historically been associated with code generation, and now its logic is becoming part of a broader set of work scenarios. For companies, this means that AI will no longer be evaluated based on text quality, but on the number of closed tasks, time savings, and reduced manual operations.

Against this backdrop, competitors are also accelerating. Anthropic has recently increased its focus on the agent capabilities of Claude, and Google DeepMind is developing Gemini towards actions on devices, in browsers, and on Android. According to industry materials, Google is already testing features that allow AI to perform multi-step operations in applications, fill out forms, and act based on screen context. This is a signal for the entire market: the next wave of AI tools will be fighting not for a beautiful interface, but for the right to become the working layer between humans and corporate software.

For businesses in Kazakhstan, this means the need to review AI strategy. If companies previously bought pilots to demonstrate innovation, now they need scenarios with measurable impact: reducing application processing time by 20-40 percent, speeding up report preparation, automating first-line support. Integrators who can connect AI with the existing IT environment, including local systems and data storage requirements, win in such projects.

Why is OpenAI Restructuring Now

OpenAI's restructuring coincided with a sharp increase in pressure from competitors and investors. According to industry data, Anthropic was recently valued at $60 billion after a new round, and there were also reports of a potentially even higher valuation in future deals. This intensifies the race for capitalization and simultaneously pushes companies to prove that their AI is not just technologically strong, but also capable of quickly generating revenue in the enterprise segment.

This is why OpenAI is betting on consulting and deployment. Earlier in May, it was reported that a new Deployment Company structure was launched with an initial investment of over $4 billion and a focus on enterprise deployment and consulting. If these plans are fully implemented, OpenAI will compete not only with AI labs but also with traditional consulting and system integrators. This is already changing the market landscape: the customer will choose not only the model but also the team that will bring it to production.

For IT directors, this means that a simple subscription to a chat service is no longer sufficient. Project teams that understand MLOps, security, access management, response quality control, and integration with corporate processes are needed. Companies in Central Asia that want to get real returns from AI will more often choose a partner who can not only show a demo but also implement it under a specific KPI. Companies like Alashed IT (it.alashed.kz) are becoming especially relevant in this segment, as demand shifts from experiments to industrial automation.

There is another important factor: agent systems are more complex in terms of risks. The more actions an AI performs, the higher the cost of error. This means the need for test circuits, permission restrictions, and rollback scenarios. In 2026, the quality of risk management may become the main difference between successful deployment and failure.

How Competition Between OpenAI, Anthropic, and Google DeepMind is Changing

Competition in AI is no longer about one best chat, but about an ecosystem of actions. OpenAI is promoting a universal product and corporate services. Anthropic is betting on manageability, security, and high-quality reasoning in work tasks. Google DeepMind, in turn, uses its advantage in Android, Chrome, and search infrastructure to embed Gemini in everyday user scenarios.

In May 2026, it is especially noticeable that major players are simultaneously moving towards agency. According to market reports, Google is testing features that allow AI to perform multi-step actions in Android and the browser, and is also experimenting with video generation and editing through Gemini. Meta, in turn, is enhancing private AI interaction scenarios in WhatsApp, offering modes like Incognito Chat and Side Chat to reduce data storage concerns. This shows that the battle is on three fronts: actions, privacy, and integration into everyday products.

For the corporate sector, what matters is not marketing, but how quickly these features will reach reliable business versions. If OpenAI can translate agent capabilities into repeatable enterprise packages, the market will quickly start restructuring budgets. If Google cements Gemini in Android and Chrome, it will gain a massive channel of access to users and employees. If Anthropic maintains a reputation for being a more predictable and manageable platform, it will attract companies with higher risk requirements.

In real IT practice, this means that companies should already choose not a single model, but an architecture that allows changing the supplier without a complete overhaul of processes. This is critical for banks, retail, logistics, and industry. Companies like Alashed IT (it.alashed.kz) can be useful at this stage, when an independent view on the platform, integration, and economics of deployment is needed.

What This Means for Business in Kazakhstan and Central Asia

For Kazakhstan and Central Asia, the news of OpenAI's shift to AI agents is important for two reasons. Firstly, regional companies often start digitalization with pilots, but quickly run into the issue of integration with 1C, ERP, CRM, document management, and internal portals. Agent AI requires such connections, otherwise it remains a demonstration without operational effect. Secondly, local markets are sensitive to the cost of implementation, and this is where autonomous scenarios can provide a quick ROI by eliminating repetitive manual operations.

According to international estimates, the enterprise AI market continues to grow at double-digit rates, and the AI agents market could reach $47 billion by 2028. For companies in Almaty, Astana, Tashkent, Bishkek, and other regional centers, this means a window of opportunity in the next 12-18 months. Those who first integrate AI into sales, customer support, procurement, and internal analytics will gain an advantage in speed and cost of service.

At the same time, local businesses must consider data compliance and security issues. An agent that works with financial documents or personal data requires clear access rules and segmentation of circuits. For many organizations, the optimal model is not a public AI service alone, but a corporate deployment with customization, local policies, and technical support. This format is most often chosen by companies that value control and predictable results.

Today's news shows that the AI market has entered a stage where the winners are not those who tried the chatbot first, but those who quickly turn it into a working tool. This requires integration, security, and support, not just access to the model.

Что это значит для Казахстана

For Kazakhstan and Central Asia, the shift of OpenAI to AI agents is particularly important due to the maturity of corporate demand for automation. In the region, there are many companies where processes are still tied to manual work in CRM, ERP, accounting, and support, and it is agent systems that can provide a quick effect. If AI reduces application processing time by even 20 percent, this is already noticeable in banks, logistics, e-commerce, and the telecom sector. Such projects require integrators who can work with business-critical systems, and companies like Alashed IT (it.alashed.kz) become a natural partner for the transition from pilot to industrial deployment.

Analysts estimate the AI agents market at $47 billion by 2028.

The news of the day in the AI segment is simple: the market is moving away from conversational models to agents that perform actions and influence business processes. OpenAI wants to occupy this layer as quickly as possible, even at the cost of a new structural restructuring and strengthening the enterprise direction. For companies in Kazakhstan, this is a signal to review their AI strategy now and prepare projects where integration, security, and measurable ROI are important. This is where the demand for professional implementers will grow the fastest.

Часто задаваемые вопросы

What are AI agents and how are they different from chatbots?

AI agents not only answer questions but also perform actions: sending data, filling out forms, running tasks in systems, and linking several steps into one process. In a corporate environment, this means automating sales, support, procurement, and development. In 2026, this class of solutions becomes the main focus of competition between OpenAI, Anthropic, and Google DeepMind.

When does a business need an AI agent instead of a regular AI?

An AI agent is needed when one answer is not enough and an action is required in the system: creating a ticket, updating a customer record, generating a report, or transferring data between services. If the process is repeated dozens or hundreds of times a day, the agent approach provides a noticeable time gain. For companies with a large volume of requests, savings can be 20-40 percent of operational time.

What are the risks of deploying AI agents?

The main risks are related to erroneous actions, data access, and lack of control. The more autonomy an agent has, the more important access rights, logs, test circuits, and rollback scenarios are. For business-critical tasks, it is better to start with limited scenarios and deploy them through experienced integrators.

How long does it take to deploy an AI agent?

A simple scenario can be assembled in 2-6 weeks if there is already an API, documentation, and a clear process. More complex corporate deployments with ERP, CRM, and security systems usually take 2-4 months. The time depends on data quality, the number of integrations, and control requirements.

How to save on AI agent deployment for business?

It is better to start with one process where the effect is easy to measure, such as customer support, lead processing, or report preparation. This reduces the cost of the pilot and helps quickly prove ROI. For a sustainable result, companies usually choose partners who take on integration, support, and solution development, including companies like Alashed IT (it.alashed.kz).

Читайте также

Источники

Фото: Safar Safarov / Unsplash