What are chatbots for and how much do they cost?
Introduction
A chatbot is a computer intelligent agent that provides interaction between business and its customers. It is located in a pop-up window on the website or in various messengers and social networks. It is also possible to implement voice communication between a bot and a person.
Five reasons why chatbots are needed:
- Customers should see that business is in touch with them 24 hours a day, 7 days a week.
- Chatbots extend sales channels and can be located in messengers and social networks.
- Chatbots reduce personnel costs, since they take on many routine tasks.
- When communicating with the bot, customers feel personal relationship to them, which increases their trust in the company.
- People like to communicate, while receiving information from the interlocutor, and not from static texts on the site.
Chatbot functions
Below are some examples of tasks that chatbots perform. Each of them can be implemented by a separate chatbot or be an integral part of a single company bot.
- 📞 Expanding the customer base. A chatbot is more efficient at collecting customer information than, for example, a form on a website. At first, the bot can only ask about the client's name and introduce itself. Subsequent conversations use the client's name, which increases the level of trust and engagement. Therefore, further information retrieval is easier. Suggestions to enter a phone number or email may appear in a variety of branches of the dialogue, for example, when a person has questions or after information about upcoming super discounts.
- 🦉 Frequently Asked Questions. Unlike static information on the site, the organization of answers can take place in the process of dialogue, for example, before answering a question about the price, the client is asked clarifying questions about the parameters of the product. Many questions are typical and instead of live employees, chatbots can answer them.
- 📅 News and upcoming events. A chatbot is a convenient tool for prompt notification of company news or future events. If the chatbot is in the messenger, then new messages will appear as notifications. When a client communicates with a bot, in a suitable time it can report relevant news: "If you are interested in this service, please note that next week ..." and further: "Leave your phone or email and we will inform you ..."
- 👫 Personnel search. The chatbot can inform about the basic requirements for the applicant, answer questions about the proposed conditions, get the necessary personal information. Then it can conduct a preliminary interview, usually in the form of multiple choice questions. The test results and the completed questionnaire are sent to the CRM system or to the email of the company's HR department.
- 🛒 Selling goods and services includes a wide range of actions that ultimately lead to a purchase. These can be recommendations for selection, intelligent navigation through the catalog, selection of quantities, various parameters and an offer of related products. After forming a shopping cart, a person can connect to the dialogue.
- 📅 Making an appointment implies the chatbot's connection with some external system that stores information about the upcoming visit (making an appointment with a hairdresser, seeing a doctor, etc.). In the simplest case, this can be a google calendar, which is jointly edited by a bot and a company employee.
- 🛠️ Product support is an important part of customer service. A chatbot can not only answer the most common questions, but also offer alternative products and services.
- 👩💼 Company presentation (organizational structure, services, products, team, contacts, etc.). Usually this information is available on the site, but getting it in the process of dialogue allows you to focus on what is directly interesting to the client.
- 👍 Reviews & Polls. Understanding the bottlenecks of a business, its products and services is very important. The bot can ask leading questions to understand with what the client is dissatisfied. With the help of a chatbot, you can conduct surveys about customer wishes for new versions of products, etc.
- 🗄 Intracorporate information. A chatbot can help new employees of the company become part of the team. With the help of a chatbot, you can organize electronic document flow, determine the dates of employee vacations, and much more.
- 📖 Testing, learning. Bots are ideal tools for testing and improving employee skills. In the simplest case, these can be test questions with multiple answers. For each question is usually given a certain amount of time so that the subject cannot get an answer on the Internet. In the process of studying, suggestions for prompts may appear, a story about student's strengths and weaknesses is stored, etc.
- 🤡 Free communication. Usually chatbots solve a specific problem. However, the bot can steer away from the main topic and talk "about life". This ties the client more closely to the company. Bots can solve psychological problems, fight loneliness and become a real friend for your client.
Buttons or text input?
According to the method of communication with a person, there are bots of two types:
- Bots with button navigation. In them, the dialogue proceeds as a result of clicking on buttons with text, which leads to the appearance of new blocks of information. The question or answer of the person is actually the inscription on the button.
- ➕ Pros:
- such bots are relatively easy to implement and inexpensive;
- people don't like to write and it is easier for them to click on buttons;
- allows you to create complex navigation systems for the catalog of goods and services.
- ➖ Cons:
- navigation sometimes turns out to be a rather long chain of button presses;
- it is immediately clear that communication is not with a person;
- may not work in all messengers with their restrictions on the number of buttons.
- ➕ Pros:
- Natural language understanding bots. The text that a person enters in the input line is assigned by the bot to one of the predefined classes (intentions). Depending on the recognized intent, further navigation or information delivery takes place.
- ➕ Pros:
- more convenient navigation, because you can go directly to the desired topic;
- more trust in the bot as a live interlocutor;
- can work in any messenger, since there is no limit on the number of buttons.
- ➖ Cons:
- such bots are more difficult to implement and cost more;
- they require many examples to pick up to correctly discern intent;
- bot needs long-term support to configure NLU engine.
- ➕ Pros:
🦉 Often, bots implement full or partial combination of button and text navigation.
Rules or training?
According to the algorithms of behavior, there are also two types of bots:
- Rule-based bots that contain a rigid algorithm for reacting to user actions. This does not mean that the user sees the same texts (they may vary), but the displayed information is completely controlled by the developer.
- ➕ Pros:
- easier to develop and test;
- bot's behavior is controlled, strange or politically incorrect answers are impossible.
- ➖ Cons:
- cannot deviate from the given program;
- sometimes doesn't look smart enough.
- ➕ Pros:
- Self-learning bots that use neural networks and other artificial intelligence techniques. They are a "black box", the behavior of which is configured by using machine learning.
- ➕ Pros:
- can look "almost human";
- are able to respond appropriately in unexpected situations;
- adjust to user behavior.
- ➖ Cons:
- behavior can be unpredictable;
- require lengthy configuration with analysis of the real users' behavior.
- ➕ Pros:
🦉 As with the navigation system, it is possible to combine rule-based logic and machine learning.
How much does a chatbot cost?
The price varies greatly depending on the complexity of the bot and the tasks it solves. Here are some guidelines, although prices on the market can vary greatly in one direction or another.
- A relatively free bot can be made using numerous constructors that are available on the Internet. It will take some time to master these tools. In this case, in the course of work may appear insurmountable restrictions inherent in the chosen designer. Such bot usually cannot work with data and is a simple navigation through screens with pictures and text using buttons or simple text commands.
- A rule-based button bot with dozens of navigation screens that send notifications about user actions (phone number entered, etc.) costs about $ 1000-2000. A significant part of the developers' costs is associated with the stages of adjustment with the customer the logic of bot's behavior, text and graphic content, several iterations of testing, etc.
- The bot, that described above and additionally uses many notification channels (google-calendar, CRM, etc.) adds $ 500-1000 to the price (if the channels are very specific, the price may rise). Approximately the same additional price fork arises when it is necessary to promptly notify users of news in various places of the bot.
- With an increase in the number of screens or blocks of information, the price may increase, while not always linearly, since the bot's logic gets more complicated. But you can estimate $ 50 for each additional screen. The price also depends on the number of languages supported by the bot. If all the texts are submitted by the customer, then the price increase is insignificant.
- Comprehension of natural language can add a wider fork to the price of a bot, from $ 2000-10000, depending on the number of recognized intentions and the duration of setup in the conditions of bot's work with real users.
- Using the voice interface (phone) will cost the development another $ 5000-50000.
- Finally, the use of sophisticated artificial intelligence algorithms can increase bot's value to five or six digits.
The base location of the bot is the company's website. Often the same bot or its adapted version is placed in various messengers (facebook messenger, instagram, telegram, etc.). Moreover, each channel has certain restrictions on the use of interface elements. For example, navigation involving a large number of buttons will work on a website or telegram, but not on facebook or instagram. Such adaptation of the bot for different channels should be taken into account when forming the price.
Finally, there are the monthly costs of hosting a bot. Depending on the intensity of its work, it can be about $ 20-100.
What do you need for a chatbot?
Naturally, you need a desire and financial resources. In addition, you should answer a series of starting questions that will help form the cost of the bot and the timing of its creation:
- Industry and clients:
- Goodman Daughter & Sons Law Firm.
- Advice and representation in family law court.
- More than 300 regular customers.
- Distribution channels:
- Company website www.goodman-daughter-and-sons.com
- Facebook messenger.
- Office phone.
- Bot languages:
- English
- Spanish
- Tasks solved by the bot:
- Company presentation:
- Our lawyers (photography, education, experience).
- Information about successful processes.
- Locations of our offices, contacts.
- Extension of the customer base (name, phone number and description of the customer's problem).
- Make an appointment with our lawyers.
- Search for information in the document database.
- Company presentation:
- Interaction interface:
- Button navigation with text input elements (name, phone, customer question).
- When searching the database - recognition of text terms from about 50 thematic domains.
- Voice interface for the task of making an appointment (calls to an office phone).
- Communication with external resources:
- Receiving information about a new client by email, saving it to the bot database.
- Integration with google-calendar about appointment, duplication in the messenger.
- Http - requests to the document database on the company's server.
- News and Discount Alerts:
- Not yet.
Next, you need to provide the content that is necessary for the bot to work: texts, images, videos. In principle, the developer can use the resources from the customer's site and even act as a copywriter (for an additional fee). However, only you know what your client needs to hear.
How is development going?
After agreeing on the concept of the bot and receiving the initial content, a prototype of the bot is created, which is provided to the customer for approval.
For a rule-based bot with button navigation, the prototype is a set of screens with buttons that navigate those screens. The screens may contain stubs such as "there will be scrolling of goods with a picture, name and description", etc. The customer tests the bot and can edit the texts and labels on the buttons right there.
The second iteration consists of filling the bot with internal logic of behavior, processing the input of text information and loading data (catalog of goods, list of services, etc.). At this stage, the bot is already quite functional.
At the third iteration, interaction with external systems (databases, CRM, google-calendar, etc.) is organized. Any of the stages, depending on the complexity of the bot, can be divided into sub-stages.
In a bot that understands natural language, it is necessary first of all to determine the list of intentions of the bot users. Examples of intentions: "Greetings", "Consent", "Misunderstanding", "I want to buy ...". For each intention, the developer prepares examples of phrases in one or another natural language. So, for the "Greeting" intent, dozens of phrases are possible from " good afternoon " to " hello, soulless piece of iron ". The more there are such examples, the better it will be. Then machine learning is carried out, as a result of which the bot will begin to "understand" not only the training examples, but also their various variations. This work requires many iterations and continues after the bot is launched into production. At the same time, the behavior of real users is analyzed from what texts they enter when communicating with the bot.