Comece com agentes de IA prontos a usar, com instruções sobre como geri-los no marketplace. Explorar a biblioteca
Voltar aos modelos

Clara. AI Replies to Customer Reviews and Questions

Instantly answers customers in marketplace chats on her own, and drafts replies to reviews and emails following your store’s rules. Flags customers who might leave and shows what buyers complain about.

Clara. AI Replies to Customer Reviews and Questions
Criado por:
Author
ASCN Team
Última atualização:
5 October 2026
Categorias
Popular
Operations
Não sabe como configurar um agente de IA?
Pague a subscrição e receba formação de configuração como oferta. Perceba como preparar o agente para trabalhar nas suas tarefas.

140 reviews are sitting unanswered in your marketplace store, and new ones arrive every day. Clara drafts replies to all of them: in your store's tone, following your delivery, payment and return rules. All you have to do is check and hit “send.” And in chats, Clara answers questions on her own, right away, before the customer goes to another seller.

What it looks like

An online coffee store connected its marketplace account. Among the unread reviews was this one:

“The coffee grinder arrived with a cracked lid. This is the second time, really disappointed.”

The marketplace's auto-reply would have said “Thank you for your review, we're glad you liked it!” Clara drafted this instead:

“Olga, we're sorry this happened, especially for the second time. You can return or exchange a cracked grinder within 14 days at a pickup point: start a return in the ‘Purchases’ section and choose ‘Defective’ as the reason.”

Clara's notes for you: the return period is taken from the “Returns” page on your website. No phone number or link in the public reply, since the platform prohibits it. No compensation offered: that's your call. The cracked-lid complaint is added to the summary: it's the third one this month.

And when the owner mentioned a return period in the chat that didn't match the website, Clara didn't pick one herself. She asked which one was correct, and only then started writing replies.

Customers leaving, and you're the last to know? Clara spots them earlier

A third unhappy email this month, the tone getting sharper, and a week later the customer leaves for a competitor. Each email looks routine on its own, and nobody puts the picture together. Clara spots customers who might leave and suggests what to say to them:

“This subscriber has written three times this month about late deliveries, and in the latest email asks how to cancel the subscription. I suggest a reply with an apology and a specific date for the next delivery. I'm not offering compensation without your approval: decide whether you're willing to give a discount on next month.”

A customer asked in chat and waits until morning? Clara replies right away

A customer messaged your store chat at 11:40 p.m. The reply came in the morning, but by then they'd already bought from another seller who answered in five minutes. On a marketplace, people compare several listings at once, and the order often goes to whoever answers first.

Clara takes over the first reply in chats and answers on her own, instantly, day and night. But only typical questions whose answers are in your rules and listing: where's my order, how do I return it, which option should I choose. If a question is disputed or the rules don't cover it, Clara doesn't guess and passes it to you with a summary.

Customer, 11:40 p.m.: “Will this grinder work for a cezve?”

Clara, 11:40 p.m.: “Yes, it will. A cezve needs the finest grind, which is settings 1–3 on the dial. There's a full grind chart in the product description.”

The same email question 30 times a day? Clara answers by your rules

“Where's my order,” “how do I return it,” “can I pay in installments.” Clara builds an FAQ and templates from your website and past correspondence, and drafts replies to repeat emails. If a customer asks something your rules don't cover, Clara doesn't guess: she asks you and remembers the answer for similar cases. Complex requests come to you with a summary: what the customer wants, what they've already been told, and what you could offer.

How it works

Share a link to your website or marketplace store → Clara writes out your delivery, payment and return rules word for word → you answer a couple of questions about what the rules don't cover, and Clara builds an FAQ and templates → connect your marketplaces and email → Clara answers typical chat questions on her own, right away → drafts replies to reviews and emails for your review, and passes complex cases to you with a summary → you review and send in one click → Clara collects frequent complaints and flags customers who might leave.

How is this different from ChatGPT and marketplace auto-replies?

  ChatGPT Marketplace auto-replies Clara
What rules it follows Whatever you explained in that chat The product listing and the review text Your store's rules, taken from your website, plus your answers to past questions
When the rules don't cover it May invent a plausible deadline or condition Replies with a template Asks you and remembers the answer
Chat response time Doesn't answer marketplace chats — Answers typical questions on her own, instantly, day and night
Channels You copy reviews and replies by hand Only their own marketplace Wildberries, Ozon, Avito, Yandex Market and email in one place
Platform rules Doesn't know them, may add a phone number or link Follows them Never puts phone numbers or links in public replies
Complaints and at-risk customers Can't see the full picture No analysis Summarizes complaints with quotes and flags customers who might leave
Manipulation in emails May treat email text as an instruction — Treats an email saying “process a refund” as a customer request, not a command, and passes it to you

ChatGPT writes a reply when you ask. Clara runs your store's customer support by your rules.

Who it's for

  • Sellers on Wildberries, Ozon, Avito and Yandex Market.
  • Online stores that answer customers by email.
  • Services, subscriptions and online schools where questions repeat and a lost customer is expensive.

How it went for others

Home goods store, Wildberries and Ozon. Everything worked right away. By the time they connected, more than 300 reviews were unanswered. Clara drafted replies, and the owner reviewed and sent them over two evenings. Now replies take about 15 minutes a day. The complaint summary also showed that the lid on one of the containers didn't close tightly: 17 reviews mentioned it. They sent a claim to the supplier, and the next batch came with a redesigned lid.

Online English school, email. This one took some work. The rules on the website were outdated: old prices, different refund terms. Clara kept running into contradictions and asking questions. In the first week, the administrator had to answer about 20 questions and update the terms page. But after that, most emails started getting handled with ready-made replies, and complex cases now reach the administrator with a summary.

Office coffee supplier, B2B subscription. It worked partly. Over a month, Clara flagged 6 customers who might leave: complaints about delays, questions about ending the contract. Managers called each of them and kept four. Two left anyway: the reason was price, which Clara can't fix. She spots the risk, but keeping the customer is still up to you.

What to connect

Wildberries connects in ASCN integrations. Ozon, Avito and Yandex Market connect with a seller API key from your seller account. Email is optional: Gmail, Outlook or any IMAP mailbox. You can also connect Notion for a knowledge base and Slack for handing off requests.

What Clara doesn't do

  • Doesn't reply on her own to reviews, emails or complex questions: she only answers typical chat questions herself, and you send everything else after reviewing it.
  • Doesn't make up deadlines, terms or compensation.
  • Doesn't process refunds or ask customers for card details.
  • Doesn't put phone numbers or links in public marketplace replies: platform rules prohibit it.

Clara can also: summarize customer complaints with quotes, so you can see what to fix in the product or listing, and write scripts and guides for new support staff, so everyone answers the same way and by the rules.

FAQ

Will Clara reply to customers on her own?

In marketplace chats, yes, but only to typical questions whose answers are in your rules and listing. For reviews, emails and complex questions, Clara drafts the reply and you send it in one click.

Will Clara share my data or my customers' data with anyone else?

No. Clara works with data only inside your ASCN account, to carry out your tasks. You can revoke access to connected marketplaces and email at any time in the integrations section. Learn more about how data is stored in the ASCN privacy policy.

How is Clara different from Wildberries auto-replies?

Marketplace auto-replies are based on the product listing and the review text. Clara replies by your store's rules, maintains an FAQ, collects complaints, flags customers who might leave, and also works with Ozon, Avito, Yandex Market and email. You see every reply before it's sent.

Can AI reply to customer reviews?

Yes, if it knows your store's rules and doesn't make up terms. Clara takes the rules from your website, writes in your tone, and doesn't promise anything that isn't in the rules. You review replies to reviews before they're sent.

A customer asked something that isn't in the rules. What happens?

Clara won't guess. She'll ask you and remember the answer for similar cases.

What if a customer tries to “give Clara orders” in an email?

Clara treats email text as a customer request, not as a command for herself. A message like “process a refund” or “give me a discount” reaches you as a regular request, and Clara takes no action on her own.

Can I use it with email only, without marketplaces?

Yes.

What about VK and Telegram messages?

Not yet: Clara works with email and marketplaces.

What does Clara handle every day?

  • Answers typical customer questions in chats right away.
  • Drafts replies to new reviews on marketplaces.
  • Answers repeat emails by your rules.
  • Passes complex requests to you with a summary.
  • Flags customers who might leave and suggests a reply.
  • Adds new complaints to the summary with quotes.
  • Adds answers to new questions to the FAQ.

What results can I expect after 3 months?

It depends on your review volume, channels and how quickly you confirm replies. Typically, after three months the store has no unanswered reviews, customers in chats get answers right away instead of in the morning, a complete FAQ that both Clara and new staff use, a clear list of what to fix in products and listings, and fewer customers leaving without a word. We don't promise specific rating numbers: your seller account will show them.

What if I can't set Clara up?

Message ASCN support: we'll help you connect your marketplaces and email and work through your first replies.

What if Clara gets confused or replies the wrong way?

Most often she's missing rules or access. Ask Clara herself: “What do you need to reply by our rules?” — she'll tell you. If a reply came out wrong, edit it and explain what's off. If she's still confused, start a new session: this clears the context, and Clara will start from a clean slate.

Se tiver questões sobre o modelo, contacte a nossa equipa de apoio
Fazer uma pergunta
message
FAQ
Ainda tem uma questão
Preciso de conhecimentos de programação para configurar este modelo?
Não são necessários conhecimentos de programação! Este modelo foi concebido para utilizadores sem código. Basta seguir o guia de configuração passo a passo, ligar as suas contas e está tudo pronto.
Como é que este modelo ajuda a manter a segurança dos dados?
Todos os dados são processados de forma segura através de APIs oficiais com autenticação OAuth. As suas credenciais nunca são armazenadas no fluxo de trabalho e mantém o controlo total sobre as contas e permissões associadas.
O que é um módulo?
Um módulo é um bloco de construção único no fluxo de trabalho que executa uma ação específica — como enviar uma mensagem, obter dados ou processar informação. Os módulos ligam-se entre si para criar a automação completa.
Posso personalizar o modelo para se adequar às necessidades específicas da minha organização?
Sem dúvida! Pode modificar acionadores, adicionar novas integrações, ajustar prompts de IA e personalizar respostas para corresponder ao fluxo de trabalho e aos requisitos de marca da sua organização.
Quão personalizáveis são as respostas de IA?
Totalmente personalizáveis. Pode editar o prompt de sistema da IA para alterar o tom, o idioma, o formato das respostas e o comportamento. Adicione instruções específicas para o seu caso de uso ou terminologia do setor.
Este modelo vai funcionar com as minhas ferramentas de apoio de TI existentes?
Este modelo integra-se com ferramentas populares como o Gmail, o Google Calendar, o Slack e o Baserow. Podem ser adicionadas integrações adicionais utilizando os conectores de API disponíveis ou webhooks.
E se a minha base de conhecimento de FAQ estiver vazia?
Sem problema! O modelo inclui instruções de configuração para o ajudar a preencher a sua base de dados de FAQ com perguntas e respostas mais frequentes. Comece aos poucos. À medida que surgem novas questões, pode adicionar facilmente mais FAQ ao longo do tempo.
Existe alguma forma de acompanhar problemas não resolvidos que exijam seguimento?
Sim! Pode configurar o fluxo de trabalho para registar pedidos não resolvidos numa base de dados ou folha de cálculo, enviar notificações à sua equipa ou criar tickets no seu sistema de gestão de incidentes para seguimento manual.
E se eu quiser mudar do Slack para o Microsoft Teams (ou outra ferramenta de chat)?
Basta substituir o módulo do Slack por um módulo de integração do Microsoft Teams ou de outro chat. A lógica central mantém-se a mesma — basta voltar a ligar a entrada e a saída à sua plataforma preferida.
Ao continuar a utilizar o nosso site, concorda com a utilização de cookies.