Error message

Notice: Undefined index: content in clean2014_block_view_alter() (line 631 of /var/domains/
Right click to download pictures

The way businesses manage information today is broken. If you have ever sat on a committee discussing your company’s information taxonomy, you will immediately understand the issues at hand:

  • Team members spend weeks coming up with a list of terms that everyone is happy with (or at least, agrees to work with)
  • The taxonomy needs to be adopted by users and incorporated into their day-to-day behavior as they look for information and upload files to your document environments
  • Within weeks you notice that people are simply ignoring the taxonomy, data is not being stored in the right place
  • Staff begin complaining that, despite the changes, they still can’t find the information they need

After all the work you and your team put into managing organizational information in a logical way, the impact is limited. This is frustrating yet incredibly common. The reason? Put simply, the most common methods used for organizing business information today just don’t fit around how humans think. Let’s explore this problem in more detail and look at how topic computing could be the solution.

The limitations of our existing ways for organizing information

In a world where vast amounts of information are produced every day, the task of managing this information is unprecedented. Not only do your colleagues constantly generate content, they also store it in a wider range of places than ever before. Key records are held in email inboxes, in content management systems, SharePoint, Dropbox, company CRMs and more. The current approaches to managing this information, and helping people find the information they need to do their work, all have severe limits. To put it starkly, despite our technology, today’s knowledge workers spend 13% more time looking for information than they did in 2002.

1.   Taxonomies

As outlined above, taxonomies are usually designed by committee. Team members estimate the best way to classify the information you hold, by general themes—whether that’s based on the content of documents, or that they belong to certain projects, departments or a mix.

The problem with taxonomies: even the best taxonomies have a tendency to fail over time. Fundamentally, as your organization grows, and as your specializations or priorities shift, the old taxonomy will inevitably become irrelevant. The taxonomy is too rigid and forces people to apply labels to content that don’t really mirror what the content is about.

2.   Folksonomies

A folksonomy turns the concept of a taxonomy on its head. By allowing users to ‘organically’ tag work-related content and store it where they deem fit, and is an example of ‘collective intelligence’. It tends to result in a tag cloud, where users can naturally decide what a document is really about.

The problem with folksonomies: one major problem with folksonomies is that they depend on user participation. However, if only some users tag documents as they upload them, the approach can quickly become irrelevant. What’s more, if people add too many tags, it rapidly becomes hard to find what you are really looking for.

3.   Search

Search in enterprise systems has improved greatly in recent years, allowing for more natural search terms that a generation raised on Google feels comfortable with.

The problem with search: put simply, you’ll only ever find results with search if you know what you’re searching for.

Perhaps more important than any of these individual problems is the big issue of information silos. Today’s workers use ever more disconnected software which may contain related information, but which cannot be brought together. We believe topic computing is the solution to all these problems.

Topic computing enhanced with AI

The problem with all existing ways of searching for content is that it fails to fit technology and organize information around how humans think or want to work. Topic computing uses AI to take a new approach however, and offers the potential for a much more effective way of discovering information.

Topic computing works by using artificial intelligence tools, such as natural language processing, to help read through all the content on your existing environments and make decisions about what content is about. While topic computing is still a young concept, we expect it to keep growing in importance.

How it would work: a topic computing system would be capable of ‘reading’ all the content held across the diversity of your business systems. It would be able to understand the language used in your content and make assumptions about the importance of a topic based on the use of certain words in the context of a piece of content.

For example, say your company runs an annual conference called “Envision”. An AI system could use natural language processing to read through all your emails, CRM messages, CMS documents and more, to find content that includes that term. What’s more, it would be able to discover the difference between important “Envision” documents and uninteresting emails by monitoring how many times that piece of content had been opened or shared.

For workers, this would revolutionize the way they find and use information. By bringing them relevant content and guessing what they’re actually looking for, topic computing could allow individuals to discover information and links they were not aware of between different parts of the business. It would allow them to see the bigger picture. And it would replace the need for imposing arbitrary tags on content that soon go out of date. Instead, topic computing would always be up to date and relevant.

Our vision for topic computing

Topic computing may be in its early stages, but we are very excited about its potential. At, we’re at the forefront of research into topic computing. Collage is our innovative sidebar which sits in Outlook, and brings users contextual information based on the content of emails they receive. By categorizing information using natural language processing, it draws relevant content to the reader from SharePoint, Exchange and a multitude of cloud-based productivity apps, helping the reader understand the context of the topic they are working on. As topic computing evolves, we’re excited to be playing a part in its ongoing evolution.

To learn more about Collage, contact me, or read more here. 

Ram Tagher
Product Manager