Editorial
Microsoft's decision to shutter Viva Topics after three years came as a surprise to no one who had followed its trajectory.
Microsoft launched the knowledge management solution in 2021 with the lofty goal of using AI to automatically organize enterprise resources and expertise by topics so people could easily find information. Despite the noblest of intentions, Viva Topics struggled to gain traction.
Users found setting up topics overwhelming and ultimately, the results didn’t deliver the goods. The expert discovery capabilities also fell flat, as the AI failed to truly understand peoples’ areas of expertise out of the box.
In the end, Viva Topics ultimately failed to motivate people enough to engage with and contribute to its knowledge management system organically.
So how did we get here? A story in acts.
Act I: Viva Topics' First Iteration: Project Cortex
Microsoft first introduced the Viva Topics predecessor, Project Cortex, in November 2019 at its annual Ignite Conference.
The announcement stated, “Project Cortex uses advanced AI to deliver insights and expertise in the apps you use every day, to harness collective knowledge and to empower people and teams to learn, upskill and innovate faster. Project Cortex uses AI to reason over content across teams and systems, recognizing content types, extracting important information, and automatically organizing content into shared topics like projects, products, processes, and customers. Cortex then creates a knowledge network based on relationships among topics, content, and people.”
With promises like that from Microsoft, expectations were off the charts. It was quickly understood that to work, Cortex depended on a fair amount of upfront work to prime the AI engine to apply topics correctly. I have written before about the challenges of bootstrapping a topics engine and it turned out to be much harder than expected to define Cortex topics. And without a good baseline, the product was unable to classify information. Disappointment began to spread.
Act II: Viva Topics Launch
Just over a year later, Microsoft tried again, relaunching Cortex as Viva Topics in February 2021. This time the announcement boasted, “Viva Topics uses AI to reason over your organization’s data and automatically organizes content and expertise across your systems and teams into related topics, such as projects, products, processes, and customers.” It was going to work this time because, “[it] builds on the Microsoft 365 apps and content you already use every day and the connections you have with people across your organization.” Rebranded as part of the Viva suite may have simplified the pricing model, but it didn’t significantly simplify topic setup; what proved necessary to create a useful baseline that the AI engine needed to accurately interpret and classify new content.
Related Article: Microsoft Is Retiring Viva Topics. Here's What You Can Do
Act III: Viva Topics Lite
In July 2023 Microsoft announced Viva Topics lightweight management using Viva Engage; this time trying to bootstrap the setup by "enlist[ing] Viva Engage admins to edit, create and delete 'simple' topics … to make knowledge management more accessible to more users.”
3 Strikes Acts and You’re Out
Despite the efforts to simplify the setup of Viva Topics, implementations stalled. At the end of the day, the missing piece was melding the core technology to the messy, human aspects associated with defining an organizational taxonomy — a prerequisite for Viva Topics to work.
And while organizations were struggling to implement Viva Topics, generative AI was taking off in the marketplace. Microsoft, long perceived to lag behind Google in the generative AI space, saw a huge opportunity to leapfrog Google with its strategic investment in OpenAI. Management and development focus redirected to the exploding generative AI market. As such, Microsoft’s generative AI offering, Microsoft Copilot, proved to be the final nail in the Viva Topics coffin.
So, does Microsoft’s announcement spell the end of the road for topics (or topic computing)? Not by a long shot. Here’s why …
Thought by Association
The idea that people think by connecting associated thoughts is an ancient one, with roots as far back as Aristotle. In the 19th century, the psychologist William James articulated the idea in more modern concepts and in the last century, this idea gained even more credence through empirical evidence provided by cognitive psychologists, including recent studies using fMRI to measure brain activity during tasks that spark associative thoughts.
Back in 1945, Vannevar Bush, then Director of the U.S. Office of Scientific Research and Development, envisioned a personalized database of associated information connected by common topics in a device he called the Memex. People would store articles, books, notes, recordings and other materials in the memex and then group them by shared topics. A modern-day example will demonstrate how this worked.
Say you are reading an article about an electric vehicle. Suddenly you think about the operation of the vehicle’s battery. You find a book about battery technology. While perusing the book, you wonder how these batteries are charged and you find a YouTube video about charging stations. While viewing the video, you are astounded about how much acreage is used to build solar charging stations, so you navigate to a website that details how much agricultural land is being used for alternative energy. In this case, all the content pieces would be stored in the database and would be connected by a common topic, say "alternative energy." Creating topics would enable you to later revisit these ideas and retrace the thought process that led to this collection. It is a clever way to retain memory and spark new ideas.
A word about topics vs. tags
While these terms are used interchangeably, there is an important distinction in the world of Microsoft 365. In layman’s terms, tags are metadata labels used to define the context and content of a piece of content. Collectively, tags provide input to define a topic, which is a more general construct. In our example above, "electric vehicle" and "battery" might be tags that are part of the topic "alternative energy." Using these relationships, Viva Topics is able to apply AI and the Microsoft Graph to locate experts and relevant content about alternative energy inside the organization.
While Bush’s device was never built, contemporary note taking products like Evernote and OneNote fulfill his vision by empowering people to associate diverse types of content into common topics using tags.
Even from our simple example, it is easy to see the benefits of organizing your own personal knowledge. The ability to retrieve information and reconstruct ideas in the future is compelling. And yet, the ongoing effort needed to maintain these systems is certainly one reason why relatively few people take advantage of these products.
Related Article: You've Got a Taxonomy. But Can You Find What You're Looking For?
The Organizational Memex
Notably, Bush never addressed the challenge of organizing knowledge on a collective scale. The memex presumed each individual would define their own set of topics and associations.
But organizations, like individuals, think associatively. Teams and departments think in terms of topics like customers, products, services and projects. So, presumably, organizing knowledge in organizations would make it easy to find materials quickly, organized by shared topics. But there’s more …
Conference
Oct
20
Gartner IT Symposium/Xpo Orlando 2025
Register
Conference
Oct
27
Gartner HR Symposium/Xpo Orlando 2025
Register
The benefit of ordering collective knowledge is far greater at the organizational level. Because when knowledge is aggregated into collective knowledge, it becomes a means for sharing ideas, experiences and expertise. Rather than just spark creativity, collective knowledge delivers productivity and efficiency gains that provide a competitive advantage. When an expert in the Argentina office defines a world-class document to answer client bids, everyone else in the company benefits. Bid responders in other offices can simply edit the winning document, without having to reinvent the wheel. It is easy to see how this could boost a business’ productivity and competitive edge.
But if organizing personal knowledge isn't easy, trying to do it collectively begets a whole new dimension of challenges. Let’s start with the challenge of defining a common tag language (i.e. taxonomy).
Case in point: a number of years ago I worked with a large U.S. federal government agency working to define their organizational taxonomy. After several years and many meetings, project leaders were able to agree on just four tags to classify their documents — and two of those were a document’s name and its date of publication.
A different organization took another approach. In this case, management assigned a small team to define a set of metadata tags that embodied all forms of the company’s knowledge. Once defined, employees were instructed to use these terms to tag all new documents, emails, meeting summaries and recordings before storing them into the organization’s document management system. The strategy failed because the employees rebelled. Burdened by an additional task that made no sense to them, most employees found a way to avoid tagging content and the project failed.
Simply put, agreeing on how knowledge should be organized is a difficult organizational challenge, not a technological one.
Related Article: Reboot Knowledge Management for the Post-Pandemic Workplace
Viva Topics to the Rescue?
The inherent difficulty in creating, maintaining and automatically applying topic definitions is what made Microsoft’s bold attempt to solve this problem with Viva Topics so exciting. Using AI to automatically apply tags to content would forever remove the burden from organizations and employees to do the grunt work. The result? Business professionals would be able, for the first time, to find what they needed quickly and easily, without all the complexity of metadata. It sounded like magic.
The missing piece of course, was the need to prime the AI engine to understand the organization’s strategy for classifying content. Assuming that the taxonomy would somehow build itself was an unrealistic dream. Microsoft realized this early. Soon after Cortex shipped, it issued guidance for commissioning a knowledge manager, a content services manager and subject matter experts at each organization to craft an initial baseline for classification categories and then maintain it over time. Solid advice, but it raised the bar to getting Viva Topics off the ground. With expectations that this would be simple to implement, most organizations were disappointed and put off plans to get serious about topics. This ‘bootstrap conundrum’ proved to be Viva Topics undoing.
This is where Viva Topics ultimately failed. The process of creating that baseline of topics (and maintaining it) turned out to be too difficult for all but the most determined organizations. So, while the technology was/is good, it proved too difficult to implement.
Long Live Topics
Despite the demise of Viva Topics, many Microsoft 365 organizations will (continue to) pursue topics because the competitive payout is so compelling. Law firms, professional services firms, construction firms, pharmaceutical companies, manufacturers and other information-rich organizations rely heavily on documents and other data to drive their business. If they can overcome the taxonomy bootstrap problem to employ topics, they could pull ahead of their competitors.
What’s the answer?
AI is certainly part of the solution, but it’s not enough. Talk of ‘auto-tagging’ content is a real possibility, but as we have seen, even the most advanced AI engine needs a baseline from which to draw inferences. And an AI engine cannot automatically build a company’s taxonomy of metadata terms and topics to bootstrap itself. Because different people use different terms to describe the world they see, building and maintaining a coherent taxonomy is still very much a human activity.
That’s why the next step in the topics journey involves focusing on the human element, which includes both methodologies and user experience tools. Tools that make it easy for experts to define and maintain tags and topics, as well as deliver employee journeys that make it easy to use topics to find experts and authoritative content quickly. Simply put, getting everyone on the same page to drive the topic management process is the linchpin to success.
The gap between ‘what’s desired’ and ‘what’s possible’ is narrowing. The ability to align the way people (and organizations) think will reduce friction in daily workflows — a concept I will expand on in a future article.
Learn how you can join our contributor community.
