Microsoft and Knowledge Management
Knowledge management (KM) is hard. This has been proven to me many times over the last 30 or so years I’ve been paying attention to the concepts. With the announcement of the abrupt retirement of Viva Topics as of February 22, 2025, I’ve been thinking about the history of KM solutions from Microsoft and where they have all gone.
But first, a few tangents…
In 1996, I took a job as a consultant at a company called Renaissance Solutions. It was started by Harry Lasker, David Norton, and David Lubin to marry the concepts of the Balanced Scorecard (BSC), knowledge management, and technical solutions. That sounds simple today, but back then it wasn’t, at least from a technical perspective. We analyzed the work our clients did to understand how to improve performance by measuring the right things (BSC), leveraging the knowledge within the organization (KM), improving processes (a form of business process reengineering – not the destructive kind), all supported with technical solutions. I only mention this wonderful experience to show that KM has been in my blood for a long time, and I’ve got at least a little cred with it. I like to think there’s a healthy dollop of KM thinking in all the solutions I build to this day.
Why is KM hard? Well, it requires a concerted commitment by an organization to do a lot of up-front work which may not pay dividends for a decent amount of time; the return on investment (ROI) is hard to prove for a while. Figuring out what knowledge the organization has, what parts of that knowledge are high-quality, which parts contribute to the overall organizational strategy, and how to quantify that knowledge and distribute it to the appropriate people at the right time takes a lot of effort. That effort means people need to be in positions to execute (in-house or outside consultants, but in the long run, in-house is a far better approach) for a long enough period of time with enough executive commitment to lay the foundations – before that ROI starts to roll in.
Organizational culture figures into all this, too, of course, and changing a culture is in my mind like turning the proverbial battleship: it takes miles of effort before it happens. To me, an organization’s culture is simply the sum of its people. No fancy initiatives change that culture quickly if the people have mindsets that don’t match the desired culture. On top of that, incentives which support all of the KM activity also tend to lag the KM efforts by a year or more, as the Human Resources function figures out how to integrate it into performance metrics and reviews.
Now, back to Microsoft, which has repeatedly tried to offer knowledge management capabilities in the SharePoint and Microsoft 365 platforms. If you’ve been watching as long as I have, you’ll probably see that I’m not hitting all of them here, but these are the most important attempts that I recall over the years.
In SharePoint 2003, we had the concept of Areas, which allowed us to focus work around the area/topic to which it related. It didn’t work all that well, but it didn’t matter that much as it totally disappeared in SharePoint 2007. It may have been promising, but it didn’t last long enough to prove out.
At Microsoft Ignite in 2015, Microsoft announced something called InfoPedia. It never saw the light of day, and the best description I can find of the hopes for it are in this article from my friend Benjamin Niaulin (@bniaulin) at ShareGate: Office 365 Redefines Knowledge Management. The idea behind this mystical portal was to provide a prepackaged approach to provide knowledge to the organization. It sounded fantastic! But other than some fairly interesting demos, it never happened.
The most recent – and in my mind, the most powerful – KM solution from Microsoft was Viva Topics. Yes, “was”, because they have announced its retirement, with no replacements planned. Viva Topics provides a way to manage knowledge objects (topics) by attaching definitions, people, and canonical content to them. That’s important, but the coolest part about it is wherever those topics are found in SharePoint pages or Teams conversations, the topic is presented as a link to that topic card. With that one link, we can provide some of the mythical KM holy grails, like Find the Expert, a notion of Communities of Practice and/or Communities of Interest, canonical artifacts and more. My guess is the uptake for Viva Topics was never high enough to provide a reliable revenue stream to Microsoft to support it. Organizations that wanted to use it had to pony up for the licensing fees as well as the KM efforts to make it actually work.
Microsoft has now decided to focus almost all of its efforts on Copilot, which is ostensibly the reason why Viva Topics is being retired. This thread Susan Hanley (@susanhanley) on Twitter encapsulates concerns that some of us have about this.
Copilot – in whichever form (Microsoft has introduced some 150+ Copilots so far – very overwhelming!) responds to queries using Large Language Models (LLMs). I’m not going to go into the technology behind any of this; there’s plenty out there if you choose to dig deeper. One of the important benefits is that – by paying a per user fee – you can connect Copilots to your organization content repositories.
From a knowledge management perspective, I see several very important issues with a total reliance on Copilots:
- Content Curation – For KM to be truly effective, there’s a curation workload to it. Only Subject Matter Exprts (SMEs) can make the nuanced evaluations about where a particular knowledge object has no value, some value, or high value. Viva Topics gave us that curation engine, as clunky as it is to work with.
- Content Management – When you turn Copilots on your content repositories, it is going to pick up everything it can find with relevance, much like a search index does. Yes, the semantic index is a bit different, but the core idea is very similar.) If you have old or outdated or incorrect content, Copilot will find it and incorporate it in its responses. Thus, effective content management becomes even more important that it has been. Many organizations are more in the content collection business than the content management business, so this challenge is now even more critical in the era of Copilot. Unless the organizations which sign up for Copilot do some serious housecleaning first (or at least in parallel), they will be getting some very unreliable results.
- Search skills aka “prompt engineering” – We’ve had truly amazing search, both in SharePoint and Microsoft 365 as well as the Web in general – for decades now. How many times have you heard “search sucks”? Search itself doesn’t suck, but search skills are greatly lacking in the general population, IMO,. If one can’t come up with a good search query to find something, how can one come up with a good Copilot prompt, which has to be even more detailed and precise? Sure, we can train people on prompt engineering, but did we ever train them on search skills? Rarely.
- Result Quality – If you’ve used any of the current “AI” tools – Copilot included – you know that they quite often – if not a majority of the time – return questionable results. They may not be wrong, but just as with search results, they may not be “right”. If we need to validate every result we get (remember – no curation and probably iffy content management, too), then we have to be SMEs ourselves to evaluate the results. If I ask Copilot to describe the iron smelting process, I’ll get a result, but is it right? I have no idea, because I’m not an iron smelter.
So, is everything dire? Am I simply a naysayer? No, not at all. I have no doubt that the AI tech we have today – as amazingly impressive as it is – is in its infancy. It will continue to improve, and at some point, we will be able to trust all its responses to be the absolutely best responses. At least more often than not.
It’s going to be a very wild ride – one that many of us won’t see to fruition. Let’s all just be realistic about where we are. Microsoft has tried to give us KM solutions many times in the past, unfortunately, none with the longevity it would have taken to truly succeed. Copilot is simply the next in the series. Let’s see where it takes us.
Any sufficiently advanced technology is indistinguishable from magic.
Arthur C. Clarke
Great article.
I tried several times to introduce elements of KM in organisations; Lotus Notes was my first knowledge repository. Almost all of them fell down because it was hard to get people to proactively share what they know, or to use knowledge tools to seek what they don’t (if they knew enough to ask the right questions).
Copilot has the advantage that it can take into broad organisational knowledge across many sources via the graph. However, until we can find better ways to capture the tacit knowledge/ expertise in individuals and processes AI will remain limited. Perhaps we need AI with enough agency to ask us the questions we didn’t think to ask.
I can only agree that it takes a huge effort to get people to share their knowledge, and often there is a very good reason why. In most cases people are measured and rewarded bases on getting things done, and time spent on knowledge sharing is at best not rewarded. In most cases the organization in reality would prefer getting more stuff done, rather than ensuring that someone else might gain that knowledge.
Sorry for sounding like a broken record, and I have seen this behavior the last 35 years and it is still very much the case đŸ˜’
Nothing worth doing is easy. One of the challenges with KM is it requires a visionary to get it going. Once it’s baked into processes and culture, it’s not all that difficult, but that rarely happens.
Yes I agree with you business leaders you would think would want to promote/reward for people inputting into the knowledge base so that people don’t have to keep reinventing how something is done.
Certainly what we do at my company, I hate it when I can’t search on the Intranet to find something but love it when people can find the answers to their questions via the Intranet.
Now *that* would be a cool Copilot: “Here are the questions you should be asking…”
i have a list of 11 copilots, where can I find information about the 150 you referenced? I’d like to see what I might be missing.
@Dean:
I pinged the Twitterverse on this, and @donasarkar was kind enough to point me to this article: Adopt, extend and build Copilot experiences across the Microsoft Cloud