If your business model involves communication with customers, voicebot can help reduce costs in this area by 2-3 times compared to solving a problem via call centre. The average cost of qualification for one lead by Voicebot is just 0.80 euro.

Implementing it you also save time and costs in lead generation campaigns on filtering out junk leads, e.g. our client - e-learning platform with more than 1 million students - ran an ad campaign and Voicebot helped them qualify 5292 leads and filter out 2500 invalid phone numbers, no-answers, and so on. Now, multiply these 2500 junk leads by the 5 minutes it takes to qualify each of them, and you’ll get 208 hours of work that could be wasted on nothing (this amounts to 4550 euro in salary by the way). Also, it is easier to control the costs with Voicebot as it allows for quick and predictable scaling without additional hires.

Voicebot helps to automate repetitive tasks, e.g. frequently asked questions or order status confirmation. It can also automate CRM/HRM system management and reporting. This prevents real consultants from burning out and saves their time for tasks requiring more problem-solving and creative skills, like handling difficult client issues or consulting on uncommon questions. Such an approach increases the performance of your team.

So, let’s imagine a situation, where a bank customer calls a support, complaining about their credit card, which is not working. There are different ways to talk about this issue:
- my card is not working;
- my card is blocked;
- I couldn’t make a payment;
- something is wrong with my card;
- why can’t I make a payment
and so on.
Voicebot can understand all these different variants, including interjections through intents. An intent is basically what the user wants or their intention. In this case the intent is the - ‘blocked card’. And all the different ways, in which the user can express it, are called the ‘training phrases’. In other words, they are the predicted phrases that the user may say. When the user’s utterance overlaps with, or matches the intent that the voicebot knows, then the voicebot knows how to react and what to say.

It all comes down to the process of intent matching, where AI based machine learning is applied. Machine learning mechanism compares the intent of the user with the intents that the voicebot knows and finds the best match. How does the matching process work? Machine learning algorithms calculate the confidence score for each intent that the voicebot knows. The confidence score is marked on a scale from 0 to 1. Where 0 means that the match is completely uncertain, and 1 means that the match is completely certain. The vb needs to find the user’s intents amongst all the variants. For this the threshold is defined in the process of the voicebot creation and testing. The threshold does not define how good the vb is in finding the variant. But it helps us to set how strict we are in letting the vb find the right intent. Opinions vary on what the threshold should be like, usually it’s from 0.3 to 0.7 depending on the complexity of Voicebot. So, again, In other words, machine learning classifies which intent that the voicebot knows is the most similar to the intent said by the user and this is how Voicebot understands people.