Convert CRM into a Knowledge Base automatically

Any business on the Internet must consider interaction with users as a fundamental piece of its customer experience strategyUsers ask, brands answer and vice versa, in a constant flow of information exchange. Now more than ever. However, we tend to think that the natural means of starting a dialogue is outside the web, through conventional channels or even the most recent ones, such as social networks.

But the conversation also takes place within the website itself where the search engine or virtual assistant is a key point of contact that the brand offers to help users resolve queries without adhering to the logic that structures our website, which, in the end, and after all, it is built on hypotheses that are not always fulfilled. If we want to deliver an optimal experience and avoid user abandonment, we must allow them to break the mould of our logic, adapting to them and not the other way around.

The great paradox is that precisely that freedom that we want to offer our users to express themselves naturally is a huge challenge for brands. We want visitors to our site to search and find, to express themselves in their next terms if they do not want to waste time browsing, to solve for themselves, if they prefer, with the satisfaction that this entails. Self-service, yes, but intelligent, which inevitably points to technology with basic cognitive skills and a robust Knowledge Base that serves as single or multiple sources of the response. Like one more employee.

What is a Knowledge Base

In this context, we understand by Knowledge Base any repository of answers, content or intentions that an engine uses as a source of search and extraction of answers to users’ questions. In reality, any stored information can be the germ of a source of knowledge in a more or less machine-readable or human-readable format.

And what company today does not have current or potential bases that can be transformed into sources of response with greater or lesser effort? Corporate documentation, articles, FAQ pages, support tickets, agent ratings and scripts, CRM tools, etc. Here the question is the effort that we can or want to dedicate to collect, write and centralize the knowledge spread throughout all corners of the company or even assess whether this effort is always the most practical solution.

We understand by Knowledge Base any repository of answers, content or intentions that an engine uses as a search source

A Knowledge Base created expressly for virtual assistance guarantees from the outset a greater orientation towards this type of human-machine interaction and better focuses the objective for which it was designed. On the contrary, this requires an initial abstraction and organization exercise and subsequent periodic maintenance. Although it is easier than it seems, not all companies have the resources for it or are at the same level of digital maturity.

What is often unknown is that gathering and centralizing corporate “knowledge” is not the only way. It depends on the use case, the choice of the technology provider and, as always, the end-user himself.

Indeed, the mission of the virtual assistant will largely determine the type of Base we need. If knowledge is very dynamic and changing, as in the case of e-commerce products, the most practical thing will be to automatically index the pre-existing content, while, if our assistant acts as a support agent, the effort to collect our policies and processes in a friendly and personalized way will be more than justified.

But, as I say, the user has the last word. We can set a mission and set goals that users will put to the test over and over again. If the analysis of real data justifies it, we will have to rethink the initial objective and if our assistant was born with the vocation of a support agent, we may also have to put him to sell because the user demands it. Therefore, some scenarios will require combining multiple sources of information, both automatic and manual.

If a company already has content stored on several systems, it should take advantage of them. Some applications will make it possible to use it without having to manually move or incorporate anything on a centralized basis.

Without the need for complex installations or technical analysis, these tools will allow to automatically extract content from external sources and structure it so that the search engine can use it, they will simply adopt the content without modifying it, offering it as a possible answer. In this way, CRM tools, such as Salesforce or Zendesk, as well as other web sources such as Twitter or Facebook or the corporate site itself can eventually become Knowledge Bases.

But is it enough to have quality content? Or perhaps the question is: Would this be enough in human interaction? Obviously not, we also need the response delivered to be accurate and relevant, the Achilles heel of much of today’s technology. The real challenge of the search experience is to understand the question, interpret the user’s intention that is freely expressed to display relevant results.

The language itself is already tremendously complex, variable, and ambiguous. Humans too. If a customer complains that their item “has arrived broken” or that it “has not arrived”, an intelligent engine must be able to distinguish intentions and display the “return purchase” process and the “package locator” respectively. The literal identification of patterns and keywords is insufficient to process the language, whether the key element is present (“has arrived”) or not (“Where is my article?”).

As if this were not enough, the rise of virtual interaction has broken spelling and has virtualized the use of new rules and words to which is added a turbulent time that has shaken our habits, altering our use of language. Consequently, the search engine must work at a deep linguistic level and rely on Natural Language Processing techniques, specifically NLU (Natural Language Understanding) to face this complex scenario with greater or lesser success.

When deciding on the search strategy for a site, make sure you choose a platform that allows you to manage knowledge, whether you decide to create it ad hoc, automatically index it or both and, above all, that has the ability to understand human language. , if we want them to repeat the experience.