Microsoft wants generative AI to be a copilot for enterprise applications

Microsoft is doubling down on its generative AI efforts, announcing today that it is integrating the technology into its Microsoft Dynamics and Power platforms. The goal is to enable enterprise applications with the power of generative AI.

Microsoft has been steadily integrating AI into its portfolio in recent years, thanks in part to the company’s expansive partnership with generative AI vendor OpenAI. For developers, Microsoft has already built the GitHub Copilot service which benefits from OpenAI’s Codex model to act as a pair programmer. Now Microsoft is extending the copilot model and bringing it to its enterprise applications beginning with Microsoft Dynamics 365 suite. 

>>Follow VentureBeat’s ongoing generative AI coverage<<

The new Microsoft Dynamics 365 Copilot service brings a range of capabilities to help automate business application operations for customer relationship management (CRM) as well as enterprise resource planning (ERP). The capabilities include text generation, sentiment analysis and workflow automation to help make it easier for Dynamics 365 users to complete tasks quickly and accurately. Going a step further, the Microsoft Power Platform, which is a low-code service for application development, is now being integrated with generative AI to help enterprise developers build their own applications faster.

“We are in the middle of quite a large hype cycle around large language models and generative AI and how it’s transforming so much of the tech industry and software,” Charles Lamanna, corporate vice president of business applications and platform at Microsoft, told VentureBeat. “We think that there should be AI copilots for every role, every function and every line of business.”

How Microsoft is bringing generative AI to business applications

Lamanna explained that the Dynamics copilot capabilities will enable sales, customer service, marketing, finance and supply chain operations professionals to benefit from generative AI. 

He said that the copilot will help users generate ideas, create and understand content quickly, and automate repeatable and mundane tasks that users encounter every day. 

Among the specific copilot integrations that Microsoft is highlighting are ones for Dynamics 365 Sales and Viva Sales to help with the sales process.

• In Dynamics 365 Customer Service, the copilot has access to the case history of a customer and is able to help support personnel with customer inquiries.
• Dynamics 365 Customer Insights and Dynamics 365 Marketing benefit from AI integration to drive customer interactions, such as a new marketing campaign.
• Copilot in Dynamics 365 Business Central will now be able to help organizations rapidly create product listings.
• Microsoft Supply Chain Center will now benefit from copilot with intelligent alert to help organizations understand and respond to potential supply chain disruptions.

A copilot is more than just GPT-3

The new copilot efforts benefit from Microsoft’s partnership with OpenAI, and make use of the Azure Open AI service.

Lamanna emphasized that the new Microsoft Dynamics copilots are not just generic implementations of OpenAI’s models. Rather, what Microsoft has done is a lot of work around prompt engineering, context setting and ranking of responses, which help enrich the core foundation models.

He also noted that the Microsoft Dynamics copilots benefit from joint work Microsoft has done with OpenAI on the Codex generative AI model for code generation. For example, the copilot for Dynamics Marketing enables users to do a segment query over a database to create marketing segments. That query requires code generation, which is done on the fly by the Codex generative AI for the specific database query.

What’s important to also highlight is the fact that Microsoft is not using its own customer data to train the core generative AI models.

“We never use customer data to train the AI model itself,” Lamanna siad. “The guarantee we have across everything is that our customers’ data is their data and we won’t use it to train the model and potentially thereby leak the data.”

Lamanna emphasized that Microsoft only uses customer data to tune and optimize personalized prompt engineering. He said that there was close to a year of work around product engineering expertise to make the models behave the way that enterprise would want them to work inside of common workflows.

Enterprises now have the ‘Power’ to build generative AI-based applications

Beyond the prebuilt copilot capabilities now coming to Microsoft Dynamics 365, Lamanna said that Microsoft wants to also help enterprises integrate generative AI into other applications.

The Microsoft Power platform is a low-code application development service that can help organizations build workflows, automations, dashboards and chatbots, among other use cases. Inside the Power platform, there’s the AI builder service that organizations can use inside of Power apps. What Microsoft is now integrating inside of the AI builder is the ability to use GPT-3 inside of Power apps.

“The whole vision of the Power platform is to democratize technology,” he said. “At this point, if you want to build an app, a workflow or a chatbot, you have to be using the GPT AI models because they’re just so powerful. “

Overall, Lamanna said that the goal at Microsoft is to help organizations benefit from the power of generative AI in an approach that is easy to consume and extend.

“We think all aspects of software will fundamentally change as a consequence of these new AI capabilities,” he said. “What we are doing at Microsoft is we are reimagining our existing offerings and we are also creating the tools for other people to reimagine their offerings as well.”