Resolve customer questions
with Text AI Agent

Customer service AI agents handle customer inquiries in natural language. They help your team automate routine support while keeping people available for harder cases. Text AI Agent uses the website pages, articles, and files you select to answer questions. It can also recommend relevant products and use enabled skills to collect details or create a ticket.

You choose what the agent knows and which actions it can take. When a request needs a person, it can transfer the conversation to your support team with the chat history available in the inbox.

AI-powered customer service uses natural language processing to interpret a question without a fixed menu. Text AI Agent considers the current conversation and looks for an answer in the website pages, help articles, and files you selected.

Answer quality depends on those sources. If a customer asks about a policy that is missing or outdated, update the knowledge base and test the answer. This keeps customer support consistent across routine support interactions. Route requests that need a personal decision to a human agent. This helps avoid asking customers to repeat themselves.

Start with a common support request, such as a question about your service or a request that needs a ticket. Add the knowledge sources the agent needs, then set its language, voice, answer length, instructions, and available skills.

Test real customer questions in preview before launch. Check whether the answers use the right information, whether skills are triggered as expected, and when the agent transfers a chat. Expand to more requests after reviewing live conversations and results.

Start with service requests that come up often and have a clear answer in your help center or knowledge base. Opening hours, account access steps, and service information are good examples.

Choose one request and give the AI agent the right sources. To complete tasks beyond answering, enable a defined skill such as collecting details or creating a ticket. Keep a route to your team for exceptions.

Yes. AI Agent can answer routine support requests outside your team’s working hours using the knowledge you provide. It can collect details or create a ticket when you enable the right skill.

Set a clear after-hours handoff so customers know when a person will reply. Review which questions were resolved and which need follow-up. This extends coverage without implying that a specialist is available immediately.

You can select website pages, help articles, and supported files for each AI Agent. On plans with multiple agents, your team can give one agent product guidance and another a different set of support information.

Use current, specific content that answers the questions customers actually ask. Text supports files including PDF, DOCX, PPTX, TXT, CSV, TSV, and Markdown. Review the sources whenever a policy, service, or product changes.

Keep the pages and files behind your answers current, and review which sources each agent can use. Text can automatically update knowledge from a website source, while uploaded files and articles should be reviewed when their content changes.

After changing a policy or help article, ask the agent a few relevant questions in preview. This is especially useful for time-sensitive support requests, where an outdated answer can create more work for customers and support agents.

Compare tools with real customer questions. Check answer quality, control over knowledge and actions, integrations, human handoff, reporting, and cost. An AI customer service platform should fit your customer service operations and existing support tools. Check the key features your team will actually use.

Text brings AI and human agents into one inbox, so your team can see the conversation when a person is needed. Test its agent builder, knowledge setup, defined skills, and reports against your customer needs before choosing a plan.

Text plans include a monthly AI Agent resolution allowance. A resolution is a conversation in which the AI Agent directly solves at least one customer question. A test, spam message, greeting, or transfer without an answer does not count as a resolution.

The allowance and cost of additional resolutions depend on your plan. Check current pricing before estimating your budget, then use your expected volume of customer inquiries to choose a plan.

AI support agents handle repeat questions and defined tasks. Human agents still need to handle complex or sensitive issues, exceptions, and conversations that call for a personal decision.

Your team can supervise multiple customer interactions and take over when needed. A clear handoff keeps the conversation history in view, so customers do not have to start their story again.

You choose which website pages, knowledge base articles, and files each agent can use. Keep those sources current, set instructions for the agent’s behavior, and test real customer queries in preview before launch.

After launch, review unanswered questions and transfers. Update a source if information is missing or outdated, then test again. For requests outside the agent’s knowledge or approved actions, make it easy to reach your team. Human intervention is better than a confident but unsupported answer.

Custom Skills let you define an action for a customer intent. A skill can collect missing details, create or update a ticket, tag or transfer a chat, or call a connected webhook or API. You choose the inputs and the permitted action.

For example, the agent can ask for the details needed for a support ticket, then create it. Test the trigger and each step in preview. After launch, use the Runs view to see whether the skill completed the task or needs adjustment. That makes workflow automation reviewable instead of hiding failures.

Text combines AI conversations, live chat, and tickets in one workspace. Custom Skills can connect AI agents to core business systems through a webhook or API for a defined action, such as looking up a record or updating a ticket.

For multi-step workflows, decide what information the agent needs, which action it may take, and what happens if the system is unavailable. Test with realistic service requests so your team has a clear place to take over. This is especially useful when connected systems are unavailable or return incomplete data.

Depending on the skill you configure, it can use details already available in the customer profile or conversation, such as the latest message, chat history, and current page. Connected tools can provide other information needed for a defined task.

Choose only the information the skill needs. For example, a ticket creation skill may use the customer’s question and the current chat, while a separate skill could look up a record in a connected system.

It can use the customer context available to it, such as the latest message, current page, and details in the conversation. A configured Custom Skill can also use relevant profile or ticket information. With a connected system, it may be able to retrieve the record needed for a defined request. What the agent can see depends on the sources, integrations, and skill inputs you choose.

Personalized support should help answer the customer’s actual question. For example, the agent can avoid asking for a detail already present in the chat or collect only what is missing for a ticket. Keep the data relevant to the task and let a person step in when context is incomplete or the request is sensitive. That protects the customer experience and customer relationships while offering personalized support.

These actions are not automatic. A Custom Skill would need access to the relevant system through a webhook or API, along with instructions that define when the action is allowed. The available action depends on that system and the workflow you configure.

With a connected system and explicit rules, an AI Agent can complete a permitted refund or account update. Test the skill with representative requests before using it for sensitive changes. Set limits and a transfer path for cases the agent should not complete, such as an exception that needs a person’s approval.

Set handoff rules for complex or sensitive issues, policy exceptions, requests for a human agent, or questions the AI cannot answer reliably. Text can transfer the chat to an available teammate or chosen team, depending on your setup.

The person taking over can read the conversation and customer details in the inbox. Review these transfers to improve knowledge gaps and confirm which requests need human judgment. A clear point for human intervention protects service quality.

Yes. Your team can see active AI chats, follow what the agent is saying, and take over a conversation. The conversation stays in the inbox, so the person joining can read what the customer and the AI Agent have already discussed.

Use this to check service quality during rollout and to step in when a customer needs personal attention. It also gives you examples to review when improving answers, skills, and handoff rules.

Your team can bring website chat, email, Facebook Messenger, WhatsApp, and SMS into a shared inbox through the relevant connections. Support agents can see the customer conversation and reply from one workspace.

Channel availability and setup depend on the integration. Check the connection you need before launch; bringing a channel into the inbox does not by itself mean every AI skill is available on that channel.

You can choose the languages an AI Agent uses and set its tone and answer length. Test common customer inquiries in each language you plan to support, especially names, policy terms, and questions that rely on your own knowledge sources.

For multilingual support, keep the underlying information clear and current. If a request needs a specialist or the agent cannot give a reliable answer, let the conversation move to your team.

Text reports AI resolutions, resolution rate, chats handled, customer satisfaction, and transfers. Read these alongside actual conversations: reply volume alone does not show whether a question was solved.

Look for repeat transfers, unanswered questions, and skills that did not complete. Compare results with your earlier support workload to assess support efficiency and operational costs using your own baseline. Review customer satisfaction too; faster replies alone do not prove a better customer experience.

Give the agent clear knowledge, a focused role, and limits on what it may do. Test vague customer queries and questions outside its sources. If it cannot make progress, route the conversation to a person instead of asking the customer to repeat it. That avoids a frustration loop in which the AI keeps giving the same unhelpful answer.

Review repeated questions, low satisfaction, and transfers. Update the knowledge base or skill when needed, and keep exceptions with your team.

Introduce it clearly as an AI agent and explain what it can help with. Use a name and greeting that do not pretend it is a human agent. Give customers a clear way to ask for a person when their question needs one.

There is no established group of four AI customer service agents. Lists vary by reviewer and by the type of service a business needs.

Compare products using your own customer inquiries. Check key features such as knowledge controls, supported actions, integrations, human handoff, reporting, and pricing rather than choosing from a fixed ranking.

A scripted chatbot follows conversation paths written in advance. An AI agent can interpret an open question in natural language, find an answer in selected sources, and choose an enabled skill for a defined task.

Both need clear limits. AI agents handle more open-ended questions, but defined skills and connected automation tools are still needed to complete tasks. If the request is outside those limits, the customer should be able to reach a person without repeating the conversation.

It can answer product questions from the information you provide, recommend relevant items, and collect details for your team. With Shopify connected, shoppers can choose a product variant and add it to their cart in chat. With an order system connected through a defined skill, the agent can look up an order status when that data is available.

Your team can see the shopping conversation and its context in the inbox. Reports can help connect conversations with sales, alongside the service results you track.

No, your LiveChat plan stays the same. The features and capabilities on this page are part of Text and are available under Text pricing plans. They are not included in LiveChat plans, including legacy LiveChat subscriptions. Availability, limits, and pricing depend on your Text plan. If you’re on a LiveChat plan and want these features, you can move to Text.