Microsoft infuses enterprise agents with deep reasoning, unveils data Analyst agent that outsmarts competitors

Microsoft has built the largest ecosystem of the corporate AI agent and is now expanding its advance with new powerful capacities that position the company to come in one of the most exciting segments of Enterprise Tech.

The company announced on Tuesday evening two important additions to its Copilot Studio platform: in-depth reasoning capacities which allow agents to tackle complex problems thanks to a meticulous and methodical reflection and the flow of agents which combine the flexibility of the AI ​​with the deterministic automation of business processes. Microsoft has also unveiled two specialized deep reasoning agents for Microsoft 365 Copilot: researcher and analyst.

“We already have clients with thousands of agents,” Vice-President of Microsoft companies, the co-pilot of companies and industry, Charles Lamanna, in an exclusive interview on Monday in Venturebeat on Monday. “You are starting to have this type of agent workforce where, no matter what work, you probably have an agent who can help you do it faster.”

Microsoft’s distinctive analyst agent

While the researcher’s agent reflects the capacities of competitors such as the in -depth research of OpenAI and the in -depth research of Google, the analyst agent of Microsoft represents a more differentiated offer. Designed to function as a scientist of personal data, the analyst’s agent can process various data sources, including Excel files, CSVs and integrated tables in documents, generating information via the code execution and visualization.

“This is not a basic model outside the shelf,” said Lalanna. “It’s a lot of extensions and adjustment and training in addition to basic models.” Microsoft has exploited its in -depth understanding of Excel work flows and data analysis models to create an agent that aligns with the way business users really work with data.

The analyst can automatically generate Python code to process downloaded data files, produce visualizations and provide commercial information without requiring technical user expertise. This makes it particularly precious for financial analysis, budget forecasts and the use of operational reports which generally require in -depth preparation of data.

Deep reasoning: bring critical thinking to company agents

Microsoft’s deep reasoning capacity extends the capacities of agents beyond the simple achievement of tasks to a complex judgment and analytical work. By integrating advanced reasoning models such as O1 of Openai and by connecting them to business data, these agents can solve more methodically ambiguous commercial problems.

The system dynamically determines when invoking deeper reasoning, either implicitly based on the complexity of tasks, or explicitly when users include guests as “reason on this” or “really think about that”. Behind the scenes, the platform analyzes the instructions, assesses the context and selects the appropriate tools according to the requirements of the task.

This allows scenarios that were previously difficult to automate. For example, a large telecommunications company uses in -depth reasoning agents to generate complex RFP responses by assembling information from several internal documents and knowledge sources, Lamanna in Venturebeat told. Likewise, Thomson Reuters uses these reasonable diligence capacities in mergers and acquisition reviews, dealing with unstructured documents to identify ideas, he said. See an example of the work agent reasoning in the video below:

Agent flow: reinvent process automation

Microsoft has also introduced agent flows, which effectively evolve the automation of robotic processes (RPA) by combining work flows based on rules with AI reasoning. This responds to customer requests to integrate deterministic commercial logic with flexible AI capabilities.

“Sometimes they don’t want the model to be freestyle. They don’t want AI to make its own decisions. They want to have hard -coded commercial rules,” said Lamanna. “Other times, they want the agent to be free and made judgment.”

This hybrid approach allows scenarios such as the prevention of intelligent fraud, where an agent flow can use conditional logic to transport the requests for reimbursement of greater value to an AI agent for an in -depth analysis against policy documents.

PET AT Home, a petty pet supplies based in the United Kingdom, has already deployed this technology for fraud prevention. Lamanna revealed that the company had saved “more than a million pounds” thanks to the implementation. Likewise, Dow Chemical has produced “millions of dollars saved for transport and management of freight” by optimization based on agents.

You will find below a video showing that the agent circulates at work:

The advantage of the Microsoft graphic

Central of Microsoft’s agent strategy is its integration of corporate data via the Microsoft graph, which is a complete map of working relationships between people, documents, emails, calendar events and commercial data. This provides agents with a contextual consciousness that lacks generic models.

“The less known secret capacity of the Microsoft graph is that we are able to improve the relevance of the graph based on engagement and how much files are closely connected,” Lamna revealed. The system identifies the most referenced, shared or commented documents, ensuring that agents refer to the sources that are authoritative rather than obsolete copies.

This approach gives Microsoft a significant competitive advantage compared to autonomous AI providers. While competitors can offer advanced models, Microsoft combines them with the work context and the explicitly explicitly explicitly explicit adjustment for business use cases and Microsoft tools.

Microsoft can take advantage of the same web data and the same model technology as competitors can, the Lamanna noted: “But we then also have all the content inside the company”. This creates a steering wheel effect where each new agent interaction more enriches the understanding of the graphic of work models.

Adoption and accessibility of companies

Microsoft has given priority to make these powerful capacities accessible to organizations with variable technical resources, said Lamanna. The agents are exposed directly in Copilot, allowing users to interact by natural language without expertise in fast engineering.

Meanwhile, Copilot Studio provides a low code environment for the development of personalized agents. “It is in our DNA to have a tool for everyone, not just people who can start a Python SDK and make calls, but anyone can start building these agents,” said Lamanna.

This accessibility approach has fueled rapid adoption. Microsoft previously revealed that more than 100,000 organizations used Copilot Studio and that More than 400,000 agents were created in the last quarter.

The competitive landscape

While Microsoft seems to lead the deployment of business agents today, competition is intensifying. Google has expanded its gemini capacities for agents and agent coding, while the O1 model of Openai and SDK agents provide powerful reasoning and agent tools for developers. Large corporate application companies such as Salesforce, Oracle, ServiceNow, SAP and others have all launched agent platforms for their customers in the past year. And also on Tuesday, AWS AWS published an AI agent, called Amazon Q to Quicksight, to allow employees to engage via natural language to carry out a data analysis without specialized skills.

Employees can use natural language to carry out an analysis of the experts at the level of the experts, ask questions whatsoever and obtain usable recommendations, caregivers

However, Microsoft’s advantage lies in its more complete approach – a solid coupling with the main company of reasoning models, OpenAI, while offering a model choice, an entrepreneurial infrastructure, in -depth integration of data on work tools and an emphasis on commercial results rather than on raw IA capabilities. Microsoft has created an ecosystem that looks like the best practice by combining personal co -pilots that include individual work models with specialized agents for specific trade processes.

For corporate decision-makers, the message is clear: the agent’s technology has matured beyond experimentation to practical commercial applications with a measurable return on investment. The choice of the platform depends more and more on integration with existing tools and data. In this area, Microsoft has an advantage in many areas of application due to the number of users he has, for example, in Excel and Power Automate.

Look at my full interview with Charles Lamanna has integrated below to hear first-hand how Microsoft stimulates his agent strategy, which these new capacities mean for business users and how organizations use agents to provide measurable commercial results: