To Make Office 365 Smarter, Microsoft's Getting More Ambitious with Internal AI Testing

If you work at Microsoft, there is a good chance that you are beta, or more likely, alpha testing new software and services. The company calls this dog-fooding and this week, the company sent out an internal memo about an update to data policies that will have the company going all-in on its own AI and ML bets.

Starting in Q1 of 2019, the company will broadly begin using its own business-related data such as email, work documents, messages, video and audio data, to help improve the company’s AI. Microsoft will be rolling out this data collection practice internally to all employees in the US next year with a global rollout, depending on local laws and regulations, in the months ahead.

There are a bunch of caveats to how this data will be collected such as omitting communications from anyone not on a Microsoft.com domain email address, items marked as personal and confidential will be excluded as well but the goal here is simple, the company needs a lot of sample data to train its algorithms.

And this shouldn’t be a big surprise. Microsoft wants to improve its AI to make it easier to detect spam, malicious messages, and be more proactive with corporate content to make Office 365 a better tool for productivity.

As you would expect, this is a sensitive topic, even for employees at Microsoft. Because all of this data is now being collected, employees are concerned that their conversations may be taken out of context or possibly expose them to unnecessary scrutinization for having “watercooler” conversations via email which could be viewed as wasting time.

Because all this data will be collected and analyzed, at some point, human reviewers will need to access the data when questions arise. While Microsoft is promising employee confidentiality, there is no doubt that this practice is making some employees uncomfortable with the practice.

But, progress can be messy and Microsoft knows that other companies are working on similar technology and it doesn’t want to be left behind. Similar practices using this type of data have helped to create Outlook’s ‘Smart Reply’ and focus inbox features and with more data, Microsoft should be able to get smarter about how it catalogs and identifies corporate content.

Microsoft is a productivity company at heart and with more than 100,000 employees, it has one of the largest collections of enterprise data in the world. By working with its own data at a much larger scale, the benefits will extend to Office 365 which is a core pillar of the company’s long-term future. Even though some employees may be uncomfortable with the practice, there is little doubt that the leanings from this practice will help it improve the products that its customers use most.