With the coming of any new wave of technology–and we’ve seen a few in the 25+ years that Step Two has been in business–comes the hopes (or hype) that it will replace existing practices. Enterprise social, chatbots and digital assistants have all claimed at one time to replace existing ways of obtaining information. It’s now the turn for some to say that “AI will replace the intranet”.
The reality of course is more nuanced. AI provides truly new ways of finding and creating information, as outlined in ourAI framework for the digital workplace. This promises not just to make it quicker and easier to find stuff, but to enable new content to be created that draws seamlessly on existing resources.
Different tasks, however, will require different approaches. Some can be best addressed by giving a prompt to an ever-present AI assistant, some will need a simple search of existing documents, while others demand time to be spent on exploring and understanding a topic.
Therefore: browsing, searching and AI will all play a role in modern workplaces.
Busting myths
Let’s start by busting not one, but two related myths.
The first is that “some people always search, and others always browse”. This myth predates modern websites and intranets, and it quickly surfaces when search and information architecture are discussed. No evidence is ever provided for this statement, either from direct analytics from the intranet at hand, or from past research done by others.
In fact, Jared Spool–one of the true gurus of the UX space–once shared data that User Interface Engineering (UIE, now part of Center Centre) had captured for a series of usability tests on public websites. Firstly, this didn’t identify any users who solely searched or browsed. And when they dug deeper into the numbers, they instead found that the correlation was that the worse the navigation was, the more people searched.
The second myth builds on the first: “since most people search these days, and don’t browse, we don’t need to worry about coming up with a good structure for the intranet”. This rests on the belief that not only is there a strong findability preference held by individuals, but that it’s shifting definitely to one method, in this case search.
We’re now hearing the logical extension of this: “everyone will be using AI, so we don’t have to worry about navigation (or search)”.
Like the first myth, there simply isn’t any hard evidence that there is a clear trend in findability behaviour. In practice, this myth is actually an expression of hope that all the work needed to create effective navigation and search is no longer needed (phew!).
Types of search tasks
The field of user experience has long provided a more useful way to look at searching behaviours, and therefore how best to meet real-world needs. There are two search scenarios:
- Known item searching, where the user has a specific item they are looking for that they have good information about.
- Unknown item searching, where the user wants information on a specific topic or tasks, but doesn’t know how to narrow down to a specific item.
This is most obvious in a library context: it’s known item searching when you have an author and book title, and unknown item searching where you only have a broad topic you want to learn about.
With the emergence of GenAI, creating content can now be seen as an act of findability, where the AI searches for suitable source material on your behalf, supplementing its generalised ‘knowledge’.
In a digital workplace context, these concepts are applied as follows:
- Known item searching for a page or document, search is the most effective method (for example: finding the maternity leave policy)
- Known item searching for an answer, AI will increasingly be the best approach, in combination with search (for example: wanting to know the number of days of annual leave provided)
- Unknown item searching, navigation remains the most effective approach, exploring the intranet or other knowledgebase (for example: wanting to understand the structure of the company)
- Creating new content, AI provides an entirely new way of utilising existing information, without the need to actually search for anything (for example: creating a summary of key employee benefits)
To round out these discussions, it’s also important to recognise that employees aren’t just doing a single action to find required information. Instead, the ‘berrypicking’ model created by Marcia Bates in the late 1990s makes it clear that multiple techniques are used in sequence to find information, with later actions informed by the results of previous steps. AI should now be included in this model of behavioural patterns.
Ensure findability for all tasks
If searching, browsing and AI all have their role in the digital workplace, then steps must be taken to ensure that each method is effective, sustaining this over time.
For navigation and browsing to be effective, intranets (or other knowledgebases) must have a user-centric information architecture (IA) that’s been tested with employees. There has long been a best practice approach for this, outlined in Step Two’s seminal book Designing Intranets. This utilises techniques such as card sorting and tree testing, and the work required need not be onerous.
For search to be effective, the search engine must be tuned and configured to match the content being indexed. As outlined by Sam Marshall, most search problems are actually content problems , so it’s critical to have effective governance in place to ensure that content is owned and updated. Step Two’s Intranet Operating Model provides a best-practice approach for establishing scalable and sustainable governance.
For AI, governance becomes paramount. No deep-learning algorithm will be able to determine which is the ‘right’ answer when there are many versions of documents, which provide different information. As the scope of the AI increases, content management practices become important not just for centralised knowledgebases such as intranets, but also across document libraries and collaboration spaces.
The good news is this: much of the work that needs to be done benefits all findability approaches.
Effective governance benefits navigation, search and AI. Creating an effective information architecture provides valuable context to search engines and AI tools.
Take a pragmatic approach
The varied needs of employees are best addressed when teams move beyond ‘religious debates’ that pit one findability method against another. Instead, it must be recognised that all forms of findability have a role to play, and will continue to do so into the foreseeable future.
Challenges around governance and site structure won’t be resolved overnight, but pragmatic steps can be taken to improve the underlying aspects that empower successful search, navigation and AI. Use this opportunity to shine a greater light on this often-hidden work, and build support for better management of information that addresses all needs.
