|

Microsoft Builds Enterprise AI Stable With Recent Patents

Given the company’s dominance in workplace and productivity tech, it may have an edge in embedding generative AI into work routines.

Photo of a Microsoft patent
Photo of Microsoft Inter-Document Attention Mechanism patent via U.S. Patent and Trademark Office

Sign up to uncover the latest in emerging technology.

Microsoft has been on a mission to make AI your favorite co-worker. With its latest patents, it shows no signs of stopping. 

The company filed two patent applications for enterprise AI integrations. First, the company is seeking to patent “customized recommendation generation using enterprise data analysis and inference.” 

Microsoft’s tech essentially watches a user’s work and learns their favorite authors and collaborators to provide personalized suggestions. The system monitors the documents that a user accesses, creates, and interacts with, as well as all the contributors to said documents, giving each person a contribution score.

The system then creates an “authorship record” and adds it to a larger database, which is analyzed and fed to a trained machine learning model to create recommendations as to which users should collaborate on certain projects. 

And in the spirit of watching you work, the company is also seeking to patent an “inter-document attention mechanism.” Microsoft’s tech uses a neural network for natural language processing to get a better understanding of documents to perform certain tasks. 

This system would allow an AI model to glean more meaning from different words or phrases between multiple documents, such as how one phrase may influence or impact another. Microsoft’s tech uses this improved understanding for tasks such as analysis or responding to a query.

Much of Microsoft’s generative AI focus has been on productivity. Alongside packing its cloud services offerings with AI, the company dove in head-first on Copilot, its AI work companion, since its debut last March. 

AI adoption in enterprise settings hinges on boosting productivity across roles, functions and industries, said Tejas Dessai, research analyst at Global X ETFs, such as to “streamline workflows, automate repetitive tasks, improve communication, and support creative functions.” 

“With its broad product portfolio, Microsoft is uniquely positioned to develop and launch solutions that address a wide range of these needs,” Dessai said. 

Even before AI swept the market, the company’s suite of office software, Microsoft 365, represented a large part of its business, said Cornelio Ash, equity analyst at William O’Neil. As it stands, Microsoft 365 has an estimated 350 million users, Ash noted.

With Copilot costing around $20 per month, enticing even half of those users to adopt the AI companion stands to bring in a massive payday. “I’d like to believe it’s going to happen, it’s just at what speed [Microsoft] can get to those numbers,” said Ash. 

While AI has massive potential to boost productivity and efficiency in enterprise scenarios, these models still pose risks. For one, generative AI models still face data security issues — they tend to release their training data if prodded in just the wrong way. These models also have the propensity to go off the rails through hallucinations or bias. 

So despite tech firms’ excitement, it’s best to proceed with caution. “In the best scenario, it’s going to be an assistant,” said Ash. “You’re still going to need someone knowledgeable in security that’s going to be able to interpret the data. You can’t just have the output go directly to the market without some sort of verification.”