A new generation of AI agents: how are advanced reasoning models changing work?

In 2025, AI models become smarter and more accessible. According to Microsoft, the use of generative AI among business leaders has grown from 55 % to 75 % in just one year. Large models, such as OpenAI o1, are already solving complex problems with human-like logic steps, comparing contracts, generating code, and executing multi-step processes. Small models developed on high-quality, synthetic data (e.g. Phi, Orca) are proving that carefully selected training can deliver high performance with fewer resources.

These improvements are leading to a new wave of autonomous agents. Nearly 70% of Fortune 500 companies already use Microsoft 365 Copilot for everyday tasks (sorting emails, taking notes), and by 2025, a new generation of agents will be able to perform tasks on behalf of users. Agents will monitor supply chains, recommend suppliers, and issue orders automatically. Thanks to tools like Copilot Studio and Azure AI Foundry, anyone can create agents without code or develop complex agents with orchestration capabilities.