IRENA: Voicebot and chatbot in the service of non-profit organisations helping refugees
The substantial refugee waves bring about the need to quickly react to various information and convey it as effectively as possible to those who need it. That is why Deloitte’s team for Intelligent Process Automation launched and configured a virtual communication platform IRENA that helps overloaded volunteers of non-profit organisations manage requests and respond to questions from Ukrainian refugees.
What exactly IRENA is capable of and how it specifically assists non-profit organisations was revealed in this interview with Jan Hejtmánek, Director of the RPA Team at Deloitte and one of the creators of the platform.
How can virtual communication platforms assist non-profit organisations, especially in a refugee crisis?
Czech non-profit organisations do not have much experience with large-scale refugee waves. In practice and under normal circumstances, there might be two women sitting in an office with two mobile phones answering incoming calls. Obviously, when hundreds of thousands of refugees come over the course of a few days, and they all need various information quickly, such a set-up is not sufficient to deal with the situation. In such cases, a technological, automated solution is required – in this case, a voicebot. It can handle some requests independently but is also backed by a call centre with human operators who can handle more complex requests that are too difficult for the voicebot.
How does platform IRENA work?
This platform combines three basic components. The first one is the voicebot and chatbot handling basic, repeating inquiries that we have mentioned. A second part is an online call centre with human operators who deal with questions too complicated for the voicebot or chatbot. And on the background of all that, there are systems to store and manage requests that help operators organise tasks and create a database of solutions to given situations, which can be used as a starting point for later assistance. All of it is online; volunteers do not have to install anything – they simply open an online application and make phone calls, take over and solve tasks, or edit the voicebot themselves if needed.
Immediate Refugee Need Assistance a.k.a. IRENA
IRENA is an omnichannel platform that includes a voicebot (a voice communication assistant) and a chatbot (a chat communication assistant), a call centre and tools for task management and planning, but also a knowledge base (a set of instructions and procedures). Non-profit organisations can use this solution during immigration crises to handle incoming questions and requests.
IRENA uses artificial intelligence, among other things, and can therefore adapt to users’ current needs. It is also scalable and available in a number of languages in addition to Czech, Ukrainian, and English.
How long did it take to programme the voicebot or the whole platform?
Creating the voicebot did not take too long, just a few days, essentially over the weekend. What makes communication assistants truly useful is their long-term development. By having us or a non-profit organisation take care of it on an ongoing basis, the tool gains added value.
In practice, the voicebot or chatbot encounters different requests. Those it knows it can resolve; others get delegated to human operators. In order to resolve more requests automatically, a person has to come at the end of the day, take a look at the overview of requests that were too complex for the AI and teach it how to resolve them, to put it simply. It is the same process as with a human operator – no coding is required.
That means that a repeating request that was not resolved automatically can be paired with a solution by an operator within the system – they add an appropriate answer, redirect it into another “tree” of questions and answers, create new sub-questions etc. This is how IRENA learns. And the more she learns, the more she can resolve alone without further assistance.
On what principle does the AI recognise what the user wants? Does it use some kind of keyword system to do this?
The voicebot and chatbot work on essentially the same principle; the voicebot just adds an extension in the form of spoken-word analysis in place of written text. However, the AI does not deal with keywords in either case but instead uses “intents” and “NER” – named-entity recognition. As the name suggests, what is important for the analysis of the inputs is not the presence of keywords only but also the identification of their meaning. We can imagine this process as sentence analysis that we know from school – AI must recognise the role of individual words in a sentence and adjust its reaction accordingly. And as we mentioned – the more a bot knows, the better it can deal with the given request.
When we received an e-mail from a non-profit organisation asking us to help refugees, we said yes, we can go and grab a phone. But we also thought about how we could make the process more effective. The solution was obvious – virtualise activities using a voicebot, a chatbot. And so, IRENA was conceived.
What makes IRENA unique, what does it bring compared to other voicebots and chatbots?
There are several differences, but three of them stand out. The first of them is its scalability which non-profits themselves can enhance. Deloitte delivers and implements the platform, and we provide all the necessary support. The deployment for a non-profit organisation only takes a few hours. But further development is up to the non-profit organisation that modifies the system to suit its requests and needs. What is most important is that they can do it by themselves; the platform’s interface is so user-friendly that our assistance is not necessary.
Another big advantage is connectivity. There is no problem to connect the systems and share data of two non-profit organisations using the IRENA platform. Organizations in Poland, Hungary, Czech Republic and Slovakia are already working together in this way. IRENA is not just for the Czech Republic, we have launched it for the whole of Europe because the immigration crisis is an European problem.
And last but not least, IRENA is highly versatile. It is not just a voicebot. It is not just a chatbot. It is a package of several tools in one. IRENA essentially needs no prerequisite; you just need to deploy the platform, and you can immediately start using it.
There is a lot of topics with which IRENA can help callers and chatters: residence legalisation, draft notice, school placement, contacts to the Emergency Services, visa and passports, what to do after arriving in the Czech Republic, how to travel to the Czech Republic, healthcare options, employment, public transport, and more…
What is the potential of the IRENA platform? Can it be used in some other way, or is it built exclusively for assistance in immigration crises?
IRENA was developed and is primarily configured for the needs of non-profit organisations in relation to the immigration crisis. Hence its name: IRENA stands for Immediate Refugee Need Assistance. However, its further possible uses are many – in state administration for instance. During the COVID-19 crisis, the need for the services of public institutions to be digital rose sharply, and IRENA is one of the platforms that can be used for this purpose. There are many questions and requests that can be handled through a voicebot or a chatbot. IRENA can be used by authorities but also hospitals for instance. Its operation is essentially the same as with non-profits; the only difference is in the content of the interactions.
There are many platforms for voicebots and chatbots. Every larger organisation has its own contact centre and technology running the platform. What we bring as an additional value is the automation of processes, connection of technologies and human work and support for process adjustment in such a way that volunteers can focus on what matters – helping people who need it.
The same applies to the private sector, to B2B. A number of companies deal with process optimisation, and IRENA is an excellent tool for automating some of them. Customers can typically contact a company providing services or goods through several channels: phone, e-mail, chat, e-shop interface, form, comment, and phone app… There are many options. And if all of those requests were to be handled by human operators, it would take a lot of time and be very inefficient. In such cases, the rule of “digital first” applies – answer to first contact digitally, automatically. And that is exactly what the IRENA approach can be used for.