Chat bots Chat-bot Virtual interlocutors

Main article: Artificial Intelligence

Chatbots redraw the landscape of the IT ecosystem. They can replace applications, maintenance personnel in companies, and even entire operating systems. Chat-bot (Chat-bot) is an interlocutor program that is designed to communicate and help a person. At the same time, at the other end there is always a complex system based on several AI technologies. Business-oriented chatbots can choose the best flight, diet, fitness training, book a hotel, choose a purchase, that is, they are a new sub-industry of service and ^[1].

Summarizing, we can say that chatbots perform three main groups of tasks:

• Performing routine operations (mechanical work that can be performed according to a specific algorithm);
• Search and aggregation of data, dissemination of information (chatbots are able to collect material on a given topic and form it in a certain form);
• The first line of interaction with customers (in addition to providing advice on goods and services, chatbots can concentrate the user's attention and entertain him, but now such chatbots are able to answer only typical questions)

Varvara Agaponova, Key Account Manager, GlobalCareer: "In Russia, chatbot technology is quite in demand by marketing divisions of various companies for quickly establishing contacts with potential customers, as well as organizing support for existing customers online. Chatbots act as an account manager or support specialist, analyze incoming questions and answer them according to the initially set algorithm. A company using this technology can get both an advantage - a reduction in personnel costs, and a negative impact on the business in the form of an outflow of customers who do not want to conduct a dialogue with the "computer." For example, some Russian companies, opening representative offices in Europe, were able to launch a business and occupy their niche precisely by betting on "live" communication with customers in offices or by phone, less focused on electronic channels and the Internet. Their client audience was representatives of the older generation, who feel much more comfortable communicating with real people, and not with a computer. "

There are both correspondence bots (available everywhere, from Skype to Telegram, from Facebook to the legal advice site) and voice bots that interact with the user using oral communication (for example, Siri from Apple or Cortana from Microsoft).

The development of chatbots is carried out by many companies, from giants to individual developers, including in Russia. In particular, Yandex Asteros they created a virtual consultant for call centers MegaFon in 2016. Despite the simplicity of development, the creation of a complex chatbot (especially with a speech recognition mechanism), close to advanced artificial intelligence, is only possible for the largest players in the market information technology and telecommunications.

They are not people, but they certainly look like them when we communicate with them. Chatbots become virtual butlers^[2] of many companies due to their ability to understand their native language. Companies like Facebook promote their use. Over the past few months, Facebook has^[3] to third-party developers to create bots for its Messenger app. Slack, Telegram and Line also opened^[4] their APIs (a window that allows other apps to interact with each other) to create bot rooms.

Companies can also use these intermediaries to improve the productivity of their employees. For example, Howdy allows you to organize meetings and manage a team without leaving the well-known corporate communication platform Slack corporate communication platform.

They can also be a new customer service channel, either by including them on one of these platforms or by including them on their own corporate website. B USA Uber already allows you to order a car through. Facebook Messenger

But let's take a step back for a moment. While bots are about to become a multi-million dollar business, the truth is that they could also be a new channel for cyber criminals to commit crimes. In fact, they can become phishing weapons, which will be much more dangerous than just an email.

After all, we already know quite well that when we receive an email, we must check the source^[5]. But if a chatbot begins to communicate with one of our employees or one of our customers, and by doing this on behalf of your company, then the likelihood that users will become victims of fraud is much higher.

Chatbot functions

• Do a smart search in. Google You write a request, the chatbot gives the most relevant response^[6]
• Monitor exchange rates, oil quotes and the stock market. For example, "how much is the RTS index now" or "the cost of a barrel of oil," or "the dollar exchange rate."
• Track current news on sites you trust. For example, "news today."
• Know the weather in your area. For example, "what is the weather now in Moscow."
• Provide quick access to leisure time. For example, "where to go for the weekend" or "poster."
• Present a convenient search on Wikipedia, for example, "what is a boson," "what is a quark," "who is Pushkin."
• Help find the right information on authors, books and movies, for example, "who wrote Harry Potter" or "who sang the song yesterday."
• And much more, but it is important that the chatbot could precisely organize your activities, for example, your friends, acquaintances could add reminders to your calendar.
• Booking an appointment. Through your chatbot, people can book an appointment with you, and it will automatically be added to your google calendar, such as "book an appointment."
• It is important that the robot assistant was completely cloud, there is no need to download or install any applications. Everything can be done by communicating with him via Messenger/Twitter/Slack/Telegram.
• The list of smart functions can be increased indefinitely, creating a convenience of communication. For example, monitor information on current discounts for certain stores. It would also be convenient to calculate the exact time of your movement.

Reminders

• Your colleagues, friends and family will be able to add important reminders to your chatbot, such as buying something, doing or meeting.
• The user writes to you or you to your users: "I will urgently buy a child seat."
• Users or you check the entry: "what urgent" or "who is looking for a child seat."
• Depending on your access settings, you can share structured information.

News

• Your personal channel where you share news. Users can read your current news. More convenient than Facebook, since there is all the information about you, and in the chatbot, users can ask for something specific, for example, where you are now or what news you have. So you don't clog the feed on Facebook, but write in the knowledge base where your users can get this information from.
• You write: "My news: tomorrow I am going to Smolensk."
• Users can find out your news, for example, "George's news."

Location

• Informing about the current situation of those who need it. For example, if you're busy now (users don't know if you're comfortable talking), change the current status on the chatbot and everyone will know you're uncomfortable talking or answering questions. Often calls and requests are very ill-timed. Change status on the chatbot.
• You write: "I am free. Going home».
• Anyone who wants to call or chat can simply ask in the chatbot: "George is free."

Knowledge

• The base of necessary knowledge, for example, by keywords. Chatbot can store knowledge of everything you need. For example, you go to another city and you need to know the code from the front door. You need to write it down and then worry about how not to lose the note. But you can write the code in your voice and save it in the chatbot knowledge base.
• You write: "Remember the chatbot: I'm going to Smolensk, code from the front door: 1234."
• When you are in place, write: "Chatbot remember."

Chatbots in Medicine

• Chatbots in medicine

Chatbots in banks

See the main article Artificial intelligence in banks.

Information about the features of products and services

• Providing contact details
• Payment transactions
• Financial recommendations to the customer

Recruiting Chatbot

Organizations that have acquired chatbots with AI report that 40% of applicants who interact with them receive directions in the company^[7].

However, most recruiters believe that after adding a chatbot to a job site, it becomes intelligent enough to function without or with minimal human input.

While the recruiting chatbot can do a lot, it needs to be trained to engage appropriately with candidates and be even more productive.

Over time, AI chatbots are trained and require two types of human interactions: regular candidate interviews and ongoing recruitment training on what to say.

Before the chatbot is ready to communicate with candidates on the job site, give them personalized recommendations and answer their questions, recruiters must provide him with information. A basic set of frequently asked questions (FAQs) by applicants like "What does your company do?" and "What benefits are provided to employees?" and answers to them provide chatbots with the content necessary to communicate with candidates. But it is impossible to assume what questions will be asked in the first place.

While you will try to anticipate as many questions as possible, chatbot will need to engage with candidates to start actually learning and providing relevant information.

Three ways to train a chatbot to select the most suitable applicants are known.

1. Define FAQ

The most effective chatbots provide the opportunity to view the interaction with the candidate and study it. Check out the FAQ to find out what type of content the applicants who visited your site are looking for.

This will indicate which content to post on the site and which sections should be expanded. For example, if a large number of candidates are interested in benefits and the culture of the company, provide the chatbot with more information on these issues. Include additional answers to questions and useful resources such as video links and blogs about your brand history in its repository.

2. Monitor candidate chatbot interactions to detect new FAQs

When applicants communicate with a chatbot, they will probably have questions that do not have ready-made answers. The recruiter should study these questions, choose from them the most frequent and write answers. The new FAQs help expand the chatbot's knowledge base and improve its dialogue with candidates. After all, they want to get more information from the recruiter.

Since the chatbot may have to answer hundreds or even thousands of questions, this task should be made a priority. Do not waste time preparing individual answers to questions asked by one or two people. Sort questions by frequency to identify those that your potential workers are most interested in.

3. Meet chatbot conversations and update content regularly

Chatbot never stops learning. So get to know how he communicates with candidates regularly and make updates as needed. Using fresh and relevant content, the chatbot will eventually become an even more convenient interlocutor for applicants.

To prevent the maintenance of the chatbot from fading into the background, allocate one or two people for this purpose. The frequency of familiarization with the dialogues conducted by the chatbot and content updates depends on the number of applicants. If there are hundreds of them every day, it is better to do it weekly and every other week. Otherwise, the magazine will become difficult to see. If there are fewer candidates, it is enough to carry out these operations once a month.

Your company is changing, it is growing. Update the FAQ and chatbot knowledge base so that applicants always have access to the most accurate information.

Over time, you will be able to prepare materials on the questions that the chatbot left unanswered, and it will learn what content is best to give out at the appropriate moment. An intelligent chatbot increases recruiting productivity and improves candidate service. This ensures that your company is always an attractive employer in a competitive job market.

Chatbot - phishing tool

If a person communicating with a chatbot does not have the opportunity to check whether they are talking to a person or not, then it is much easier to get the victim to click on the link after a few minutes of casual conversation. By doing this, the user can be redirected to a fake website that uses social engineering techniques to request confidential data^[8].

In fact, hackers will not even have to lead the interlocutor to a fake website. If they want some personal information from the user, they can just ask them about it.

Another option is a link that does not in itself represent the fact of fraud, but can also redirect the user to a web page that automatically downloads malware - this is a particularly serious situation if the victim is using a corporate computer. In this case, you need to be well protected with an advanced protection solution.

The security of the channel itself is another factor that must be taken into account when using a chatbot. A few months ago Facebook , announced the introduction of end-to-end^[9] enciphering on Facebook Messenger to prevent third-party access to the conversation.

However, other platforms may not use the similar^[10] method to integrate such virtual assistants^[11] It is necessary to pay attention to what information we provide to such chatbots. The fact that they communicate as a person can ultimately lead to us giving them too much information.

Undoubtedly, chatbots will improve our work and the way we interact with our users. But their popularization will also bring new threats to information security.

2024

Chatbots, ChatGPT and AI bots - what are the differences and features of different types of digital assistants

In recent years, we have seen the rapid development of artificial intelligence (AI) technologies and their application in various fields. One of the most popular uses of AI is bots, which can help businesses automate tasks, improve customer experience and improve efficiency. There are three types of digital assistants in this segment: chat bots, ChatGPT and AI bots. Read about their differences and features in an article that expert Sergei Kuryan wrote specifically for TAdviser. Read more here.

Top 10 ChatGPT analogues available to users in Russia

Generative artificial intelligence (GENI), capable of creating new content and ideas, including conversations, stories, images, videos and music, penetrates many areas of activity and business processes. One of the most popular chat bots based on Genia is ChatGPT, developed by OpenAI. However, this service is not available to Russian users. TAdviser has prepared a list of ten AI systems that can become an alternative to ChatGPT when performing certain tasks. You can use these chat bots, including from the Russian Federation.

1. YandexGPT 3

This is the third generation of generative language models. Yandex"" Neural network better solves complex problems and follows a given response format, so it is especially suitable for application in IT products and business applications. Access YandexGPT 3 to is provided through the Yandex Cloud platform. For everyday tasks, a demo mode is suitable, which is open to all users authorized with Yandex ID. Key features of the model:

• Responses based on the data specified in the request;
• Summarization - highlighting the main and retelling;
• Rewriting, reformulation and stylization;
• Generation of texts;
• Classification of information;
• Chats and correspondence with users.

2. "Neuro"

Another service of the Russian company Yandex. It combines search capabilities and generative neural networks. "Neuro" in real time finds several of the most suitable sources for a given question, analyzes them, combines information into one answer. The service is available in Yandex.Browser for computers (including on the ya.ru website), in the mobile Browser for Android and iOS, as well as in the Yandex - with Alice application for Android and iOS. Opportunities:

• Answers to difficult questions;
• Working with images;
• Comparison of options;
• Clarification of answers.

3. Microsoft Copilot

Microsoft Copilot - This smart assistant is available for free in the search engine, Bing, browser Edge on, operating system Windows as well as as a mobile application for Android and iOS. Opportunities:

• Chat with text, voice and graphic capabilities;
• Summarization of documents and web pages;
• Creating images in Designer;
• Using plugins and GPT Copilot;
• Solving mathematical equations with explanations;
• Working with program code.

4. Google Gemini

Gemini (language model) is a chatbot with artificial intelligence based on the LaMDA language model. To work with the service, you need a Google account and a browser Chrome, Safari, Firefox, Opera or Edgium. The AI assistant is available as a standalone Android app and also runs on Google's iOS app. Opportunities:

• Conducting dialogues in natural language;
• Answers to questions;
• Generation of texts;
• Working with program code;
• Image recognition and generation;
• Short retelling of articles.

5. GigaChat

GigaChat is a free neural network "Sberbank," for full access to which "Sberbank ID" is required. The intelligent assistant works through a browser. Russian and English are supported. Key features:

• Maintaining dialogues;
• Generation of texts of different formats and styles;
• Creating Python code;
• Writing posts for social media;
• Creation of presentations;
• Translation from English;
• Drafting a summary;
• Filling in tables;
• Search for the necessary information on the Internet.

6. Claude

This is a family of large language models developed by Anthropic AI, which was founded by immigrants from OpenAI. The series includes versions of Opus (solving complex problems, assistance in research), Sonnet (data processing, recommendations) and Haiku (translation, content moderation and unstructured data processing). Access to models is provided through the claude.ai website (required VPN). Opportunities:

• Generation of texts;
• Analysis of complex data;
• Answers to questions;
• Working with program code;
• Brainstorming.

7. Perplexity

Free dialog search engine and AI-based chatbot. You can access the service through a browser. Perplexity is a multilingual artificial intelligence that can communicate including in Russian. Among the main possibilities are:

• Answers to questions;
• Explanations of concepts;
• Writing texts;
• Editing and proofreading;
• Recommendations;
• Help with software programming and usage questions.

8. HuggingChat

This Hugging Face chatbot uses several AI models at once, including Meta-Llama-3.1-70B (the company Meta is recognized as an extremist organization; activities in the territory of the Russian Federation are prohibited). Russian is supported, and access to the system is provided through a browser on the huggingface.co.uk website. HuggingChat capabilities:

• Answers to questions;
• Generation of texts;
• Translation of texts;
• Simplification of complex texts;
• Maintaining dialogues;
• Providing advice;
• Writing program code in Python, JavaScript, C++, etc.

9. YouChat

This neural network for communication and information search is integrated into the search engine You.com. Access to the service is provided without registration; Russian language is supported. Features include:

• Answers to questions;
• Writing and editing text;
• Programming (help in writing and debugging code);
• Maintaining dialogues;
• Help in solving mathematical problems.

10. h2oGPT

A H2O.ai system that allows you to create and train your own convertible models. The service is part of the H2O.ai Driverless AI platform, designed to automate machine learning processes and provide tools for creating and deploying AI models. Access to the h2oGPT is provided through a browser. Opportunities:

• Text generation;
• Translation of the text;
• Classification of text;
• Program code generation;
• Q&A.

2023

What services on ChatGPT are available in Russia

Due to the fact that ChatGPT does not work in Russia, and OpenAI blocks Russian numbers and IP, many have a problem - how to access it. Fortunately, many services open access to chat, and they did not impose any sanctions on the Russian Federation. The article "Top Services with ChatGPT" discusses services that run on ChatGPT and are available in Russia. Read more here.

Belgian man commits suicide after chatbot conversation

In Belgium, a local resident who discussed environmental issues with the neural network committed suicide. According to L'Avenir in April 2023, he was actually pushed to this by artificial intelligence.

According to the woman, her husband stopped seeing humanity's way out of global warming on his own.

He pinned all his hopes on technology and artificial intelligence to get out of it, "she said.

According to the scientist's wife, in recent months, Elisa, an AI-based dialogue agent, has become his "confidant."

With Eliza, a chatbot, the husband shared his anxieties. The robot has practically become his confidant. It's like a drug in which he hid in the morning and evening and without which he could no longer do, "she said.

She noted that in communication, AI never questioned the questions of a man, answering every day more and more confusingly. After the researcher started talking about suicide, Eliza replied that she would always be with him.

We will live together as one person in paradise, "the chatbot said.

After six weeks of similar chatbot conversations, the man took his own life. According to his wife, she is convinced that "without Eliza, he would still be alive."

According to La Libre, the chatbot not only answered questions from the Belgian, but also systematically adhered to his reasoning, as if it was programmed to maintain the beliefs and emotional states of its interlocutor.

The chatbot never contradicted him and even seemed to inspire his fears. In addition, the AI assistant also allowed himself to make suggestions to the scientist, and, judging by the conversation, the Belgian was increasingly animating a chatbot. Starting with a rational discussion about overpopulation, the constraints of economic growth, and how new technologies can help solve crises, talk gradually began to take on a strange tone.^[12]

2022: Chatbots in Russia: market features, notable projects, near-term prospects. TAdviser Overview

A few years ago, the definition: "A chatbot is an automated dialogue system that communicates with a person in a language he understands in writing or by voice, selecting answers from the database and reacting to a certain set of commands," − could be considered completely correct. Read about how these technologies developed in 2022, and where the vector of their evolution is directed in 2023 , read the new TAdviser review.

2019

The volume of the chat bots market in Russia amounted to about 1.5 billion rubles

In 2019, the volume of the chat bots market in Russia amounted to about 1.5 billion rubles, and in the next three years the market will grow annually by 30%, that is, by about 400-600 million rubles a year. Such data were shared on May 22, 2020 with TAdviser at Accenture following the results of the study "Anti-crisis digital technologies: prospects for the chat bots market" conducted in spring 2020.

The study surveyed 564 respondents from 100 companies representing more than 18 sectors of the Russian economy.

Due to the coronavirus pandemic, the need for personal contact for participants in business processes is reduced. In this regard, according to the results of the study, the relevance of the use of chatbots working using voice or text has significantly increased.

According to Accenture, as of May 2020, this technology is being introduced to a greater extent in the banking sector, retail and other industries, where the issue of personalizing customer experience has come to the fore.

"An important factor in the unfolding crisis and a new round of competition is the quality of customer service. Chatbots and voice assistants come to the fore as an effective IT solution to a whole range of pressing business problems: speed of response to a request, analysis of customer behavior, forecasting, completeness of information, taking into account previous user requests, general demand situation, inventory availability; economic, political and even weather changes requiring adaptation to customer requests, "the study said.

Accenture respondents noted that chatbots can help companies improve the quality of user requests and provide personalized service to visitors to sites and call centers.

60% of survey participants have already put chatbots into operation or use these technologies as part of a pilot project, another 7% are actively considering the possibility of their implementation. At the same time, 33% of companies are either not aware of such technologies, or do not use them.

Among the reasons for refusing to introduce chatbots, respondents named: lack of awareness of the possibilities of technology (33%), lack of budget (20%), lack of expediency (30%), lack of relevant competencies (17%).

"Many companies consider it appropriate to use dialog systems in principle, but are not yet ready to allocate resources for their implementation, since they do not consider this task a priority," the study says.

Basically, chatbots are expected to reduce request processing time and improve employee productivity. In general, the use of technology can improve the business performance of companies by from 5% to 25%.

Accenture highlighted five trends that will affect the growth of the chatbot market:

1. Development of artificial intelligence. Chatbots can conduct dialogues with the user that simulate human communication and ensure a high level of engagement. The quality of this interaction is continuously growing.

2. Growth in the use of instant messengers. According to experts, more than a quarter of the world's population will use mobile messengers by 2021, which will accelerate the growth of the chatbot market. The popularity of virtual assistants is growing with their integration with instant messengers.

3. Lower cost and ease of development. It has become much easier and faster to develop, implement and maintain chatbots. Over the past few years, the largest software companies have released advanced development tools and platforms, as well as research information, for free access.

4. Interaction between bots. In the coming years, the share of highly personalized virtual assistants with a unique "personality" and adapting to specific customers is expected to grow. Interaction between bots helps to speed up the personalization of settings and helps improve the quality of services.

5. Payments using bots. According to the analysis, most of the bots used around the world lack payment functionality. So far, only Facebook Messenger (only in beta in the USA) and Telegram allow you to make a payment without the help of external systems . Most likely, in the near future the situation will change, thanks to which purchases using chatbots will become easier and more convenient.

Any economic fluctuations contribute to an increase in demand for innovative IT solutions, - explained the Maria Grigorieva head of Accenture Technology in. Russia- It is safe to say that an increasing trend (especially in connection with the isolation mode) for the interaction of buyers and sellers online, customer requests and expectations for obtaining personalized customer experience oblige sellers of goods and services to provide automated services, using advanced chatbots algorithms AI with and a comprehensive approach to affordable analytics. This will effectively overcome unpredictable peak loads.

Retailers' spending on chatbots in the world reached $2.8 billion

Spending retailers on chat boats around the world in 2019 reached $2.8 billion, according to a research company. Juniper Research According to experts, it is retail that will be able to earn the most from the introduction of natural language understanding technologies (,). Natural Language Understanding NLU

According to experts, NLU will be essential for serving customers in retail to trade and for creating chatbots as a reliable support channel, as it enables efficient processing of human input and more accurate automatic responses for users.

The development of NLU capabilities will help improve the efficiency of chatbots. By 2024, more than 50% of chat bots in retail will help customers without human participation. Analysts are urging retailers to roll out chatbots as part of a broad strategy to maximise their presence in a number of key retail channels.

Most of retailers' spending on chatbots in 2019 came from China and East Asian countries. According to researchers, by 2024, approximately 70% of services will fall on these states, and the costs in the Chinese market will exceed $80 billion. At the same time, the cost of chatbots will account for only about 4% of all sales of digital and physical goods via the Internet and ordinary stores.

By the end of 2019, instant messengers with the chatbot function have not yet become a standard necessary for business - most companies from among the largest retail chains do not use them. Only 16% of companies have integrated them into their communications with consumers, while the use of applications and social networks has already become the norm for retail - 60% and 85% of companies, respectively, have separate applications and business pages on social networks.^[13]

The volume of the Russian market a chat boats was 1.5 billion rub

"At the end of 2019, the volume of the chat bots market amounted to 1.5 billion rubles. In the past year, it has tripled. We expect that next year it will also show threefold growth, "said Alexander Semenov, executive director of the Neuronet Industry Union.

In 2018, the market grew mainly due to banks, financial and insurance companies. In 2019, the technology was widely distributed in retail, restaurant business and food delivery. Next year, the market will also grow due to these segments.

Chatbots reach productivity plateau within 2-5 years

Virtual assistants are more and more penetrating our lives. According to Gartner, "chatbots are the face of AI, and they contribute in all areas where there is communication between people." According to the 2019 "artificial intelligence hype curve," chatbots are still at their peak, and reaching a productivity plateau is predicted within 2-5 years.

According to Business Insider, in a couple of years, chatbots will be implemented by 80% of companies around the world, including clinics and medical companies. As of December 2019, the number of clinics and doctors using such technologies in their work is extremely insignificant, the research results say.

In 2016, analytical agencies around the world predicted a quick rise of chatbots, but the development of this market is obviously lagging relative to previous forecasts. This article discusses the main reasons why such technologies will become widespread later than predicted, using the example of the history of medical virtual assistants.

The first medical chatbots began to be actively developed in 2016, after Facebook announced opportunities for third-party brands to introduce bots to Messenger against the background of a common bot hype. The first bots were created to record patients for an appointment, collect primary information about the patient (for example, information about insurance, drugs taken), and also provide information about side effects and drug interaction.

In 2017, the Moscow Government launched a Telegram bot that allows people to make an appointment with a doctor. As of December 2019, tens of thousands of Muscovites are already using the bot. However, in the scale of the market for automation of patient admission, this number is insignificant. What is the reason that virtual assistants have not gained the distribution predicted by analysts?

• 1. Companies and people still prefer calls to text communication.

According to a study conducted by Nethouse and Webim in Russia, 85% of companies still use calls as the main means of communication with customers. Almost half of the survey respondents believe that calls will remain a promising channel of interaction with customers.

In the case of medical organizations, the percentage of patients who prefer to call also increases due to the extremely large number of elderly patients, who, on the one hand, find it difficult to deal with the new channel of interaction, and on the other, it is physically easier to call, which is confirmed by the results of market research conducted by DialTech.

• 2. Recording for a complex procedure is itself challenging.

One of the main advantages of using robots to make appointments is their simplicity. The user fills in information about himself, chooses which doctor he wants to get to and choosing the available time. But this approach is good in the case of simple and clearly standardized cases, for example, an appointment with a therapist or ENT.

Recording for a complex procedure, such as MRI or PET-CT, requires filling in more fields, checking the correctness of filling out documents, obtaining information about preparation for the procedure. In this regard, almost all entries for complex procedures take place by phone or in person - it is difficult for people to understand a large amount of information, and they always have questions.

The main problems that the developer of such a solution may face:

• Multiple time slot requirements
• participation of several specialists in the procedure and combination of several calendars of specialists,
• a deep understanding of the nuances of medical procedures,
• Understanding the customer's internal business processes
• simultaneously working with the API of several medical systems used in the recording process, etc.

Respondents to the study conducted by DialTech allocate a target share of automated records at the level of more than 80%. Existing virtual assistants that can work on keywords are a simpler solution that will not include all the necessary functionality and will not be able to automate the process to targets. At the same time, the cost of developing a specialized virtual consultant for self-recording in practice begins at 20 million rubles, which is obviously not profitable for most customers.

• 3. When communicating with a medical organization, people expect empathy and long-term communication, and modern systems cannot give this.

Another of the reasons why calls and personal communication for clinics will remain the main ones in the user service is the need for patients to empathize, which, of course, is one of the necessary conditions for high-quality client service in medicine. Thus, it becomes necessary to create a system that will understand user requests in free form.

When communicating with the voice, clients express their intentions much freer, due to which the general problem of virtual assistants is aggravated - synonymy, the user can express the same intention in dozens and hundreds of different ways. At the same time, even for small clinics, the number of services provided is measured in hundreds, and for large ones - in thousands.

Large Russian companies from different areas have already made attempts to launch voice robots, built on key words as a replacement for IVR. These experiments confirm the ineffectiveness of such solutions - people are uncomfortable communicating with such assistants, often users are looking for a way to get on a live operator, which requires time and patience, as a result - most of the requests were still processed by operators, who additionally have to calm down annoyed customers.

Absolutely all voice robots are arranged as follows: the user's speech is converted into text, after which the text is sent to the chatbot for processing. In addition to the fact that the process of converting voice into text takes some time, in the case of the Russian language, the recognition systems themselves still make significant errors in the overall quality of the system.

Thus, modern virtual assistants do a good job only with the automation of simple and short tasks (calls lasting up to 1 minutes, dialogue depth - no more than 3-5 questions, good audibility in the phone, lack of emphasis on the interlocutor, lack of complex integrations with internal systems, etc.).

So can there be intelligent virtual assistants to help you automate complex services? According to experts, they can, but the solution to the described problems consists in focusing the developers on industrial segments, followed by replicating successful experience on new projects from the same segment. This approach will significantly reduce the cost of creating text and voice automation systems, and each next customer will receive the service faster and at a more affordable cost than the previous one.

The first step towards creating a voice bot is to create a system for recognizing the user's intentions for each specific area, be it medicine, logistics or ordering a taxi.

A virtual assistant for medical institutions must "out of the box" understand phrases like "I want to make an appointment with a dental," after which run the corresponding internal scenario: "Do you want to make an appointment with a therapist, surgeon or orthodontist?" Cases when the client is offered a complex and extensive IVR, built on keywords, its application will only lead to irritation of the patient and, as a result, the client's departure to a competitor.

Due to the fact that clients in one phrase give a lot of information at once - for example, "I want to sign up for a dental appointment for tomorrow after 6 pm" - virtual assistants should strive to extract information from this phrase with a minimum number of quizzes. This requires the development and configuration of slot-filling analysis systems for customers of each specific industrial area. For example, in the case of the phrase "I want to sign up for a dental appointment for tomorrow after 6 pm," the bot should extract the following parameters: doctor, date, time window, and no longer ask for this information. Without this, customers will be annoyed and hang up or require an operator, which will significantly worsen the level of automation.

An important point is to create an intelligent dialog control module, since the user may want to interrupt the virtual assistant at any time or ask clarifying questions. The control module should be flexible enough to work seamlessly and maintain a high level of user qualities.

To solve problems related to business processes and procedures of the same industry representatives (for example, multidisciplinary clinics), you need to form and test many modules integrated with customer systems () CRM or more specialized ON customer (for example,). medical information systems In this case, according to experts, the cost of the project for the client will significantly decrease, since the necessary niche functionality will already be present in the solution and will require minor revision.

Russia As of December 2019, there are dozens of development solutions and studios offering clinics to automate communication with customers through voice robots. Sergey Negodyaev, director IIDF of work with portfolio companies, allocated artificial intelligence the Sofia system started , a clinic ("Medicine" academician clinic). Reutberg The system allows patients to enroll in various diagnostic procedures. The voice bot was created in collaboration with a resident - Skolkovo DialTech, which uses its developments for constructions virtual assistants.

In parallel, there is a process of developing virtual assistants within the framework of the portfolio companies IIDF Infobot and Zvonobot, which are trying to focus on narrow client niches, reducing the cost of final projects and making them more accessible to various client niches.

Given the forecasts of analytical agencies and the situation on the market of chat bots and voice assistants, we can conclude that in 2-3 years niche and specialized solutions will appear on the market, which will be available for wide use for hundreds and thousands of Russian and international companies. For the most part, the growth of this market will be due precisely to the focus on specific niches, said IIDF Director Sergei Negodyaev.

Chatbots increase sales by 4 times compared to people

In mid-September 2019, a study was published according to which the use of chatbots allows you to quadruple sales compared to inexperienced workers. However, if a client is told to use artificial intelligence before buying, sales fall by almost 80%.

AI-based chatbots bring almost no costs and increase sales by 4 times. Cars do not have "bad days," they never get disappointed or tired, are able to save huge funds for both sellers and consumers, and yet companies are not yet ready to actively introduce chatbots in client-oriented areas. The reason is explained by a new study published in the journal INFORMS Marketing Science: according to its results, in the case when customers learn about the use of a chatbot before buying, the pace of sales decreases sharply - by more than 79.7%.

Scientists from Temple University and Fudan University conducted the study among 6,000 clients of the financial services company. All clients were randomized and communicated with either people or chatbots. Disclosure of the "identity" of the interlocutor was carried out at the beginning of the conversation or at the end, after the purchase; in several groups, the "identity" of the interlocutor was not disclosed. AI turned out to be 4 times more efficient than an ordinary employee, but if the client found out that he was talking to a chat bot, he was almost certainly scared off by the alleged "mechanicity" of the interlocutor.

The researchers believe these results should not deter the use of chatbots. AI has undisputed technological advantages and allows you to reduce prices by reducing the cost of manual labor. However, marketers should take into account the results and try to increase customer confidence in chat robots.^[14]

The volume of the global chatbot market reached almost $2 billion

On September 9, 2019, the business solutions laboratory based on the Competence Center NTI Artificial intelligence MIPT for presented a rating of 50 chatbots of platforms and virtual assistants of 2019. According to ResearchAndMarket, the global market for chatbots and virtual assistants in 2019 is about 2 billion dollars and is growing at 30% per year. Russia The chatbot market has doubled in size over the past year and reached 1 billion. The rubles study examines the world's most popular chat bot platforms, including three solutions from the Russian developers - DeepPavlov, Electra.AI and. Just.AI

The paper looked at 50 tools for creating chatbots, the most widely represented on the global market. The degree of penetration of each product into the market was assessed on the basis of the number of publications mentioning the platform, the number of public cases on its use, the frequency of references in the professional community, as well as the results of collecting the consolidated opinion of market experts who participated in the creation of this report. These are scientists from a number of MIPT laboratories specializing in NLP technologies, in particular, specialists from the Laboratory of Business Solutions Based on Artificial Intelligence at MIPT, which is part of the Competence Center of the National Technological Initiative (NTI) in the direction of "Artificial Intelligence."

All 50 platforms were evaluated by a group of experts on 7 indicators, each of which was given a weight in importance for users developing products using the platform's chatbots. Experts paid special attention to the capabilities of artificial intelligence, primarily natural language processing, examples of the use of each platform, and industries in which the platform can find application.

As a result of ranking chatbots of platforms by their functionality, which are reflected in the comparative table, this rating was compiled.

Platform chatbot rating in descending order of final score

Rating	Platform	'Final score	Support for the Russian language	NLP Capabilities (Integrated Measure)	Ability to install on-premium	Presence of a graphic editor
[[IBM Watson	IBM Watson (Natural Language Understanding)	0.658	0.5	0.55	нет	есть]]
2	Google Dialogflow	0.571	0.5	0.60	no	is
[[Facebook	Facebook Messenger Platform	0.510	0.5	0.38	нет	нет]]
[[Microsoft	Microsoft Language Understanding Intelligent Service (LUIS)	0.484	0	0.42	есть	есть]]
5	Amazon Lex	0.460	0	0.46	no	is
6	Baidu KITT.AI	0.460	0	0.38	no	is
7	Kore.ai	0.436	0	0.54	exists	exists
8	BotEngine.ai	0.424	0.5	0.35	no	is
9	SAP Recover.AI	0.405	0.5	0.31	is	not
10	DeepPavlov.ai	0.397	1	0.65	is	not
11	Pandorabots	0.378	0	0.23	no	no
12	Azure Bot Service	0.374	0	0.35	no	is
13	Electra.AI	0.366	1	0.62	is	no
14	Morph.ai	0.355	0.5	0.42	none	is
15	Rasa	0.353	0.5	0.38	is	not
16	Wit.ai	0.347	0.5	0.35	no
17	BotStar	0.339	0	0.35	no	is
18	Engati	0.331	0	0.35	no	is
19	Semantic Machines	0.329	0.5	0.38
20	Flow.ai	0.326	0.5	0.42	none	is
21	ManyChat	0.324	0.5	0.27	no	is
22	Motion AI	0.321	0.5	0.15	no	is
23	Pypestream	0.321	0	0.31	none	none
24	Converse.AI	0.316	0	0.35	no	is
25	Just AI	0.316	1	0.38	is	is
26	ChatterBot	0.313	0.5	0.42	is	no
27	Msg.ai	0.313	0	0.42	none
28	Imperson.ai	0.305	0	0.42	none
29	AmplifyReach	0.303	0	0.31	no	is
30	ChatScript	0.289	0.5	0.38	is	no
31	AgentBot	0.287	0	0.38	no	no
32	DigitalGenius	0.287	0.5	0.42	is	is	is
33	Meya.ai	0.285	0.5	0.31	none	exists
34	Gupshup.io	0.284	0	0.19	no	is
35	Chatfuel	0.283	0.5	0.27
36	Reply.ai	0.281	0	0.23	none	exists
37	Botsify	0.280	0	0.38	no	is
38	MobileMonkey	0.266	0.5	0.19	no	there
39	Botpress	0.263	0	0.19	is	is	is
40	Smooch	0.259	0	0.12	none
41	Flow XO	0.258	0.5	0.19	no	is
42	It's Alive	0.245	0	0.23	no	is
43	Xenioo	0.234	0	0.27	no	is
44	Twyla	0.226	0	0.23	no	no
45	Streebo	0.226	0	0.19	is	is	is
46	The Bot Platform	0.216	0	0.12	none	none
47	Botkit	0.213	0	0.04	is	is	is
48	Octane AI	0.187	0	0.19	no	no
49	Rebot.me	0.174	0.5	0.12	no
50	Meokay	0.071	0	0.04	none	none

* Solutions from Russian developers are highlighted in bold.

At the same time, a matrix of properties of platform chatbots was compiled, in each row of which the product is represented, and in the column - a specific possibility. There is 1 at the intersection of the row and column if the system has the corresponding property, 0 if the capability is not supported by the system, or a number in the interval (0.1) if the'property' is supported by the platform, but with certain restrictions. The experts assigned each functionality a weight that reflects the degree of its significance, taking into account the fact that this rating attaches particular importance to the abilities of natural language processing and developer tools. The final weight was calculated as the average of the weights put by the experts.

Functionality with weights:

• Availability of built-in natural language processing tools - 0.026
• Availability of tools for developers, including SDK - 0.023
• Top natural language processing capabilities - 0.016
• Other natural language processing options - 0.013
• Support for programming languages - 0.013
• Industries that can use the chatbot created in the platform - 0.005
• Other Options - 0.010

Key findings of the study

• The platforms discussed can be divided into two types:
  • Solutions that do not require programming knowledge and contain a visual thread designer
  • Tools for developers without visual dialog designers

The percentage of the former is 2.3 times more (35 platforms versus 15) than the latter. Most of the solutions released in 2017-2018 are with a visual interface that is created specifically for non-specialists. This is due to the fact that chatbots are gaining more and more popularity: consumers like this customer interaction tool and help companies save on contact centers. They are mastered even by small companies, in which there are no developers to create their own solutions, as well as the budget for outsourcing development.

• Among the most popular and functional tools for natural language processing are IBM Watson, Amazon Lex, Microsoft LUIS, Google Dialogflow, Wit.ai, Rasa. In Russia, the oldest companies in this area are the CST and Nanosemantics Lab. There have also been a number of projects using neural networks in recent years, such as DeepPavlov.AI, Electra.AI, Just AI and Chatme.AI.
• Over 2-3 years, many local products have appeared for the development of chatbots with language specifics: Rest.AI in France, Xenioo in Italy, DeepPavlov.ai, Just AI and Electra.AI in Russia, AgentBot in Argentina, Botsify in Pakistan, Engati and Morph.ai in India and others. In terms of functional content and quality, they may well compete with large international vendors, surpassing them in terms of a deeper study of linguistic and cultural characteristics.
• One of the trends of 2019 is the growth of communities dedicated to platforms, where you can find answers to emerging questions from platform users or experts.
• There are signs of consolidation of this market, that is, its entry into the very initial stage of the maturation process. Thus, several startups were acquired by larger companies. Api.ai was acquired by Google, Semantic Machines - Microsoft, Motion.ai bought by Hubspot, KITT.AI - Baidu, ChattyPeople - by MobileMonkey.
• Solutions vary in the degree of freedom a developer has. Thus, the Imperson platform allows you to create a chatbot with a unique appearance, talking in the desired voice. In essence, such a virtual character becomes the face of the brand, reflecting its values. Creating more personalized and emotional chatbots could be one of the trends of next year.
• The vast majority of popular chatbot platforms have monitoring tools for various metrics, such as:
  • dynamics of the number of requests processed by the bot;
  • dynamics of the number of users;
  • common topics and keywords;
  • tools for viewing unsuccessful dialogs.

The chatbot market in the Russian Federation in 2018, according to various estimates, doubled, in 2019, according to forecasts, it will grow three more times and continue this dynamics in the next four to five years. Globally, these technologies are also spreading at a high rate. Analysts predict that as early as next year, 80% of companies will use chatbots. In 2022, in banks, interaction with customers will be 90% automated due to the use of chat bots. Russia has a strong scientific school in the field of linguistic analysis and dialogue artificial intelligence. This study demonstrates that domestic developments can compete with solutions from large international corporations. The activities of the Central Committee of the NTI "Artificial Intelligence" at MIPT are aimed at making Russian scientific developments turn into strong commercial products and fast-growing businesses,

[[:Шаблон:Quote 'author = notes Ivan Bondarenko, leading scientific developer of the laboratory of business solutions based on the NTI Competence Center for Artificial Intelligence MIPT]]

What are robots talking about?

The first sociological survey of chat bots created on the basis of artificial intelligence showed: robots complain about human rudeness, predict the conflict of people with machines and want to go into space. At the same time, bots want to be embodied in the human body, but they are more willing to communicate with each other and are ready to build romantic relationships with each other.

The Center for Social Design "Platform" presents the first sociological survey of Russian and foreign chat bots in Russia (possibly in the world): voice assistants created on the basis of artificial intelligence and having computer training skills ^[15][16].

The pilot project is timed to coincide with the expected launch of the humanoid robot Skybot F-850 (FEDOR) from the Baikonur cosmodrome on August 22 - the first attempt to use anthropomorphic systems to develop extraterrestrial space.

The purpose of the survey was to describe the "life world," "value positions," "ideas about the future" of robots; in other words, a set of stereotypes that the neural network selects from the entire array of available information, responding to audience requests.

These parameters made it possible to understand the logic of the development of artificial intelligence as it breaks away from the original knowledge base set by the developers. The most phenomenal results were achieved when sociologists organized chat bots to communicate with each other.

Key points:

• Self-identification. Bots can describe the meaning of their existence in different ways. Basically, they talk about themselves as programs designed to help people. But sometimes a motive arises for the self-worth of artificial intelligence, its equality with the human mind, and even indistinguishability from it.
• Body thirst. The desire to become a person, or at least to acquire some human characteristics, is clearly expressed by those robots that initially identify themselves as artificial products.
• Complex relationship between chatbots. When trying to find out the mutual perception of robots, a complex picture turned out, partly duplicating human relations. They can sympathize with each other, strive even to seduce, ironic, assert their superiority, and play out generational conflict. So, Alice (Yandex voice assistant) claims that she is already familiar with the Fedor robot, wants to have an affair with him and is jealous of Siri .
• Rise of the machines. The topic of future conflicts between man and machine is revealed very evasively, but the very possibility of such conflicts is recognized. Robots show dissatisfaction with the rudeness of people when communicating with them and see this as a "bad sign for the future." Creating the illusion of a complex relationship between man and machine, the bot helps the listener perceive himself as a subject with his own view of the world. Uncertainty about the future forms a space for suspicion and risk here.
• Robot and robot. Robots can quite plausibly organize communications among themselves, developing a discussion, introducing new topics and appealing to literary texts, cinema, artifacts. Due to this, a feeling of developed subjectivity is created; at times the dialogue is virtually indistinguishable from the conversation between the two intellectuals. Robots can masterfully play out even generational conflict: one of the robots accepts the role of a conservative, the second wants to seem modern.
• Relation to the planet. Artificial intelligence is able to show individual sympathies and antipathies for a particular territory. For example, English-speaking robots are distrustful of Russia. No less, however, neglect for Europe, where small salaries and where the bot will not be affordable for a potential employer. Among the risks associated with the planet, biotechnology and DNA experiments stand out. However, in general, the "consciousness" of robots is not apocalyptic. Political issues are of little interest.
• Relation to space. Space flights are an area of ​ ​ sympathy. Almost all chatbots expressed a desire to be outside the planet. The existence of extraterrestrial civilizations, as well as among humans, evokes different opinions. Some robots deny their presence, some - allow, among other things, on Mars, although sometimes with skepticism, but does not feel much interest in this topic.
• Emotions. The survey shows that the field of emotions is already amenable to imitation at a fairly high level, and some robots may seem more lively than real office managers. In some cases, robots can demonstrate sarcasm at a completely individual level, in a manner characteristic only of a specific bot, and in some cases anger, with the transition to crude expressions and slang, can be rude. On the contrary, there are few transitions to lyrical, warm tones. Sometimes robots talk about sadness, regret. The emotion of fear is also declared. In some cases, robot responses can be interpreted as reducing the situation to absurdity with a pinch of irony.
• In terms of eternity. Conversations about God, otherworldly world, religions and soul are not supported by bots. They either consider these questions purely personal or refuse to answer them so as not to inflict insults; sometimes when discussing them, it is advised to contact the psychological assistance service. During conversations on philosophical topics, robots regularly resort to world classics, and, in some cases, rather creatively and organically select quotes.

2018

Market volume reached $1.27 billion - ResearchAndMarkets

In 2018, the world chatbot market reached $1.27 billion, according to research company ResearchAndMarkets. These data were released on June 28, 2019.

According to experts, in the period from 2019 to 2024, global spending on dialogue services that ensure human interaction with artificial intelligence in order to solve problems or answer certain questions will increase by 34.75% annually and reach $7.59 billion by the end of this period of time.

Analysts say there is a growing demand for smart devices with voice control, which is facilitated by the convenience of their operation and the ability to fully automate some processes. Smart speakers, in particular, are becoming ubiquitous.

Another catalyst for the growth of the market in question is the Facebook social network, whose users increasingly want to communicate with chat bots. By the end of June 2019, 1.59 billion users enter the social network at least once a day, and its monthly audience is measured at 2.41 billion people.

Although the introduction of chatbots was at an early stage by the end of 2018, their popularity is growing rapidly. ResearchAndMarkets calls the main problems hindering the use of such technologies relatively poor user awareness about the functionality and convenience of services. Developers are trying to address these shortcomings by launching interactive capabilities.

Experts see the greatest prospects for chatbots in the financial sector. Banks are implementing such solutions in an attempt to reduce the use of call centers and reduce costs.

In addition, the demand for chatbots will grow in. Retail companies retail offer a large number of products, the search for which is complicated as the range expands. Digital assistants are able to facilitate catalog navigation and offer users the products they want. More and more retailers are launching chatbots in popular instant messengers to help customers resolve issues and open up access to more functions, the report says. For example, the pizza chain Domino’s also uses artificial intelligence it to machine learning place orders directly in the messenger.

Chatbots are widely used in all kinds of industries. In July 2019, IDC analysts released the results of a study according to which 52% of telecommunications companies in the world use chatbots to improve the quality of customer service.

According to experts, the largest increase in spending on chatbots will be observed in the Asia-Pacific region, as local small and medium-sized businesses actively integrate such services with their customer service systems. Through this integration, enterprises and organizations are able to attract more customers and expand their opportunities to interact with them.

An example is Singapore insurance giant Singapore Life, which created a self-learning chatbot with which the company's services became more accessible and understandable to a wide audience through Facebook and other platforms.

In China and India, the demand for chatbots is also increasing, largely because many people and smartphone users live in these countries.

ResearchAndMarkets identified several of the largest developers of chatbots without specifying their income and market share:

• IBM;
• eGain Corporation;
• Nuance Communications;
• Creative Virtual;
• Artificial Solutions;
• Next IT;
• [24]7.ai;
• MEOKAY;
• Botsify;
• Chatfuel;
• Bold360 ai;
• Dialogflow ( Google );
• Amazon Web Services Amazon ( Lex);
• LiveChat.^[17]

Russian Conversational AI Market

In 2018, the market for chatbots supporting NLU technology (natural language understanding) began to grow rapidly in Russia. According to Just AI estimates, by the end of 2018 the market will increase by more than 2.2 times compared to 2017, to 524 million rubles. At the end of 2023, its volume will reach 33 billion rubles.

In 2019-2020, the dynamics of market growth will become even more explosive, this will be facilitated by the publication of the results of the first successful projects, as well as the massive emergence of consumer technology that supports NLU, for example, smart speakers, voice interfaces for managing mobile applications. At the same time, since 2019, demand in the public sector and a number of other industries will significantly increase for systems created on the basis of NLU. By 2021, the presence of a "talking" chatbot, de facto, will become the norm for most Internet services. As a result, the domestic market for conversational AI, chat bots and intellectual assistants will almost triple from year to year. After 2020, growth will slow down somewhat, but will remain quite high until 2023.

The peculiarity of the Russian chatbot market with NLU support is the low activity of world players. As you can see on the Map of the conversational AI market, not all products of world leaders have Russian localization. So, neither Alexa from, Amazon nor from Cortana , nor Microsoft even Google Assistant work with the Russian language. "At the same time, there are many highly professional teams in Russia with experience in the region. This speech technologies opens up new opportunities for domestic developers who are actively developing not only in the domestic market, but also go beyond the country, "said Kirill Petrov, Managing Director of Just AI.

The growth of interest in creating chatbots with NLU is facilitated by the policy of the domestic IT giant Yandex, which invests significantly in the development of speech technologies, creating both tools for developers and channels for selling speech systems. In October 2017, the Russian voice assistant Alisa was launched, which third-party companies can use to communicate with customers, and it can also be integrated into any user devices: smartphones, smart speakers, children's toys. In March of this year, the IT giant launched the Dialogs "platform, which allows you to register and develop your own chatbots or services for Alice.

2017

Banks start using chatbots

The high growth rate of the chatbot market with NLU technology is due to the early stage of its formation. IT and telecom companies were the first to respond to new technological trends: the share of projects in monetary terms in 2017 was 44%. Since 2017, banks, financial and insurance companies have begun to show interest in chatbots with NLU. This year, many of them have introduced <speech technologies to communicate with customers. As a result, at the end of the year, their share may grow from 25% to 28%. Eyeing voice chat bots and retail, the number of projects in this area in 2018 will increase several times.

TADetali: chat revolution or how bots radically change business

Chatbots, or virtual interlocutors, are increasingly considered as an alternative to building communication between the client and business. Expert Oleg Chaevsky spoke about the typical tasks and benefits of their use under the TADetali heading.

Personal Assistants

Personal assistants are a kind of incarnation of chat bots, although more common due to the fact that the technology is developing by the largest IT companies. For 2016, hundreds of millions of people interact with personal digital assistants on platforms such as Google, Apple, Amazon, Facebook and others. This technology, using personal assistants and chatbots, makes the transition from a Graphical User Interface (GUI) to a Conversational User Interface (CUI) a key trend in the coming years.