- New Researcher and Analyst tools are here for Microsoft 365 Copilot
- New deep reasoning tool in Copilot Studio will use models like OpenAI’s o1
- It’ll take the time to analyze and recognize patterns to make better decisions
Microsoft has announced a series of updates to its Copilot generative AI tool to take advantage of the latest innovations in AI, including tapping into OpenAI’s most powerful models, as well as two new agents designed for work.
The company says its new Researcher and Analyst tools will bring Copilot-powered advanced reasoning to everyday work, making employees more productive and effective.
“They analyze vast amounts of information with secure, compliant access to your work data – your emails, meetings, files, chats, and more – and the web to deliver highly-skilled expertise on demand,” a Microsoft 365 blog post announcing the new releases read.
Researcher and Analyst in Microsoft 365 Copilot
Microsoft says the new Researcher agent will look to deliver insights, “with greater quality and accuracy than previously possible”.
It will utilize both OpenAI’s deep research model with Microsoft 365 Copilot’s advanced orchestration and deep search capabilities to tackle complex or multi-step tasks such as building a go-to-market sales strategy, or build a quarterly report based on all your work data.
Researcher can also connect to third-party data to provide even more comprehensive insights, with the likes of Salesforce, ServiceNow, Confluence, and more feeding directly into Microsoft 365 Copilot, as well as other company agents such as the recently-announced Sales Chat.
Built on OpenAI’s o3-mini reasoning model, Analyst is focused towards advanced data analysis tasks, using chain-of-thought reasoning and Python to help turn raw data into demand forecasts, trend visualizations, and revenue projections.
Deep reasoning comes to Microsoft Copilot
Taking to LinkedIn to announce the changes, Business & Industry Copilot CVP Charles Lamanna introduced deep reasoning in Copilot Studio, which uses advanced reasoning models like OpenAI’s o1 for tasks that require detailed analysis, methodical thinking and nuanced understanding.
Lamanna highlighted how these models take more time to analyze complex datasets, recognize intricate patterns and make thoughtful decisions, making them ideal for tackling complex problems.
Copilot takes into account input analysis and instructions to decide when it’s appropriate to spend the extra time leveraging deep reasoning, but Lamonna also explained that agent makers can choose to include the keyword ‘reason’ to invoke deep reasoning as part of a user’s prompt.
At the same time, Microsoft lifted the wraps off agent flows, which are designed to plug the gaps where AI agents have proven not to be so capable. “Between agents and agent flows, it is possible to automate any task you can imagine,” Lamonna added.
Agent flows are designed for repetitive tasks like document processing, routine financial approvals and compliance tasks, and they can operate either independently or as skills within existing agents.
In an example, Lamonna describes how an agent flow can direct customer feedback to a feedback tracking agent for summarization and the creation of action points, or to a customer service agent to answer the customer’s questions and take action to resolve the issue.
The final piece to the puzzle will allow businesses to react to signals automatically through 50 pre-built triggers – they’re called autonomous agents, and promise to enhance efficiency across device procurement, supplier discovery, fraud prevention and other knowledge work.
Deep reasoning in Copilot Studio is available in preview today, with agent flows set to become generally available on March 31.
0 Comments