Implementation of artifical intelligence services in cloud call centre agent panel

Většina call center potřebují k provozu speciální hardware a kanceláře. Agenti, kteří odpovídají volajícím na často kladené dotazy, musí hledat v souborech, které jsou mnohdy na různých uložištích. To způsobuje dlouhou dobu čekání pro volající. S novými cloud technologiemi je možné vytvořit call centrum založené pouze na cloudovém prostředí, bez jakýchkoliv drahých hardwarů. Agenti tak musí pracovat pouze minimálně díky webovým prohlížečům nebo Microsoft Teams. Naše oddělení už vytvořilo webovou aplikaci, ve které agenti mohou odpovídat volajícím a vidí aktuální transkripci a překlad hovorů. Toto řešení používá Amazon Kendra, který je však velmi drahý. Nedávno Amazon vydal nové řešení jménem Amazon Connect Wisdom. Cíl této práce je vytvořit proof of concept webové aplikace, která má stejné vlastnosti jako výše zmíněná, ale místo Amazonu Kendra používá Amazon Connect Wisdom.Most call centers currently require expensive hardware and office spaces to operate. Also, agents who answer customers' questions must manually search for answers in files commonly scattered across multiple locations. This prolongs answering time which, resulting in longer calls and thus larger waiting queues. With new cloud technologies, it is possible to create a call centre hosted purely in cloud environment that can be used in production without the need for any specialized hardware. Therefore agents can work remotely since they can answer calls via web browsers or even via Microsoft Teams. Our department has already created a Proof of Concept web application whose main purpose is to see if agents may receive real-time article recommendations, but it also supports call answering, real-time call transcription, and translation. This solution is working, but to provide article suggestions, it must use Amazon Kendra, which is too expensive to be used only for this purpose. Since then, Amazon has released a new solution called Amazon Connect Wisdom which is interconnected with Amazon Connect, and it can recommend FaQ articles during calls based on caller's questions. The goal of this thesis is to create Proof of Concept web application that has the same features as the application described earlier, but it must use Amazon Connect Wisdom instead of Amazon Kendra. It should also compare Amazon to other cloud providers and evaluate if Amazon is the right provider. The thesis has two parts. In the first part, NLP and its history are introduced. Multiple NLP cloud providers are evaluated, and an ideal provider is recommended. The second part focuses on building a web application for agents. Firstly, the architecture of the overall solution is described. After that follow chapters regarding infrastructure set up, cloud services configuration. The next chapter describes back-end and front-end code. Then review from cloud expert Matthias Schardt is noted and the thesis ends up with a conclusion if this PoC has been successful and recommendations if the project shall continue above PoC