Knowledge in the Modern Workplace

Are our tools actually making us better?

In the last few years, hundreds of software products have been launched with the goal of improving how teams create and manage knowledge. There are universal search tools, powerful notes tools based on tagging, and even alternative apps for creating and collaborating on documents. But are these tools actually making us more productive?

At my last job, I was the Director of Product for a startup and was responsible for making sure that people were on the same page. Every day, I received hundreds of messages from Slack, JIRA, document updates, and more. I had to make sure that the important information got to the right place, so I spent hours essentially copy/pasting information from one tool to another. I was desperately looking for a solution that could help me and my team.

I tried many purported solutions and tested numerous note taking tools, knowledge management systems, and productivity apps, but I never found anything that truly helped. Based on my attempts and conversations with other people trying to do the same thing, I've come up with some general observations about the state of workplace knowledge management.

Modern collaboration is not scalable

The first observation is that both our current tooling and how we work today is not very scalable.

First, let's talk about tooling: every SaaS tool is their own source of truth. Let's say that you're developing a feature--you need to look at the spec in Google Docs, then try to find the conversation about the feature in Slack, and then go to JIRA to look at the actual ticket requirements, and then go off into Figma to look at the design. Each of these places is sure to have their own back-and-forth set of comments and updates. On a surface level, it makes sense that conversations go into Slack and designs go into Figma, but now the developer has to go to two places to figure out what's going on. As the number of specialized tools increase, the amount of fragmentation only increases as well.

The other part of collaboration that isn't scalable is simply due to people. When you add a new person to your team, now there is another person that needs to be kept up to date. That person not only has to communicate with everyone else, but they also have to check all of the same tools in order to understand what's going on. Fred Brooks in his book the Mythical Man Month famously talked about how adding a new person to a team can actually slow that team down. Many companies recognize this and cap team sizes (see Amazon and their two-pizza teams, for example.)

All of this makes knowledge management systems for teams incredibly important. But they're not that great.

Many existing knowledge management systems are too manual

One huge problem with most knowledge management systems is that they are too manual, making them hard to keep up to date.

Let's take the Google Suite as an example. Google Docs is a simple tool to use--you can just open up a document and type. However, each Google Doc is embedded into a folder hierarchy in Google Drive. Those folders and that hierarchy are usually a mess. Each folder may have a whole slew of documents, and discovering the correct document inside of that folder structure is a fool's errand. It's usually easier just to find the Slack message or email where it was shared.

Modern versions of Google Docs/Google Drive don't fair that much better. Notion and Coda are both immensely popular, and to their credit, they do bring many enhancements in the way that you can link documents together and surface pages in a wiki structure. However, they suffer from the same fundamental problem--when someone adds a document, they still need to figure out where it should go. Very often, these systems just go out of date since maintaining the folder hierarchy is a ton of work. An expertly laid out information hierarchy often goes out of date in a few months.

Some tools take manual knowledge management to the extreme. Tools like Roam Research or Obsidian expect you to tag and link everything inside of your notes and documents in order to provide value. This works for a small subset of people, but it is impossible for the general public to use effectively because they require too much work.

Many of these tools are super popular for tools-for-thought enthusiasts, so they end up with droves of happy users. Surely there is a way for these types of tools to be useful in a team context, right? Well...

There's a gulf between personal and team knowledge management

Highly successful people usually end up creating their own system for storing and retrieving knowledge. These systems are highly customized for each person, and they usually require a lot of manual work.

For a specific type of person who is willing to put in work manually tagging their information, tools like Roam and Obsidian give them outsized returns. The people who effectively use these tools have their own system of what they want to tag and when they tag things. They also know what they want to get out of the tool.

Similarly, there are Notion super-fans who love specific templates that let them organize their information just the way they want it. Maybe it uses a specific way of tagging pages, or maybe it relies on cards and the kanban views. Whatever the case, they've found a specific application that fits their mental model perfectly.

This often translates terribly into a team context. Everyone's mental models are specific and unique. One person may want to tag a page one way, but someone else doesn't think of it in the same manner. A system that works for one person doesn't necessarily work for someone else. If teams adopt a manual organization strategy, usually one person ends up being responsible for the knowledge base, and they are the only person who knows where everything should go.

Teams require a shared understanding of knowledge in order to collaborate effectively, and so shared knowledge often just decays to the least common denominator. The tools that have helped individuals just don't work effectively for teams.

The common denominator in most cases are chat apps like Slack, leading me to my next observation:

Chat has become the de-facto knowledge base

Chat apps like Slack started off as a faster alternative to email, but they've quickly grown and expanded over time. Now, many teams rely on tools like Slack to be their source of truth in addition to being a communication tool. If it didn't make its way into Slack, it didn't happen.

One reason why this happens is that there is a lot of knowledge that lives in the periphery of existing SaaS tools. When developing a feature, for example, there is the product spec, but conversations about the product spec invariably will involve the dev design document or the associated tickets. There is a lot of work that happens around existing knowledge artifacts, and that type of work tends to be in conversation form. These conversations go into chat apps like Slack and get stuck there.

Many teams that I've talked to have lamented this change. One VP of engineering told me that he was having enormous difficulties getting his team to document processes in their internal wiki. His team would have conversations every month about common issues like how to reboot their database cluster. These topics were available on the wiki, but they frequently went out of date and were not updated. Instead, people would ask on Slack and the same conversations would be re-hashed.

As chat apps subsume more and more content, their flaws become readily apparent. People complain of an explosion in the number of Slack/Teams channels, all of which may have updates, making it noisy and tough to keep up with. These systems expect you to be available all of the time, making it difficult to do focused work. And the info inside of chat apps is also highly ephemeral, scrolling off of the top of the screen and making it a chore to find something later.

This is a problem because chat apps are decidedly disorganized. That's why teams often turn to search. If only there was a great way to search everything, then you wouldn't have to put in so much manual effort in organizing and categorizing. Right?

Search by itself is not the answer

There are a ton of companies that are convinced that if you just search everything, then your team would be able to be on top of everything. That is just not true.

Search only works for topics that you know about. So much of team knowledge management is about making sure that everyone has the right visibility into information. If someone on a team never heard about a decision or wasn't in a meeting, how do they know to search for that information? The information discovery problem is unsolved with search, so it's not enough by itself to solve team knowledge management.

In addition, search (especially in a third-party tool) often has an issue with stickiness. A user may search for something right after installing the new hot search tool and then forget to use it the next time they may need it. Or, they may try it for a while and run into a few cases where the search tool did not give them the information they expected--and then they're done. Search only has a few chances to get the results right before people just give up and end up asking someone on the team who knows what’s going on or the place they should be looking.

So what is the solution to team knowledge management? Are we stuck? No, we are not, and I believe the solution lies in generative summaries.

Generative summaries are the way forward

The fallback for when someone can't find something in a team context (whether in a manual knowledge management system or through search) is to rely on that one person who knows what's going on. That person then compiles the knowledge they have in their head and writes a short answer summarizing the knowledge that the original person wanted to know, synthesizing together info from multiple sources.

When an effective team lead needs to make sure that everyone knows some specific topic, what do they do? They usually call a meeting and discuss it, and then they send out a summary of the topic discussed, along with the related resources that were discussed.

Both of these human processes result in a short, condensed response that synthesizes existing information and links to other team artifacts. This output is noticeably different from a search output or a specific document in a knowledge base. If we can generate these types of responses automatically by summarizing existing work and then sharing it out with the team, we can finally stop relying on manual processes to make sure that everyone knows what's going on.

These types of generative summaries are the way forward for team knowledge management. Teams will have AI-agents that listen to all of the work that's happening across every platform, and these agents can synthesize reports about what's going on and also create summaries on-demand when asked questions.

That's why we're building Maestro AI—we’re keeping teams in sync with generative summaries of everything going on in your team. Over the next few posts, we’ll share more about our vision for Maestro, and why we’re so optimistic that the next decade will see a revolution in what teams can accomplish together.