lab setting with a gloved hand placing a petri dish under a microscope
PHOTO: ThisisEngineering RAEng

Any vaccine goes through rounds of clinical testing to vet for efficacy and potential side-effects, with the results made public for all to see. We are all probably far more aware of this process as we look to the newly developed COVID-19 vaccines to usher an end to this year-long crisis. The clinical tests play a critical role in ensuring the ongoing safety of the vaccine as well as in building public trust.

Much like vaccines, Microsoft 365 has the potential to make equally far-reaching changes within the workplace, though without the life or death consequences. As the dominant platform, Microsoft is constantly delivering new functionality that quickly become the de-facto way of working for many businesses. Many of these new features are driven by algorithms and AI, and the same questions must be answered: do they work in practice, and what are the side-effects?

With new features coming straight out of Microsoft Labs and into production, I ask the question: have these features been clinically tested within workplaces? If so, will Microsoft be releasing the results?

The Algorithms Working Behind the Microsoft 365 Scenes

Algorithms have been driving what staff see within Microsoft 365 for some time. The Yammer feed is a perfect example. It provides a curated view of updates that seeks a balance between relevance and serendipity. These algorithms have proven hard to perfect. While all social tool vendors introduce regular refinements, the stakes are relatively small. In the worst case, somebody may miss an interesting item, but social tools typically share nice to know items rather than need to know.

Microsoft Graph has also aimed to make connections between people and documents, and people with people, based on the files they work with and the topics they cover. These connections are surfaced in Delve, "My feed" in SharePoint, and increasingly in search.

This is where it gets a bit experimental: how do the algorithms in Microsoft Graph work, and do they return relevant and useful results? Or do they share stuff that was assumed to be secret or private? While the technical details are well-understood, I’m not aware of any clinical results from Microsoft to show that it adds meaningful value to how businesses operate.

Related Article: Beware the Lasting Impact of a Microsoft 365 Non-Decision

Circumventing Internal Communications

Bigger changes are happening in the world of internal communications. With the release of the SharePoint app bar comes "My news," which contains a list of recommended news stories. Recommended how? Similar features are making their way into Teams and products such as Microsoft Viva.

This is a big deal for internal communicators, as it takes control out of their hands and places it entirely in the realm of algorithms. On what basis are news stories recommended? Are they the most important and most relevant items?

A big debate is currently underway around the algorithms that underpin Twitter and Facebook, as it’s recognized that these technologies have the ability to shape what we hear and ultimately what we think.

The same considerations apply within businesses, particularly in larger firms. Has Microsoft shared how the algorithms operate? Can they be tuned or changed to match the circumstances of an individual business?

Related Article: How Internal Communications Is Cutting Through the Noise

Microsoft Viva: Shaping What We Know and Feel

Microsoft Viva holds the greatest potential to demonstrate the power of algorithms and AI within the enterprise.

Viva Insights, for example, aims to help firms better support staff, addressing both wellbeing and productivity. The initial announcement promises: "Viva Insights for individuals help employees stay connected with their colleagues and protect time for regular breaks, focused work, and learning …. Viva Insights can help a manager see if their team is at risk of burnout and provide recommendations like encouraging your team to turn off notifications, set boundaries in their calendar, and set daily priorities to focus on what matters most."

These are big claims. Even more so when you consider how varied job roles are within firms, and how different work cultures can be between firms (and countries). This product aims to literally shape how people feel. The potential dangers are self-evident.

Again, has Microsoft outlined how these algorithmically-based recommendations are made? Has it conducted clinical testing in a suitably wide range of organizations, with the results publicly released? Or has it only been tested in white-collar organizations operating in the U.S.?

Viva Topics goes even further. It “uses AI to automatically organize company-wide content and expertise into relevant categories like projects, products, processes, and customers … No need to search for knowledge — knowledge finds you.”

If Microsoft has found a way of using AI to solve the problem of knowledge management, the benefits will be transformative. For decades, firms have struggled to "know what they know," and for staff to be consistently informed at the point of need. Many KM projects have foundered in the attempt of making this work at scale, across complex organizations.

The risks are equally huge, however. What if an employee reads an outline of a safety procedure provided by Viva Topics, and is then severely injured as a result? As the information is provided as a fact, on the corporate network, the employee can reasonably assume it’s correct, therefore making the company liable for the injury.

Topics is the very visible face of research that started with Microsoft Cortex, rapidly progressing to become a mainstream offering within Microsoft 365. Does the reality match the marketing, and will organizations understand how to get the most from it, while mitigating the considerable business risks? This could be the greatest impact of AI on how businesses work that we’ve seen to date.

Related Article: Is Viva Topics the World's First Topic Computing Solution?

Shining a Light on Algorithms

All of this is more than just an academic debate or a play on words in the pandemic era. Microsoft is by far the dominant player in the digital workplace. Moreover, many firms naturally assume that whatever Microsoft releases is the new best practice, and that it works without question.

However, if these features — rushed from the labs into production — don’t work, they may literally break the way we work, inflicting huge damage on firms and their ability to operate.

To proceed with confidence, this is what needs to happen:

  • Microsoft should conduct meaningful clinical trials on key new features and release the results publicly.
  • It must provide details on how each of the algorithms work, communicated in a way that is understandable to both IT and business teams.
  • It must be possible to monitor the activity of each of the algorithms, to see if the results are useful for staff.
  • Businesses must be able to tune the algorithms, to better match the organization and its goals. 

Under the leadership of Satya Nadella, Microsoft has sought a creditable balance of innovation and ethics. No longer the Evil Empire, the firm needs to be aware of the risks of rapidly progressing lab experiments directly into enterprise production.

Let’s get ahead of the issue before the first big crises appear, and make the most of new capabilities while being completely transparent on how they work under the hood. We desperately need this type of innovation, but we need it delivered in a way that surfaces and mitigates the business risks.

Microsoft, are you ready to be a good corporate citizen when it comes to enterprise algorithms and AI?

Related Article: IBM and Microsoft Sign 'Rome Call for AI Ethics': What Happens Next?