Why knowledge is critical for chatbots to be successful

By Aprill Allen on Nov 16, 2017

In a world where consumers increasingly demand instant gratification via their prefer communications channel, service providers in business must align to this expectation. Simply directing customers to a portal to complete a form, expecting them to send an email or presenting them with complex phone-based navigation options, delays resolution and puts pressure on support and service organisations.  Chatbots and Virtual Agents have emerged as key technologies in the fight against these more traditional support channels but this kind of automation requires a lot more thought and preparation than how it seems.

Chatbots and virtual agents are only as good as the information and the search functionality that underpins them. If you’ve planned, or provide, a customer-facing service portal, you may already be well on your way, because the same knowledge base your customers use for self-help is the same knowledge base your bot (or virtual agent) will search. Behind the scene, a chatbot may have a set of structured questions and answers driving its side of the conversation, or there may be more intelligent functionality such as natural language search and machine learning that contributes towards building the bot’s knowledge base over time.

 Either way, contrary to the fears people are expressing, this service automation won’t replace your human agents, altogether, but the nature of their work will change. They’ll do far less repetitious and reactionary work and more proactive knowledge work and relationship building; your workforce availability for improvement projects will increase; and the most important goal is to see your customer satisfaction increase with fast and effective access to self-service help.

Here are a few important things to think about before you jump into a long term commitment with your first virtual agent.

Customer context is king

Remember a customer may describe issues differently to how your technical teams might. So, when you’re designing your virtual agent’s dialogue flows or writing knowledge base articles, capture the words the customer used when they explained their problem and use these in the question title.

But, content still matters

Use plain language in your bot’s answers and keep responses short. If the answer involves steps your customer must take, 3 is a good limit. You don’t want to overwhelm them, but it’s ok to offer more information via a link to a full-size article or policy document.

Data cleanliness

If you’ve got an existing knowledge base, ensure your titles are meaningful to the customer (as above) and any knowledge asset classifications or descriptions are uniform and up-to-date. This metadata contributes to making answers findable for your chatbot or virtual agent.

Search isn’t set-and-forget

It’s common for people to forget that search engines need regular tuning to ensure they’re up to the task. Get your head into the config of your search tooling to see if there are ways you can improve it through altering weightings on fields to affect the results displayed, for example, or by adding new key terms and phrases to the thesaurus. Have a routine to check failed searches, as well, because this will give you clues about what answers your customers can’t find and which answers may need synonyms added. As an example,, provides the ability to track and optimise search performance.


Think about how your virtual agent looks from the customer perspective. Is it friendly and does it fit in with the branding of your organisation? Your bot’s approachability is influenced by the name you give it and the voice and tone it presents in those conversations with customers. So, think about the kind of personality that fits with your brand. Your marketing colleagues will be able to help out here and may be able to provide you with a style guide to refer to when you’re writing your dialogue flows and knowledge base articles.

Finally, starting with a holistic knowledge strategy that takes any organisational changes and training into account will help set you on the path to success. It takes a village to raise a good virtual agent, so think about the behaviours and processes that you may need to influence before you find yourself struggling with a lack of engagement on all sides.

To get more tips on how to start the knowledge base your virtual agent will need...