AI saves the helpline agents’ time finding answers in seconds

Irina Kolesnikova
March 16th, 2022

An agent chooses the category of TitanCS agent assist tool

The state helpline 1247 is an information number operating next to the emergency number 112, which offers information and advice in Estonian, Russian and English 24 hours a day in situations where life, health, and property are not endangered. The senior specialists of the state helpline provide and collect information related to the environment, roads, rescue cases, and currently also the coronavirus. 

The service is extremely busy during the COVID-19 crisis: agents answer up to 1000 calls per day. The service needs AI assistance to get callers the right information as quickly as possible due to the situation with pandemic restrictions and quickly changing recommendations as well as the different preparation levels of the staff.

The problem: Information changes too fast and comes from too many sources

During the coronavirus crisis, public services providing information to people met several challenges. First, the situation has been (and still is) changing quickly: updates and governmental decisions appear often, are not regular and are sometimes suddenly. Second, many sources are providing this information: official websites of public organizations, governmental websites and many others.

Even well-prepared agents could be confused and overwhelmed with that avalanche of amendments and changes. In order to fix that, MindTitan created the 1247 agent assistant. To provide an analogy, the tool works like a tree catalog: starting from a general situation, it moves to more and more specific ones, until it delivers an answer.  However, the main goal is to allow the agents to just choose one right key phrase and get the answer, or to maintain the way through the catalog as a backup plan.

The expected outcome of the project

An AI-powered agent assistant, providing for a quick search of relevant information at hand and at one place. This tool will allow decreasing the training time of agents, at the same time increasing the quality of services, and thus making life better for people in Estonia.

The 1247 helpline assist tool in action
The 1247 helpline assist tool in action: an agent chooses the best answer much quicker.

The ASR and NLP models

  1. The main key phrases.
    First, when an agent of the 1247 helpline gets a call, the automatic speech recognition (ASR) model transcribes the caller’s request into written text, which is later classified using a natural language processing model that automatically offers topics of the call for the agent to confirm. To provide answers to the inquiries, MindTitan builds solution flows that guide the 1247 agents through the client dialogue, helping to ask the right specifying questions and give the correct answers. For that purpose, MindTitan data scientists analyzed over 600 call transcriptions. Six main key phrases describing topics for starting solution flows were discovered, such as “Vaccination”, “Travelling during COVID-19 pandemic”, “Isolation and close contacts”, etc.
  2. The further division.
    Those six general topics (aka general solution flows or, to put it simply—situations) then are divided into 50 subcategories. The further division goes with a similar scale, elapsing 3-5 levels of clarification averagely. To keep it as simple as possible and to provide information as swiftly as possible, some solution flows are straight and contain only 1 level.  
  3. The solution.
    When the agent receives the topic hint, the right direction can be chosen and the agent, either asking some specifying questions or not, quickly guides the caller to the answer or recommendation needed.

Conclusion

The main goal of the 1247 agent assistant is to perform higher efficiency of services with less human effort. 

The project is expected to be launched soon, all the algorithms are ready, functional, and are in testing. The MindTitan team is looking forward to getting the first feedback on the working system by the end of 2022. The AI-powered agent assistant for helpline 1247 is expected to save a significant amount of time for helpline 1247 is expected to save a significant amount of time for callers, reduce training requirements for agents and boost the quality of the information service provided.

Do you want to meet us?

Do you want to meet us?

Harry Liimal
HEAD OF BUSINESS DEVELOPMENT

Harry has years of experience working in business development, strategy, and digitalization for some of the world’s largest companies. He is a peoples’ person, enjoys fast-paced environments and multiple responsibilities. He is especially curious about transformation and disruptive technology.

Harry holds a Master’s degree in Law from the Tartu University and is acquiring an MBA at the Estonian Business School. He has also studied innovation management in Japan at Nagoya University of Commerce and Business.

Talk to us

Do you want to meet us?

Do you want to meet us?
Go back