Blog

Articles on customer service, artificial intelligence, and their interaction.

Wikibot chatbot is now available on Albato, a leading no-code service integration platform

Wikibot chatbot is now available on Albato, a leading no-code service integration platform

Wikibot is a smart assistant for support services that learns from your documentation and responds to user queries in chat or via email, acting as a first-line support specialist. The service's advantage lies in its quick start and virtually no "magical" settings to get started. Just provide us with a link to your knowledge base, and within a few hours, you can start asking the bot questions and receive answers. Instead of a complex script constructor, in Wikibot, you train the bot like a new employee, showing it how to respond to different types of questions. If you don't have a knowledge base, you can create training examples for the bot in Google Sheets or on the Wikibot portal. The unique feature is the use of the large language model LLM under the hood. Therefore, the bot understands questions in any formulation, except genuinely debatable or complex ones, and responds like a human. Regardless of the formulation, the chatbot understands the question, finds information in the knowledge base or FAQ, and answers the user. Using Wikibot for customer support will enable you to: - Reduce the overall ticket resolution time by providing instant answers to typical questions. - Increase customer loyalty by improving service quality and support availability at any time of day or night. - Free up your specialists' resources to handle more complex and crucial cases. ## Integrating Wikibot with Albato Integrating Wikibot with Albato opens up many useful scenarios, such as: - Adding an intelligent assistant to your service that understands natural language, the most popular interface used by approximately 8 billion people worldwide. - Modern helpdesk systems and website chats are already integrated with Wikibot. However, sometimes you need to send only certain types of user questions (tickets) to the bot. Albato is excellent for this. In Albato, you can configure ticket filtering by any fields, as well as additional actions. - Using Wikibot in your favorite applications, such as Google Sheets. ## Example of Integrating Wikibot with Google Sheets Let's implement a simple example where the user writes a question in the first column of Google Sheets and receives an answer from their Wikibot in the adjacent cell (second column). To repeat the example, you need to register with Wikibot and create your bot. The process of creation is detailed in the [documentation](https://docs.wikibot.pro/). In the bot settings, in the API Keys section, you need to add a key. ### Example of Integrating Wikibot with Google Sheets Let's implement a simple example where the user writes a question in the first column of Google Sheets and receives an answer from their Wikibot in the adjacent cell (second column). To repeat the example, you need to register with Wikibot and create your bot. The process of creation is detailed in the [documentation](https://docs.wikibot.pro/). In the bot settings, in the API Keys section, you need to add a key. ![1 1-e1705356394794.png](https://cms.wikibot.pro/uploads/1_1_e1705356394794_418bb20e6a.png) In the example, we will use the bot [Wikibot_support_bot](https://t.me/Wikibot_support_bot), which was trained on the Wikibot website. Let's create a new bundle, which will retrieve the value of the cell from Google Sheets, pass the question to Wikibot, and write the answer to the adjacent cell in Google Sheets (the final version of the scheme is shown in the screenshot). ![2.png](https://cms.wikibot.pro/uploads/2_2459a715f8.png) Now, let's delve into each step. ### Add the "Google Sheets: New Row Created" step. In the settings, you need to select the document where the first column is "Question" and the second column is "Answer Generation." For such a bundle, it's essential that the Google Sheets document is shared for editing via a link. ### Add the "Wikibot: Ask Wikibot a Question" step. Here, you will need the API key created in the bot settings. ![3.png](https://cms.wikibot.pro/uploads/3_7a4747a7cf.png) Set "Created a new row: Question Column" as the Query parameter. In the Ticket ID field, set 1, as it is not needed in this example. ![4.png](https://cms.wikibot.pro/uploads/4_c2d5168fab.png) ### Add the "Google Sheets: Update Row by Number" step. Set the "Ask Wikibot: Answer" as the value of the answer column. ![5.png](https://cms.wikibot.pro/uploads/5_939a0e6c81.png) ### Result Run the bundle and start typing questions in the first column of the document; after 1-2 minutes, answers will start appearing in the second column. Thanks to this bundle, you can quickly assemble a database of your bot's responses for further analysis and improvement of its performance. ![6.png](https://cms.wikibot.pro/uploads/6_67978c0113.png) ### Conclusions Ask questions in the comments about the possibilities of applying natural language artificial intelligence in your products. Wikibot specialists are available [https://t.me/use_wikibot](https://t.me/use_wikibot).

tomleto
Comparison of Wikibot with competitors in Russia

Comparison of Wikibot with competitors in Russia

Chatbots with artificial intelligence for user support in helpdesk are rapidly gaining popularity. It seems that popular helpdesks offer a huge choice of AI chatbots. We compared market leaders and found a couple of options where you can try creating and training a bot with your own data. The products included in the comparison are: Wikibot, JustAI Aimylogic, Chatme.AI, AutoFAQ, SupportAI, Nanosemantika. ## Comparison Criteria - **Available Large Language Model (LLM)**: ChatGPT, YandexGPT, Custom Network. - **Bot Training Capabilities**: Testing the bot with your own data (otherwise, bot training is a custom development and can be expensive), parsing popular knowledge base sites (Notion, Confluence), and files. - Integrations with helpdesk, chats, messengers. - **Orientation towards technical support**: Copilot, Prompter, working hours, flexible response templates. ![Comparison of Wikibot with competitors in Russia.webp](https://cms.wikibot.pro/uploads/Sravnenie_Wiki_Bot_c_konkurentami_v_Rossii_f182e9e2b0.webp) ![Untitled (1).webp](https://cms.wikibot.pro/uploads/Bez_nazvaniya_1_e88b76269f.webp) ## Conclusions Most products for creating AI chatbots are actually custom development. Many have a visual constructor, but it only allows creating scripted bots. You can create and train an AI bot with your own data for free only in Wikibot and JustAI Aimylogic. Contact us if you need more comparison characteristics or an analysis of chatbots that are not included in the comparison. We will conduct an analysis for you.

tomleto
Open letter to hosting services about a support chatbot

Open letter to hosting services about a support chatbot

Hosting services usually have an excellent knowledge base, so an AI chatbot will help them reduce customer response time. This letter aims to establish a connection. Hello, folks! We present Wikibot—a smart assistant for customer support services that learns from your documentation and responds to user queries in chat or via email, functioning as a first-line support specialist. Our advantage lies in quick deployment with almost no "magical" settings required. Just provide us with a link to your knowledge base, and within a few hours, you can start asking questions to the bot and receiving answers. Under the hood, we use a large language model (LLM), allowing us to focus on speech interaction instead of intricate configurations. Our interface relies on natural language processing, enabling communication with users in their native language. Therefore, our chatbot understands queries in any formulation, except for genuinely controversial or complex ones, and responds like a human. Regardless of the question's formulation, the chatbot understands the query, finds information in the knowledge base or FAQ, and responds to the user. We propose using Wikibot for your customer support. This will: - **Reduce waiting time** for customers with instant answers to common questions. - **Increase customer loyalty** by improving service quality and support availability 24/7. - **Free up resources** of your support specialists for more complex tasks, thereby enhancing their efficiency. **Wikibot—an intelligent assistant crafted with a passion for customer support** We listen to our customers and evolve the product together. In the last six months, we've added: 1. Exception lists—rules for when to transfer to an operator, for example, questions about "money." 2. Working hours and non-working hours. 3. Integration with popular Help Desk systems, knowledge bases, chats, and messengers. 4. Answer algorithm settings, working in prompt and copilot modes. Check out a sample chatbot: [Excel_Wikibot](https://t.me/Excel_Wikibot) (Excel Assistant) Read real user reviews about our product on [Startpack](https://startpack.ru/application/wikibot). Please forward this letter to the organization's executives. Let's make customer support smarter with Wikibot!

tomleto
Comparing chatbot builders: Botmother vs Wikibot

Comparing chatbot builders: Botmother vs Wikibot

# Comparing Botmother and Wikibot: Two Popular Chatbot Builders Chatbots are a distinct type of application that integrates into messengers and customer support chats on websites. Users appreciate being able to perform necessary actions without leaving their favorite messengers. In this article, we will explore use cases of two popular chatbot builders: Botmother and Wikibot. ## Botmother — a Constructor for Creating Scenario-Based Chatbots To create a bot, you need to describe the bot's scenario in a visual constructor. At each step of the scenario, a screen is created that users see with buttons and texts. The bot's logic is set up once and works the same on all platforms: VKontakte, Telegram, WhatsApp, etc. Botmother has extensive capabilities: - **Collecting user surveys** — allows configuring fields that users need to fill out. - **Payment acceptance** — enables selling products and services, receiving payment directly in the messenger using popular payment systems (Robokassa, PayOnline, bePaid, YooKassa). - **Statistics** — provides useful information about users. You can track at which stage users drop off. - **Mailings** — allows sending news and offers to users. - **Switching to an operator** — needed to connect an operator to the chat, preventing customer loss and avoiding negativity if the bot doesn't understand anything. - Integrations with popular CRMs. ## Wikibot — a Constructor for Chatbots that Understand Natural Language The Wikibot service allows creating a bot for automatic responses to chat clients in natural language, replacing a first-line technical support specialist. It uses existing documentation on the company's products or services for training. Advantages of using the service: - **Easy to start**. To create an AI bot, simply provide a link to the knowledge base or product website. - **Natural responses**. The chatbot, using a large language model, understands questions and provides answers in natural language. - **Reduced staff costs** and, for users, reduced waiting time for responses. - **Improvement of the knowledge base**. The service provides analytics on questions that the AI did not find answers to in the documentation. This information forms the basis for improving the support department's knowledge base. - The bot automatically learns when articles in the knowledge base are updated. - The service supports integration with website chats and helpdesk systems. ## Conclusion The functions of Botmother and Wikibot hardly overlap. If you need to collect user data, receive payments, inform users, use Botmother. If you need a bot that reduces the workload on the support department and helps users solve problems, use Wikibot. By the way, some scenario-based chatbot builders have already implemented integration with Wikibot, such as Talk-me. If you are unsure about your choice, feel free to reach out to us.

tomleto
Create AI chatbot in Google Sheets

Create AI chatbot in Google Sheets

1. Go to the Google Sheet using the link https://bit.ly/create_wiki 2. Make a copy of the document (File -> Make a copy). 3. Fill in the document with your questions and answers. 4. Open access to the document for everyone (Share -> Anyone with the link -> Copy link). Click the "Copy link" button. 5. Open Wikibot on Telegram at https://t.me/tom_test_wikibot, click "START," and then click the "I want to create a new bot" button. 6. Paste the copied link and send it. After a few minutes, the bot will learn, and you can start asking it questions. Creating a bot in Google Sheet is possible thanks to the Wikibot service.

tomleto
Open Letter to BI Companies — Project DashboardAI

Open Letter to BI Companies — Project DashboardAI

Hello, Analytics, Managers, and Founders of Business Intelligence companies! I am developing the Wikibot service, which learns from documentation and advises users in chat as a first-line support specialist. Our chatbot understands questions and responds like a human. The world is moving towards the use of AI assistants. People like getting results by simply giving commands in their native language. I propose that, together with my team, we create the product Dashboard.AI, which will have the following functionality: - **Report Search.** A manager can ask the bot to find a specific report. For example: "Find a report on advertising expenses for the past year." Based on the list of reports in the data catalog, Confluence, or another wiki, the bot sends a link to the chat. - **Writing SQL Queries.** Instead of searching for a specific report, a manager writes to the bot in the chat: "Profit in Moscow for electronics from May to July." The bot sends the necessary metrics and a link to the relevant report, with applied filters. If there is no suitable report, it writes the SQL query itself. - **Automatic Dashboard Creation.** A manager can ask the chatbot to create a dashboard. The bot, asking a few clarifying questions, will choose the optimal template and assemble all the necessary SQL queries and filters. ## Implementation **Report Search.** The list of reports is usually located in the knowledge base, documentation, or data catalog. Modern products, such as SiteGPT, ChatBase, WonderChat, Wikibot (hereinafter the agent), excel at semantic search: Documentation is indexed and stored in a vector database. When a user enters a query, the vector base finds the 5-10 closest vectors (semantic search). Each vector represents the name and description of the report. The user's query and the top close vectors are sent to the LLM - a large language model (LLaMA, Falcon, Google Bard, Anthropic Claude), and it selects reports that fit the user. **Writing SQL Queries** Modern LLMs already write SQL fairly well (an article comparing GPT, Claude, Bard: [Great SQL Bot Bake Off](https://www.linkedin.com/pulse/great-sql-bot-bake-off-comparing-big-llm-beasts-code-ian-thomas/)). To do this, you need to ask a question and pass a description of the tables in the prompt. After receiving the user's query, the agent can take the table names from the documentation and all the fields of the table from the metadata of the DBMS. Currently gaining traction are projects: - DB-GPT - allows generating queries to databases using local LLMs - PandasAI - makes pandas (and other popular data analysis libraries) conversational, allowing you to ask questions to your DataFrame in natural language. **Automatic Dashboard Creation** GPT, given a data table, can already choose the most significant and suitable columns for visualization. A comprehensive example of how to do this is described in the article "Create AI-powered dashboards" [Create AI-powered dashboards](https://www.luzmo.com/blog/ai-powered-dashboards-tutorial). Adding to this dashboard templates, various best practices in the form of heuristics, and clarifying queries from the bot, you can get a Junior BI developer. ## Looking to the Future In the implementation section, I described what can be done right now. A couple of ideas for the future: If you teach the "Dashboard_AI" project to create dashboards, the next step will be the ability to transfer dashboards from one system to another. Currently, a huge amount of resources is required to transfer a company's analytical reporting to an alternative BI product. In addition, it will be possible to create dashboards immediately for several popular products (Tableau, Power BI, Superset, Metabase, Redash). A personal analyst who helps managers quickly get the necessary metrics and find insights can become a highly demanded product. Contact me if: - You are interested in adding the above functionality to your product. - You want to create a similar product in collaboration with our team, based on popular BI products (Tableau, Power BI, Superset, Metabase, Redash). The world is moving towards the use of AI assistants. People like getting results by simply giving commands in their native language.

tomleto
In October 2023, Wikibot became Product of the Week #1 on ProductRadar

In October 2023, Wikibot became Product of the Week #1 on ProductRadar

# How the Idea Came About It was hard not to notice the boom of ChatGPT and artificial intelligence in general. We started experimenting with language models and capabilities, trying to understand what AI can and cannot do, its advantages, and limitations. We realized that we achieved quite a good result from the start and had a huge potential for improvement. We then started brainstorming specific product use cases, leading us to a solution for the support department. # Time from Idea to First Client Our initial goal was to find three clients within the first month, and that's roughly what happened. Among this trio was Skillbox. We were very fortunate with them. We initiated our sales by sending proposals to potential clients' email addresses, and one of the recipients was the public email hello@skillbox.ru. To our surprise, they responded! It went something like this: - "We are developing a cool AI-powered chatbot, and we propose you try it." - "Okay, let's give it a try!" That's how we delved into real user cases and "prod-prod" 🙂 So, don't be afraid of such straightforward approaches as direct sales! # Most Challenging Moment and How We Overcame It The most challenging moment is what we are currently facing, strangely enough, related to AI 😀 We are very close to a clear understanding of how to close the first line so that we can scale. There is a real understanding that we are finding p/m fit, but imagine that AI is like a child, albeit a brilliant one, on whom we depend. Therefore, we have to deal with its whims from time to time. But we will cope, and our efforts will pay off. [![wikibot ceo](https://cms.wikibot.pro/uploads/sasha.png)](https://www.youtube.com/watch?v=LRgREgbPIg)

tomleto
Copilot: Bridging the Gap Between AI and Human Interaction in Helpdesks

Copilot: Bridging the Gap Between AI and Human Interaction in Helpdesks

**Copilot** is a concept where human and AI complement each other (akin to symbiosis in biology). Usually, a person formulates the task, and AI handles its parts. This article explores the interaction between AI and specialists in helpdesk chats. In essence, a helpdesk serves as a dispatch center, assisting clients with their pressing issues. Typically, it includes: - Tools for receiving user queries from various channels (Website, Telegram, Email, etc.). - A unified tool for responding to user queries and configuring flows for different question types. User queries are commonly referred to as incidents, going through several stages and closing when the user's issue is resolved. - Analytics tools for operators, question types, and issue sources. Analytics helps improve the entire support system. - A knowledge base where support service managers add answers to frequently asked questions and other information for consistent and high-quality responses from all operators. - Automation tools - triggers, integrations, bots. Let's delve into the last point, specifically the creation of chatbots for technical support. To ensure high-quality technical support, companies need a support department ready for peak loads. However, maintaining a support department ready for peak loads is expensive and usually not cost-effective. In 2023, a new solution emerged for providing quality chat responses in natural language without significant development costs - chatbots based on large language models (such as ChatGPT, LLaMA). They inherently understand natural language and only require the company's technical support knowledge base. Such chatbots easily handle typical questions, and complex ones can be transferred to operators. How these bots work is detailed in this [video](link-to-video). And here we come to the concept of symbiosis or Copilot. Many professionals believe that AI will not radically replace humans in the coming years, but Copilot will emerge in various professions: - GitHub created a superb assistant for developers, GitHub Copilot, which completes functions, checks code, and generates simple logic. - Midjourney did the same for designers and copywriters - quick assistance in creating drawings or prototypes. - Tesla introduced Copilot for drivers. ChatGPT is a specialized Copilot for text-related work for everyone. In Google Workspace, Duet has appeared to help business users solve typical tasks with documents and emails. **Chatbot and Support Specialist Together** In the simplest case, a chatbot works like this: - IF it finds an article with the user's query answer in the knowledge base, - THEN provide the user with an answer based on the information in the article, - ELSE switch to the operator. Of course, there are many nuances: handling greetings, farewells, profanity, clarifying questions, etc. However, there are more interesting symbiotic scenarios. For example, when an operator receives a question, in their response text field, the bot writes an answer, and the operator decides whether to send it or make corrections. Also, the AI bot can connect to a regular scripted bot on one of the branches. This is convenient when there are too many response options at one step in the flow. Interestingly, one of our clients independently invented Copilot. The client was hesitant to deploy the bot to real users initially, so they instructed support specialists to send all user queries to the bot. Thanks to this idea, problematic questions were identified within a month, the knowledge base was improved, and the chatbot was introduced into real support processes. Thus, I've outlined three Copilot scenarios in support chats: 1. AI answers simple questions; humans handle complex ones. 2. AI generates an answer, and humans decide whether to send it or enhance it. 3. AI integrates into scripted bots. I believe that any helpdesk should possess AI capabilities to compete in the future market. It would be great to organize an open discussion on helpdesk development. Join the post and share your vision: - Do you think Copilot is needed in helpdesks? - Are you integrating AI into your helpdesk? - What are your plans for helpdesk development in 2024? I know that most users dislike old chatbots. A car used to be slower and less stable than a horse. We are creating new intelligent chatbots to solve your problems faster. I would appreciate your responses: - Share your latest experience interacting with a bot. - Were there cases where a bot answered your questions well? - Do you want to know if you're chatting with a bot, or is it enough to get a proper answer to your question?

tomleto
Creating RAG-powered chatbot using LLM, Vector DB, and a website parser

Creating RAG-powered chatbot using LLM, Vector DB, and a website parser

DataLearn is the most useful and my favorite platform for DATA specialists in Russian. I wanted to speak on it several times, but each time I hesitated because I speak very poorly (I have had cerebral palsy since childhood). This time I found the strength to overcome myself. We've created a cool project in which we build a self-learning chatbot using a large language model, Vector DB, and a website parser. As a result, we get a chatbot that indexes the site and answers questions about it like a human. All source code is open source. We would appreciate your likes and reposts on any social networks. [![Check out the project](https://cms.wikibot.pro/uploads/Snimok_LLM_fcbd9d0d85.PNG)](https://www.youtube.com/watch?v=8IRKx3d7tZY) **Source Code:** [GitHub Repository](https://github.com/TomLetoAI/chat-example)

tomleto
Trying out Automatic.chat - a service for creating an AI chatbot for a website.

Trying out Automatic.chat - a service for creating an AI chatbot for a website.

The number of services addressing the challenge of creating an AI consultant based on knowledge databases is growing. We introduce Wikibot, an AI-powered chatbot for customer support. Drawing from our experience, we will examine the most interesting products and services available in the market. Comparison is conducted in August 2023, and the functionality of services is subject to constant change, so it's essential to consider this when reading the article. Let's dive in! ![Image2](https://cms.wikibot.pro/uploads/Risunok2_90989a54c4.webp) ## Homepage Attempting to create an AI chatbot using our documentation in 10 seconds. Note that on the landing page, there is a "Create a Bot" button. If our focus is on assisting company support services, there's broader applicability: orientation towards knowledge about the website. ![Image3](https://cms.wikibot.pro/uploads/Risunok3_90e7362b35.webp) The portal features a convenient, animated bot creation wizard guiding users through the process. +1! ![Image4](https://cms.wikibot.pro/uploads/Risunok4_f0dc81f31c.webp) ## Pricing Plans Instead of message packages per month, they offer pricing plans. It's unclear what happens when the dialogue limit is reached. A plan comparable to ours costs $149. In August 2023, our price is around $80. However, our current pricing is per response, whereas here, it's per dialogue. It can be argued that this service is more cost-effective for clients. A seven-day trial is convenient and clear. +1. In our experience, clients typically decide whether they will use the service within 2-3 days. The "Removable Footer" likely refers to chatbot advertising and a link to the service. +1 ![Image5](https://cms.wikibot.pro/uploads/Risunok5_1536x983_1e016b67d3.webp) ## Portal Selecting a bot, the current tariff for it, clear dialogue residue statistics with the bot, menu, tariff selection, settings, and data. Very minimalist design. Nothing unnecessary. ![Image6](https://cms.wikibot.pro/uploads/Risunok6_5749c23ba9.webp) ## Customization Messages upon entering the site and a welcome message when opening a dialogue are separated. Separate message customization in the input field. Avatar image is conveniently and clearly designed. Option to show sources, appearing as buttons in a separate block after the response. Coloring bot responses in a user's chosen color. Convenient. Button customization for calling the chat by size and color. System message setting, presumably part of "Prompt," assisting the bot in better performance on a specific site. The default is: "You are a helpful assistant with artificial intelligence. Your goal is to kindly help customers answer any questions they have about our website using the provided context. If the answer does not match the context, just say 'Hmm... I don't know.' Ignore chat history if it's not essential." No language setting for the response. It likely determines the language of the question, and the response is generated accordingly. I added the phrase "Give me an answer in French" to the system message and received a response in French. This is a very flexible tool for individual bot configuration. +1 ![Image7](https://cms.wikibot.pro/uploads/Risunok7_08f6700ca4.webp) ## Message History Viewing interactions with the bot in the form of chat selection and displaying the dialogue as seen by the user. No filters, not even message search. Exporting conversation history to Excel. A simple log of questions and answers. ![Image8](https://cms.wikibot.pro/uploads/Risunok8_3305124066.webp) ## Source Settings Ability to index the site automatically, work from a file, or select specific links and pages. Retraining button available. No option for automatic retraining of the chatbot on a schedule. ![Image9](https://cms.wikibot.pro/uploads/Risunok9_ed56e6f1ab.webp) ## Integration Scripts for embedding on the site. Customization options, various embedding options: as a button or a separate window. Option to share the bot with a separate page through an individual link. ![Image10](https://cms.wikibot.pro/uploads/Risunok_10_1536x878_c42934d81e.webp) ## Feature Requests Requests about user issues and new features are available from the portal. You can see the real-time development process, and a board with the development team's tasks is accessible. Testing GPT responses ![Image11](https://cms.wikibot.pro/uploads/Risunok11_1_1_5da7f3ce84.webp) Seeking information on the site, providing correct links. It does not mislead. ![Image12](https://cms.wikibot.pro/uploads/Risunok12_d85a464e16.webp) Semantic search also works. The site does not have the term "creator," but it has "founder." The AI chatbot responded correctly. ![Image13](https://cms.wikibot.pro/uploads/Risunok13_6776042db2.webp) ## Under the Hood - GPT Handles it well. Managed to create an AI chatbot in 1 minute that understands questions and provides answers, including links to relevant documents. ### Key Considerations and Features: - On the landing page, emphasize the "Create a Bot" button and collection of documentation links. - Bot creation wizard with training and explanations on usage. - Pricing based on plans, not response packages. This is much simpler for the client. - A seven-day trial. A clear history for users. - Footer with information about our service, removed on higher tariffs. - Our chat, embeddable on the site. We integrate with existing popular chats. - System message for flexible AI customization for individual users. - Automatic language detection for responses. - No automatic retraining. - Maximum proximity to users and transparency in development. The service evolves together with clients.

tomleto