Editorial
Imagine the situation when you logged onto your intranet and rather than being presented with a carefully crafted browsing space, you have a simple chatbot field asking you what you wanted to know about? We saw this when Google was first launched for internet searching. Could this happen again for intranets?
A Google Moment
Intranets have been with us for over 30 years. While much has changed about how intranets have been designed, built and delivered, there are some constants. The intranet is where employees expect to find authoritative information for their job, and where important company news can be communicated. It has attracted significant investment, aimed at making it easier for employees to use.
Generative AI has not been ignored by the intranet community, but rather seen as another adjunct to the main facility. It has even been regarded as providing a “facelift” for tired intranets. It can offer intranet editors and authors writing assistance and enhance intranet search with natural language queries. There are also claims that it can turn information into knowledge through improved search and summary capabilities.
But what would it take to achieve a “Google moment” for intranets? There would be no need for a carefully crafted browsing user interface, underlying information architecture, end user surveys and workshops, etc. In its simplest form, you will be creating your own “google moment,” when Google supplanted the flashing heritage websites with a single search window. Instead of a suite of ranked weblinks, the chat bot prompts you to ask a question, optionally engaging in a conversation as you guide it to the information you are looking for.
What Does 'Good' Look Like for Intranets Today?
Well-designed intranets have been regularly assessed and compared for intranet awards. Since 2001, Nielsen Norman Group has showcased the best intranets each year through its Intranet Design Annual Awards. The criteria includes:
- A good overall user experience, based on expert evaluation and user feedback.
- Innovative features or solutions that address specific business or user needs.
- Clear and consistent design, with a strong visual identity and branding.
- Well-organized information architecture, with intuitive navigation and search.
- A high level of content quality, relevance, and freshness.
These key performance indicators (KPIs) have been consistent for intranets.
NN/g expanded its criteria in its its 2023 Intranet Design Awards to include AI-powered features. It identified a rich mix of new capabilities to augment existing intranets. Many of the applications provided chatbot Q&A functionality to support employees in locating documents, hybrid/remote working, selecting collaboration tools, career advice and more. AI was also used to support intranet managers in retiring aged content, improving search results and tagging content topics.
The question is, will generative AI make many of these KPIs irrelevant?
For the first time in decades, Nielsen Norman Group has chosen not to provide a 2024 awards edition. It claimed that the organization needed more time to research and understand its audience in order to continue providing actionable insights. As experts in the field, are they seeing a disruption forming?
Related Article: The Best Intranets Today Are Swiss Army Knives
What Would It Take to Have an Intranet ‘Google Moment’?
There are many valid opinions as to why an AI chatbot can’t replace the current browsing interfaces that intranets provide. Let us look at some common ones and how they could be addressed:
AI Chatbot Reliability
Large language models (LLM), like ChatGPT, operate on probabilities. In areas that may have poor coverage, they are likely to “just make it up.”
However, hybrid AI chatbots designed to work inside the enterprise can be different. In this case, hybrid means that private content, i.e. intranet content, can be used as context giving supplementary information to the underlying LLM; it is not used to train it. If this supplementary information is prioritized over LLM content, the risk of unreliable or inaccurate information is minimized. Products like Microsoft’s Copilot Studio can provide references to your underlying document base. For an AI chatbot that will potentially be interpreting and summarizing formal policy documents, it will be important for the chatbot to reference these documents as the “official policy.”
This still doesn’t completely eliminate the potential for hallucinations. Air Canada discovered this the hard way when a customer challenged its decision not to take responsibility for its own chatbot’s advice.
Assurance and Credibility for Employees
Intranet practitioners have been mindful of the need for good governance. Nothing should change for an AI chatbot. The supplementary data source used by the chatbot must be also tightly governed. We have heard horror stories about how AI chatbots have unearthed unintended private information that may have been lacking the appropriate security settings.
Kurt Kragh Sørensen, IntraTeam Network owner, reminds us how Microsoft’s Delve search product caused embarrassment by exposing private files. Therefore, it is now even more important that intranet content is effectively and transparently governed for accuracy and authenticity.
Prioritized Security and Privacy
For a majority of organizations, intranet content is freely available to all staff. Intranet designers may build in “personalization” features like prioritizing local news articles based on geographic regions. However, this wouldn’t prevent employees from looking for news in other regions. If information on the intranet is being secured to specific job roles, this could be problematic for an AI chatbot to manage. Information should be secured outside the intranet/AI chatbot.
Integration With Other Enterprise Applications
It is preferential to embed intranets within business workflows. Digital work consultancy StepTwo has an award category specifically for intranets integrated into business applications. The intranet is commonly used as a portal to employee services provided by other applications, including booking leave, reporting safety issues, etc. The role of the intranet is to provide a convenient and singular place to access these services. Providing service information, along with the appropriate integration facilities, should allow employees to access required internal service applications directly.
Maintenance and Training for Bots
AI bots need to be maintained in the same way that intranets do. A good proportion of intranet maintenance is related to keeping content fresh, relevant and accurate. This same well-maintained content will ensure that the chatbot is up to date.
Conference
Sep
16
HR Tech Conference Las Vegas 2025
Register
Conference
Oct
27
Gartner HR Symposium/Xpo Orlando 2025
Register
We hear a lot about the requirement to “train” AI chatbots. However, in the case of intranet content and the hybrid AI approach, intranet content is not made available for training the foundation LLM model. Think of the foundation model as being responsible for creating the natural language functionality and “general knowledge” content. The supplemental intranet content will provide the context and content for the responses to be shaped by the LLM.
Replacing the intranet front end interfaces and underlying information architecture with an AI chatbot might appear a bridge too far for many intranet managers. A better way to think about it is that AI chatbots will provide them with more time for concentrating on the core facility of the intranet: the content itself. This way, we can better leverage the unique collaborative human capabilities for generating original and engaging content.
Learn how you can join our contributor community.
