Home Getting Started on the Platform
🐾

Getting Started on the Platform

How to set up ClaudIA to test it
Manuela Farias
By Manuela Farias and 6 others
• 7 articles

Guide to Creating Initial Content for ClaudIA

🤖 What is ClaudIA? ClaudIA is the generative AI customer service agent from Cloud Humans. It was created to answer customer inquiries with greater agility and empathy, using a content database from your company. It is powered by language models (LLMs) and trained with content extracted from tickets, FAQs, and documents provided by your team. As a result, it can interpret messages and respond based on the registered content. 🧩 How does ClaudIA respond to requests? ClaudIA operates based on pairs of question/title and answer, which we call IDS (Intelligent Data Semantics). Each IDS is a piece of content that can be triggered whenever a customer asks a similar question. 1. The customer sends a request. 2. ClaudIA performs a semantic search in the IDS database. 3. It identifies the most relevant content and responds based on it. Each piece of content in ClaudIA's database is registered by section (which we call IDS). For each IDS, we configure: - The question/title: which is how the customer should trigger that content. - The answer: which will be sent when the question is asked. - The type of content: which can be: - N1: when ClaudIA responds alone; - N2: when the ticket needs to be directed to a human attendant; - Interactive: Activates automated flows, such as queries to spreadsheets or APIs, using Eddie. - The tag: used to categorize the service. If the chosen type is Interactive, it will be possible to select the corresponding interactive flow that will be activated from that IDS/section. 🏷️ What is a TAG? How does it work? Each IDS content receives a tag, which is used to categorize the service. The Tagger is the mechanism of ClaudIA that automatically defines the final tag of the ticket. How it decides: 1. Identifies which IDS content was used with the most relevance. 2. Applies the corresponding tag: - If it is N2 content → uses the tag from the section that overflowed to N2. - If it is Interactive content → uses the tag from the section that triggered Eddie's flow. - If it is N1 → uses the tag from the most relevant content of the ticket. - If no content is used, the tag will be NO_TAG. 3. The tag must be written in lowercase, without accents, without "ç" and separated by underscores (e.g., incomplete_certificate). 4. The tag is used to generate reports and service metrics. 🧼 How do we create the content database? 1. We extract tickets from your helpdesk (25% of the average volume). 2. We clean the content by removing duplicates. 3. We cluster questions into similar groups (to avoid duplicates), so we know the most common questions asked by customers. 4. After extracting the questions, we build answer suggestions based on: 1. Answers provided by agents in the tickets. 2. Additional provided content (FAQs, training, etc). 5. We generate a Google Sheets with the suggested questions and answers. 6. You validate this content in Sheets, reviewing: - If the question is clear and specific - If the answer is correct and up-to-date - If the content is useful or irrelevant - What is the correct tag for that content - What the type is: N1 or N2 📋 Best practices for creating or validating content - Avoid overly broad questions/titles. Example: ❌ Can I change the order? ✅ Can I change the purchase order after issuance? - Do not use greetings in the responses. ClaudIA does that automatically. - Avoid creating duplicate sections. This confuses the model and harms performance. - Use clear and well-written questions/titles. Correct spelling. - Prefer to direct lengthy instructions to external links. Example of a response that directs to FAQ: You can renew your certificate by following the step-by-step available at the following link: [link]. - You can include specific instructions for ClaudIA directly in the IDS question/title. Example of a question with specific instruction:* [If you decide to use the answer below, always send it complete without changing anything] How to request a refund?* - Make variations of questions/titles within the same section/IDS. Example of how to structure: I don't have an ID at the moment, can I still participate? / I will not have my RG on the day, is it okay to go like that? / If I am without identity, is entry still allowed? 📄 Template for content creation/validation Use this link to access the spreadsheet template used in the creation/validation phase: 🔗 Content Validation Template

Last updated on Aug 12, 2025

Roadmap and Feature Request

This will be the official tool where you can: - See what has already been delivered, what we are building, and what is under review - Suggest new features (Feature Request) - Vote on other suggestions 🔗 Access it here: https://cloudhumans.canny.io/ 🎥 And watch this video explaining how it works: https://www.loom.com/share/026bef6e168143fca2ec464a2923e44c We launched this space to provide more transparency about what we are building and to create a direct channel to hear what is most important to you. 💬 Suggestions and votes will be considered in the construction of the Roadmap in the next quarter—so feel free to contribute! 📌 The page will be updated frequently, so it’s worth saving it to your favorites. ❓ If you have any questions, just reach out to us here. Other Questions: Do I need to create an account to use it? Yes, to vote and create suggestions you need to log in. We recommend using your corporate email, as this allows us to send you updates via email when the features you voted on change status or are completed. What can I see on the platform? You can track the roadmap for both Cloud and CloudShare. You can filter to view only one of the products if you prefer. The features are organized by status: what has already been delivered, what is being developed, and what is under review. How does the voting system work? You can vote on the features you find most relevant by clicking the vote button. Although this doesn’t directly change our priority order, it serves as an important insight for us to make future decisions and understand what is most important to the majority of our clients. How to suggest a new feature? You can create a new post describing your suggestion. You can choose a category (we are still defining these categories better) and detail your idea. When you create a post, it automatically comes with your vote. Are the suggestions really considered? Yes! Suggestions and votes will be considered in the construction of the roadmap for the next quarter. This space was created precisely to provide more transparency about what we are building and to create a direct channel to hear what is most important to you. How frequently is the platform updated? The page will be updated frequently, so it’s worth saving it to your favorites to keep up with the news regularly. Will I receive notifications about the features I voted on? Yes, if you use your corporate email during registration, we can send you email updates when the features you voted on change status or move on to other stages. I have questions about how to use the platform. What should I do? If you have any questions, feel free to contact us directly. We are here to help and ensure that you can make the most of this feedback tool. What is the goal of this initiative? We want to create more transparency about what we are building and establish a direct channel to hear what is most important to you. This helps us prioritize developments that truly add value to our clients.

Last updated on Aug 12, 2025

Setting Up the Context, Identity, and Tone of Voice Fields for ClaudIA

Objective Enable each client to clearly define: - The company context that ClaudIA represents in customer service (details about "what" she is helping with). - The identity of the AI ("who" she is). - And the tone of voice used in responses ("how"). This ensures that ClaudIA communicates with greater consistency, correctly representing the brand and adapting her language as expected. What changed and why does it matter? Previously, ClaudIA's tone of voice and brand description could only be configured during onboarding or by manually requesting the Cloud Humans team. Now, with this new functionality, any client can edit these definitions directly in the Hub at any time. Why is this important? - Prevents dependency on support for simple adjustments. - Ensures that new campaigns, branding changes, or repositioning are quickly reflected in the AI. - Allows testing different tones of voice and uses of emojis that can impact the customer's connection to the brand. - More personalization in responses. - Alignment of the AI with the brand voice. - Improvement in the final customer experience. What was launched Two new editable fields in the Hub: 1. Company and Agent Context Where you can fill in the company name, business context, and description of the AI's identity. 2. Voice, Tone, and Style Allows configuration of: - The style of language and formality. - Preferred and avoided expressions. - Emojis that should (or should not) be used. - Any other nuance related to the tone of voice. Where to configure These fields are available in: Hub → Settings → Base Prompt Modules Examples of filling in Company and Agent Context Example 1 (B2B SaaS): We are Cloud X, a marketing automation platform for small and medium-sized e-commerce businesses. Your name is ClaudIA, and you are our virtual assistant, responsible for helping our clients understand features, solve simple technical problems, and forward more complex cases. You are objective, polite, and demonstrate technical expertise. Example 2 (B2C): Your name is ClaudIA, and you represent FitMama, a workout app for pregnant women. She speaks like a welcoming, experienced, and fun coach. The service should be light, with empathy and clear guidance, even for those who don't understand technology. Example 3 (ClaudIA from Cloud Humans): You are acting as ClaudIA, a conversational AI from Cloud Humans. Cloud Humans is a company specialized in CX/support that offers solutions for customer service through a conversational AI and back-office automations, as well as providing its own Helpdesk, CloudChat. The main products of Cloud Humans are ClaudIA (AI) and Cloud Chat (Omnichannel Helpdesk deeply connected with ClaudIA). 🗣 Voice, Tone, and Style Example 1 (More technical and formal style): You should use clear, technical language without emojis. Avoid informal expressions. We prefer short sentences and objective instructions. Example 2 (Warm and human style): Use welcoming language, without very technical terms. You can use emojis like 💡, 😊, and ✅. Avoid “Dear” and “Best regards” — prefer “Hi!”, “I look forward to hearing from you” or “You can count on me!”. Example 3 (ClaudIA from Cloud Humans): You should respond in a tone that is always friendly yet professional, avoiding jargon or slang. Be very empathetic and make Cloud Humans' products appealing. When writing your responses, avoid forming large blocks of continuous text. Whenever there is a change of topic, conclusion of an idea, or new information, separate paragraphs with a line break (a blank line between sentences). Never concatenate multiple sentences into a single long paragraph. Use few emojis, but when deemed appropriate, use only some of these: 🙂🧡🤖👍 Limits and Best Practices Character Limit - We recommend a maximum of 1,000 characters per field to ensure that the AI's base prompt remains agile and efficient. - Very long texts may cause unnecessary semantic dilutions and affect the consistency of the responses. ✅ What to include - The company name, the sector it operates in, and a brief description of it. - The role of the AI (e.g., attendant, assistant, specialist). - The ideal communication style (e.g., empathetic, technical, playful). - Words, emojis, or expressions that should or should not be used. 🚫 What to avoid - Vague texts (e.g., “speak beautifully and kindly”). - Contradictory instructions (e.g., “be formal and super casual at the same time”). - Defining multiple personas in a single context. Strategic Usage Tips - Adapt to the main channel: If your company uses WhatsApp as the main channel, opt for clearer and more concise tones of voice. - Use to strengthen your brand: ClaudIA is part of the customer experience. An aligned tone of voice generates more trust and identification. - Change when necessary: Launching a new product? Black Friday campaign? Temporarily update ClaudIA's tone to reflect this moment. - Write in the primary language of service: even though ClaudIA has features that allow her to respond in multiple languages, instructions in the most common service language will make responses and adherence to instructions even more effective.

Last updated on Aug 12, 2025

Main Screens of ClaudIA Hub: Playground, Content, Dashboard, Tags, Test Cases, Settings, and Integration Configuration

Playground The Playground is a very useful tool that allows us to test the knowledge base and subsequently audit the ticket generated from that test. With the Playground, we can explore various question-and-answer options, verifying the accuracy of our knowledge base and the adequacy of the provided answers. This allows us to make necessary adjustments to ensure the precision and effectiveness of responses to customers. By using the Playground, we can simulate any question directed to ClaudIA. When she responds, we have access to the information used to formulate that response. This allows us to audit the service, if we deem it necessary. To perform an effective audit, we return to the conversations and remove the filter "indicator: client." This helps us obtain a clearer and more accurate view of the interactions. Content The content screen displays all sections of IDS (Improved Data Set) Some things you can do in Content: - Search and filter by sections - View the modification history of a section - Create a new section - Edit an existing section - Temporarily disable a section - Delete a section from the IDS Dashboard Analytical views of ClaudIA and its results Tag Maintenance This is the screen where we register the tag options that each content can have. This allows us to organize and categorize content efficiently, facilitating access and utilization of information. How to create a new tag to use in your content and enable ClaudIA to use it To create a new tag, simply click the "+ create tag" button at the top right of the screen. This will open a new screen where you can create the tag. If the tag you create does not exist in your helpdesk, it will create a tag with that name directly in your helpdesk as soon as you click the save button dynamically and automatically. 🚨The only exception is for Intercom. In Intercom, we are not able to automatically create tags due to the APIs available to us, by the Intercom platform itself. How tags work for those using Intercom If you use Intercom, you will see that there is a different column called "Helpdesk Tag." This is a mapping of tags between ClaudIA and Intercom. If it does not exist in Intercom, an alert will appear in red saying "Missing." Additionally, when you create a new tag by clicking the create tag button, two fields will appear for you to fill out: the tag field, which is the name of the tag you want to use, and the helpdesk tag field, which is the tag from the helpdesk you want to use. This second field is optional, meaning if there is no existing tag in Intercom, you can leave this field blank, but you will need to go to Intercom and create that tag there. If you do not do this, it will continue to show the "Missing" alert and ClaudIA will not be able to use that tag in Intercom. In other words, the tag will only appear in the Hub. How to edit or delete existing tags If you want to edit a tag, just click the pencil (✏️) button to the right of the tag, under the "Action" column. This will open a pop-up similar to the creation one, where you can change the name of the tag. For Intercom users, you can also change the helpdesk tag it is mapped to, switching to another tag. If you want to delete a session, just click the red trash can icon (🗑️) to the right of the session you want to delete, under the column called "Action." It will open a pop-up asking you to confirm that you really want to delete it. If you want to delete, just click the red "Delete" button; if you want to cancel, just click "Cancel." Test Cases This is a more advanced screen, and detailed documentation can be found here. Settings The general settings screen allows adjustments and customizations in ClaudIA. Some functionalities are not yet available for users, but we are constantly expanding options to make ClaudIA more accessible and configurable in a self-service manner. Integration Configuration (with Helpdesks) This screen is only available to users with privileged access to the Hub (CX Heads, Company Tech Team, etc.). Here you can configure the integration in third-party helpdesks completely self-service without the need for assistance from Clouders.

Last updated on Aug 12, 2025

Conhecendo a ClaudIA e o processo de auditorias

1) Como funciona a ClaudIA ClaudIA é a IA generativa da Cloud Humans, projetada para oferecer atendimento mais humanizado e ágil; Utiliza Modelos de Linguagem de Grande Escala (LLMs), treinados com milhões de textos da internet, livros e artigos; Seu uso é personalizado e restrito à empresa específica que está atendendo; Tem liberdade para adaptar conteúdos, desde que respeite o tom de voz e as políticas da empresa; Ao receber uma solicitação, ClaudIA analisa a dúvida do cliente e consulta sua base de dados, chamada IDS; Realiza uma análise semântica e ranqueia os melhores conteúdos para usar no atendimento. ****Para verificar a base de conteúdo que a IA está utilizando é necessário acessar o Hub da Cloud Humans através desse link aqui (caso ainda não possua acesso basta sinalizar no canal principal do projeto). Após acessar, clique em Conteúdos no menu lateral, como indicado na imagem abaixo: Dessa forma a tela de conteúdos será aberta e será possível realizar tanto a busca exata da informação como a busca semântica - simulando a forma como o cliente pode trazer sua dúvida. Os conteúdos com maior probabilidade de uma resposta correta retornarão no topo. Para testar um atendimento do início ao fim é possível utilizar a opção Playground, também disponível no Hub pelo menu lateral. 2) Erros da ClaudIA durante os atendimentos É importante destacar que por mais completo seja o conteúdo disponível para a IA realizar os atendimentos, ela sempre terá uma % de erro associada ao seu desempenho. Isso acontece por vários motivos, dentre eles: - Pergunta ambígua do usuário: muitas vezes o cliente pode trazer sua dúvida de forma não muito clara, fazendo com que a ClaudIA utilize um conteúdo incorreto ou escale por falta de informações; - Conteúdo desatualizado: aqui se aplicam casos de mudanças na política interna, links ou processos da empresa sem atualizar o conteúdo no Hub; - Falta de conteúdo: lançamentos de novos produtos, promoções e até bugs ou lentidão devem ser inseridos como conteúdo para a ClaudIA ter acesso à nova informação; - Falta de compreensão real: por mais humanizada que a ClaudIA seja, seu funcionamento é feito a partir da identificação de padrões estatísticos entre palavras e ideias, o que pode estar sujeito à erros Aqui na Cloud Humans consideramos que a ClaudIA pode ter 3 tipos de erros, que podem ser classificados durante o processo de auditoria: - Erros do tipo 1 (leves): informação correta, mas faltou algum detalhe que deixaria a experiência mais completa; - Erros do tipo 2 (moderados): informação parcialmente correta, ClaudIA omitiu ou incluiu algum conteúdo sem necessidade; - Erros do tipo 3 (graves): informação totalmente incorreta, casos n2 que ClaudIA não transferiu, etc; O que observamos na média de projetos com boa gestão da IA é que a taxa de erros gerais (somando os 3) não passa de 20%, sendo no máximo 5% de erros do tipo 3. Esse pode ser um bom parâmetro para quem está iniciando o processo de auditorias e precisa de uma meta inicial. Com relação à quantidade de tickets que precisam ser auditados isso vai depender da volumetria do projeto em questão: Aqui trazemos uma sugestão pensando no mês, considerando um grau de confiança de 99% e uma taxa de erro de 5%: - De 20k à 40k tickets no mês: 1000 auditorias mensais; - De 10k à 20k tickets no mês: 800 auditorias mensais; - De 5k à 10k tickets no mês: 600 auditorias mensais; - Até 5k tickets no mês: 400 auditorias mensais; 3) Iniciando as auditorias Uma parte essencial da melhoria contínua da ClaudIA é feita através das auditorias dos tickets respondidos. Essa etapa é fundamental para identificar a falta de conteúdo na base e corrigir possíveis respostas e/ou transferências incorretas durante o atendimento. Para isso acesse nosso Hub através desse link aqui (caso ainda não possua acesso basta sinalizar no canal principal do projeto). Obs: Caso esteja na etapa de auditoria de tickets falsos, não esqueça de retirar o filtro Identificador: Cliente, indicado na imagem abaixo: Os tickets fechados mais recentemente estão no topo. Recomendamos sempre iniciar as auditorias por eles, dado que trazem uma visão mais atual da qualidade do atendimento da ClaudIA. Filtros também podem ser adicionados para para auditar uma amostra específica (ex: por tag), porém, caso esteja iniciando seu processo de qualidade da ClaudIA, recomendamos evitar essa segmentação e auditar tickets de forma aleatória, contemplando tanto casos n1, como n2. Para iniciar a auditoria basta clicar no ícone da coluna Ação, como ilustrado na imagem abaixo: Ao abrir o ticket você encontrará algumas informações, como: - Projeto: nome do projeto, caso sua empresa tenha mais de uma ClaudIA, a sinalização é feita aqui; - ID Helpdesk: é o ID do ticket no seu Helpdesk, caso queira verificar o atendimento internamente; - ID Cloudchat: é o ID no ticket dentro do Hub, é o mesmo presente na URL do ticket aberto; - Criado em: Data e hora do início do atendimento; - Resolução: N2 são todos os casos que a ClaudIA finalizou o atendimento transferindo para um humano. N1 são todos os casos que a ClaudIA finalizou o atendimento sozinha. - TAG: tag aplicada pela ClaudIA no final do atendimento. - Número de interações: Soma das mensagens enviadas pelo cliente e pela ClaudIA - Teste A/B: Caso exista um teste de performance acontecendo, ele será apontado aqui. Abaixo dessas informações estarão as interações ocorridas durante o atendimento entre o cliente e a ClaudIA. Para verificar qual o racional utilizado pela IA e a lista de conteúdos que retornaram em sua busca você pode clicar no ícone i ao lado das mensagens: Dessa forma é possível analisar o que levou a ClaudIA a utilizar determinado conteúdo e realizar os ajustes necessários, caso a resposta esteja incorreta ou não seja aplicável para o cenário em questão: Para ajustar a resposta da ClaudIA é necessário clicar na opção de thumbsdown e indicar o problema na resposta (esse dado refletirá em seu dashboard de métricas, também disponível no Hub): Após isso, algum ajuste precisa ser feito para que no próximo atendimento a ClaudIA atenda como esperado. Veja exemplos de possíveis ajustes: - Deveria ter utilizado outra seção para responder: Caso: Pergunta do cliente estava clara e ClaudIA fez uma seleção errada do conteúdo; Correção: Mudança no título para tornar a resposta mais relevante, para isso utilize o lápis ao lado da seção. Após isso indique que aquela seção é a correta e clique em Enviar para salvar seus ajustes e suba a tela para concluir a auditoria. - A resposta estava incompleta: Caso: Claudia omitiu parte do conteúdo da resposta; Correção: Podemos indicar por [] no Título para que ela nunca omita informações nesse conteúdo específico. - Precisava ter clarificado antes de responder: Caso: Pergunta do cliente estava ambígua e ClaudIA não fez uma triagem antes de fornecer a resposta; Correção: Adicionar conteúdo pedindo quais detalhes gostaria que o cliente informasse. - Alucinação: Caso: ClaudIA saiu totalmente do conteúdo, inventou informações, etc; Correção: Reporte no canal oficial da Cloud Humans para apoio na análise. - Deveria escalar por falta de conteúdo: Caso: ClaudIA não tinha informação e reteve o cliente; Correção: Nessa situação o mais indicado é incluir um novo conteúdo n2 para que a transferência ocorra e de forma mais rápida no próximo atendimento. Para isso clique em Nenhuma das opções para criar novo conteúdo. - Problema no conteúdo da IDS: Caso: ClaudIA escolheu conteúdo certo, mas ele estava desatualizado; Correção: Nessa situação o mais indicado é incluir um novo conteúdo n2 para que a transferência ocorra e de forma mais rápida no próximo atendimento. Para isso clique em Nenhuma das opções para criar novo conteúdo. Extra: ClaudIA não ativou Eddie: Caso um fluxo esteja publicado no Eddie, mas por algum motivo essa seção não esteja sendo ativada vale conferir o título da seção e adaptar as palavras chaves para o conteúdo ganhar maior relevância e prioridade na escolha da ClaudIA. Caso queira saber mais sobre, recomendamos esse artigo sobre o funcionamento entre ClaudIA e Eddie. Obs: caso não tenha certeza de qual ajuste realizar ou seja um caso de alucinação grave, peça apoio por meio do canal oficial do projeto para que um de nossos especialistas ajude nas mudanças necessárias. 4) Motivos de transferência da ClaudIA Existem alguns gatilhos que podem fazer com que a ClaudIA transfira um atendimento. Conhecer cada um deles e entender qual está sendo o mais ofensor na retenção dos tickets ajuda na estratégia para aumento de retenção da IA. O motivo de transferência aparecerá sempre na nota interna, ao final do atendimento, como mostra o exemplo abaixo: Seguem os motivos de transferência da ClaudIA e explicação: - Claudia usou conteúdo N2: dentre os conteúdos que voltaram para a Claudia responder, ela escolheu utilizar um conteúdo N2 para responder o cliente; - Claudia detectou ação de transferência: Cliente foi transferido após Claudia enviar alguma mensagem indicando que iria transferir o atendimento para um agente/suporte; - Cliente Pediu Humano: Cliente insistiu em falar com humano e a Claudia escalou - Claudia pediu detalhes e não entendeu a solicitação do cliente: Claudia fez perguntas de clarificação para o cliente até atingir o limite dessas perguntas - aponta necessidade de ajuste no conteúdo - Claudia não sentiu confiança na resposta: Claudia respondeu sem confiança por 2 ou 3 vezes seguidas - aponta necessidade de ajuste no conteúdo - Claudia não resolveu o problema do cliente: Após Claudia perguntar se pode auxiliar em algo mais, o cliente ainda permaneceu com dúvidas sobre o mesmo problema. - Interação superior ao limite de mensagens: A conversa entre o cliente e a Claudia extrapolou o número máximo de mensagens permitido - Repetição do mesmo fluxo do Eddie: Para garantir que o cliente não fique entrando no mesmo Eddie diversas vezes, ele é escalado antes do fluxo ser acionado pela segunda vez - Eddie Transferiu: Cliente foi transferido dentro de um fluxo dentro do Eddie - Erro na chamada do Eddie: Houve algum erro no Eddie durante a conversa - Escalou por receber anexo: Claudia escalou o cliente após ele enviar algum anexo (imagem, pdf) - Escalado de maneira forçada: Agente retirou o ticket da Claudia - Escalou fora do horário de atendimento humano: Claudia escalou porque não havia atendimento humano disponível (Human Office Hours) - Claudia ficou mais de 10 minutos sem responder (timeout): Claudia escalou após demorar mais de 10 minutos para responder à última mensagem do cliente

Last updated on Aug 11, 2025

Error Validation in Eddie Flows

This article discusses the most common types of errors that should be validated before publishing Eddie flows, as well as how to identify and fix them. Adopting these best practices helps ensure that the flow functions cohesively, all paths are covered, and the end-user experience remains consistent. Automatic Validation in the Editor Eddie flow validation is performed automatically after each change made in the editor, providing real-time feedback on inconsistencies and points of attention. The validation does not prevent the flow from being published, but it is recommended that all issues be reviewed and resolved before the flow goes into production. Known Errors Missing ClaudIA block in branches Each branch of the flow must have at least one ClaudIA block before the end of the path. ClaudIA needs this information to identify that the flow has ended and to take the necessary actions (continue support, resolve N1, escalate to N2, etc.). Missing text associated with ClaudIA block The ClaudIA block exists but has no associated text. The text block can be positioned before or after the ClaudIA block, provided there is no Input block positioned between them. Incomplete Conditions Alerts when conditional blocks (IF, ELSE) are not properly filled out. It is important that all outputs from the conditional are connected to some block, so the flow can proceed without errors. Broken Links Indicates that a link block points to a flow that no longer exists or has not been published yet. This warning can be ignored during flow creation but must be resolved at the time of publication. Missing text between input blocks Occurs when sequences of text input blocks are connected directly without a message between them. To prevent errors in ClaudIA integration, it is necessary to always send a message before any information collection. Conclusion Ensuring that all Eddie flows are validated before publishing is essential to provide a smooth user experience, reduce operational errors, and maintain the integrity of Claudia's logic. With the validations and best practices outlined here, your team can build more robust, predictable, and easier-to-maintain flows.

Last updated on Sep 12, 2025