PostReplyAI customer support starts with the knowledge you can trust.
Give a chatbot the business information it needs, test it on real support questions and keep your team available for the conversations that need a person.
Prepare useful answers before turning on automation
Start with information your team already relies on: product details, service scope, delivery guidance and answers to common questions. Organize it into a knowledge base and make sure each source is current. Contradictory documents and missing policies make it harder for an AI assistant to give a useful answer.
Connect the knowledge base to your chatbot
Create a knowledge base, add sources and wait for ingestion to complete. Link the relevant knowledge to the chatbot and use the testing interface to ask representative questions. The application retrieves workspace knowledge to provide context to the AI model; simply uploading a file does not confirm that ingestion or retrieval succeeded.
- Test questions with a clear answer in your source material.
- Test ambiguous questions and information the source does not contain.
- Review the answer quality before enabling a customer-facing flow.
Make the handover to a person part of the workflow
Decide which enquiries your team should handle directly. For a conversation requiring review, disable AI and assign a teammate. Keep automation focused on questions you can evaluate. AI can make mistakes and should not be treated as a guaranteed source of factual or policy decisions.
Understand what your AI allowance measures
An AI token allowance measures input and output processed by the model, including instructions and conversation context. It is not a count of customer messages or resolved tickets. Longer histories and longer answers use more tokens. Current plans show their monthly allowance on the pricing page; further AI requests pause when the applicable allowance is reached.
Run a small, useful acceptance check
Use a set of common support questions plus a few difficult exceptions. Record whether each answer uses the correct business information, whether an uncertain answer needs a teammate and how much AI usage the test consumes. Provider availability and regional support must also be confirmed for the people and customers who will use the service.
Frequently asked questions
Does the chatbot use my business information?
You can link a knowledge base to a chatbot so it can retrieve relevant workspace information for its answers. Source ingestion and retrieval need to succeed, and you should verify the result in the testing interface before launch.
Can a person take over an AI conversation?
Yes. You can turn AI off for an individual conversation and assign it to a teammate. Define when your team should do this before enabling an automated workflow.
Does one token equal one message?
No. Tokens measure parts of the model's input and output. Instructions, retrieved knowledge and conversation history also contribute to usage, so messages of different lengths can consume very different amounts.
Give your next enquiry a complete customer workflow.
Start with one messaging channel. Connect the CRM, spreadsheet or calendar your team needs and build a workspace around how your business serves customers.
Create your workspace