Keyword Spotting is designed to detect relevant keywords in spoken and written text. This article is about the spoken word – and specifically on the telephony channel. The complex data that travels over the voice channel in real time can be interpreted live by an Artificial Intelligence. The recognized topics can be used immediately, but also completely anonymized later for evaluations.
Since this technology has only recently been able to be used in practice, the areas of application are still partly unexplored – but for this very reason extremely exciting and promising. We would like to show you three examples of both real-time and historical evaluations that are already being used by our customers today.
With real-time recognition, customer dialogs take on a new dynamic. Predefined keywords are immediately recognized in ongoing telephone calls. If, for example, the word “supervisor” is mentioned, ECCO takes notice: It suggests a forwarding or automatically transfers the call. The following three scenarios show further possibilities.
Useful links and information on the tablet
In the contact center, every second counts. So if a service agent first has to search for information in knowledge databases, valuable time is always wasted. Wouldn’t it be much more practical if relevant links were suggested automatically? This is exactly what real-time keyword spotting can be used for.
For example, if the system hears that a:n caller:in is complaining about insurance premium adjustments, the AI automatically suggests the table with the current index adjustments and other information to the service agent. This way, the customer is immediately informed about how the price adjustment came about. This saves time and improves the customer experience.
Tips for cross- & upselling
Keyword spotting can also keep its ears pricked during sales calls. For example, if the keyword spotting agent notices that the customer has just decided to buy a fitted sheet, the service agent’s screen will immediately show that matching comforter covers from the same series are available as an offer. In this way, an agent never misses an opportunity to arouse maximum buying interest.
Checking compliance automatically
In a similar way to selling options, keyword spotting can also wait for certain core elements of a conversation to be addressed. For example, if the agent is required to point out certain privacy terms, the system can keep spotting until that topic has been discussed. And through automatic detection, it also immediately logs that all policies have been followed.
This capability of the AI system can also be used to ensure that new agents are trained efficiently at the outset. In this way, they can see right away in their first calls whether they are adhering to the expected call flow and all the points to be covered.
The possibility to draw helpful conclusions from the recognized keywords is partly linked to live spotting. Each recognized term can also be analyzed in its quantity afterwards. Of course, additional keywords can also be defined that do not trigger an event live and instead only affect statistics.
The most obvious benefit of quantifying addressed topics is, of course: comparisons. Which topic was addressed most often this month? What did the numbers look like the month before? In which projects were which topics relevant? So there is an automatic categorization of conversations – with the quality of this assignment going much deeper into detail than would be the case with manually selectable categories. Every mention, every detail and every cross-reference is counted. A corresponding weighting of keywords per project then enables detailed analyses at a previously unattainable level.
Linking with telephony data
Since the AI system is fed with information from the telephony channel, the keywords counted can also be assigned to a queue, hotline or project in retrospect. And this without drawing conclusions about individual callers.
Top keywords can be calculated in absolute or percentage terms and compared with the telephony key figures. Which keywords were used in calls with particularly long waiting times? At which time of the day did the number of positive or negative keywords increase? Did the topics discussed change with increasing call time? With flexible reporting, we enable you to draw exactly the conclusions that are important to you.
The key data mentioned so far allows you to use completely new types of metrics. Every change in the handling of callers can now be measured. For example, a question might be: How did call duration change for certain topics compared to the previous month with new strategies? Keyword spotting provides the answers fully automatically by paying attention in the background during the conversations in both months. The trend development in correlation with telephony key figures is available at the push of a button. You are free to introduce completely new measurable indicators here.
Keyword spotting works completely privacy compliant, as the keywords are taken from the live texting without customer information. This very texting can thus already be discarded at the end of the conversation.
Our keyword spotting is part of link|that ECCO. All mentioned features can be enriched by a sentiment analysis. Feel free to have a look at our product page to learn more.