Microsoft brings an AI-powered Copilot to its business app suite

Microsoft today introduced what it's calling the "next generation" of AI product updates across its business apps portfolio. They touch on both Power Platform, Microsoft's set of low-code tools for building apps and workflows, and Dynamics 365, the company's suite of enterprise resource planning (ERP) and customer relationship management (CRM) tools.

In an interview with TechCrunch, Charles Lamanna, CVP of business apps and platform at Microsoft, described the updates as the logical next step on Microsoft's automation journey. Powered by tech from AI startup OpenAI and built using the Azure OpenAI Service, Microsoft's service that provides enterprise-tailored access to OpenAI’s API, the new capabilities follow the rollout of OpenAI text-generating AI models in Power Platform four years ago and the more recent debut of generative AI capabilities in Viva Sales, Microsoft's seller experience app.

"Over the last four years, we've been on a journey to bring generative AI and foundation models to the workplace," Lamanna said via email, noting that Microsoft has a longstanding partnership with OpenAI to commercialize the vendor's tech in Microsoft's own products and through the Azure OpenAI Service. "And we've now reached the point where the tech and product can enable transformative outcomes for customers."

In Dynamics 365, Microsoft's launching what it calls Copilot (borrowing branding from GitHub's Copilot service), which -- broadly speaking -- aims to automate some of the more repetitive sales and customer service tasks.

Image Credits: Microsoft

For example, in Dynamics 365 Sales and Viva Sales, Copilot can help write email responses to customers and create an email summary of a Teams meeting in Outlook. The meeting summary pulls in details from the seller’s CRM, such as product and pricing information, Lamanna says, and combines them with insights from the recorded Teams call.

"We securely and intelligently access information from customers’ CRM, ERP and other enterprise data sources at runtime," Lamanna added. "We use large language models to combine the enterprise data with underlying knowledge to produce responses tuned for each customer. Importantly, we don't use customers’ data to train the models."

Over in Dynamics 365 Customer Service, Copilot can draft "contextual answers" to customer queries via chat or email and provide an "interactive chat experience" for customer service agents that draws from knowledge bases as well as case history. These complement the new "conversation boosters" feature in Power Virtual Agents, Microsoft's chatbot builder, which lets companies connect a bot to resources like a website or knowledge base to use that data to respond to questions that the bot hasn't been trained on.

In turn, conversation boosters complements a new "GPT" model in Microsoft's AI Builder tool that lets organizations embed text-generation features into their Power Automate and Power Apps solutions. Lamanna says that, for example, a researcher could use it to summarize text from weekly released reports and have it sent to their email, while a marketing manager could tap the GPT model to create targeted, generated content ideas by entering specific keywords or topics.

Given Microsoft's recent foray into generative text -- i.e. Bing Chat -- one might be reluctant to build an app using the company's tech lest it go off the rails. But Lamanna asserts that conversation boosters and the GPT model -- plus Copilot, for that matter -- are "grounded in reality" by each customer's CRM, ERP and other data sources.

"AI-generated content is always clearly labeled, and users are encouraged to verify the accuracy before using it. When relevant, we also cite the sources from which the answer was retrieved to better enable the user to verify the accuracy of the response," Lamanna said. "We have monitoring and controls in place to allow us to quickly respond with manual intervention in case any issues slip through the above lines of defense."