Start with approved facts
Use only current information about products, policies, delivery, returns, and contact routes.
Implementation guide
A practical website AI chatbot setup checklist.
Use this checklist to turn a chatbot idea into a controlled customer-support workflow. It focuses on the details that determine whether an assistant is useful, safe, and easy to improve.
Protect provider access
Keep customer-owned provider keys on the server and use revocable public integration keys for websites.
Design the handoff
Set the questions that need a person and make the escalation route clear before launch.
1. Define the job before choosing the wording
List the repeat questions that genuinely slow down your team or delay purchases. Good first candidates are delivery guidance, returns policy, product basics, opening hours, and service preparation. Do not ask a chatbot to decide refunds, make medical or legal claims, or promise live stock unless it has an approved data source.
2. Build a small, accurate knowledge base
Write concise business facts in plain language. Remove old offers, conflicting policy text, and unverified statements. PlugginAI lets an owner manage knowledge sources so the business can decide what the assistant is allowed to use.
3. Keep credentials out of the browser
A provider API key should not live in a public web page, a WordPress front-end setting, or a chat widget. Save it through the dashboard so it can be encrypted server-side, and use a revocable WordPress or widget key for the website connection.
4. Test the hard questions
Ask the assistant about returns, edge cases, unavailable information, and requests that should go to a person. Check that it says it is unsure rather than inventing an answer. Test the widget on mobile and desktop before publishing it site-wide.
5. Review the workflow
Use the inbox and conversation controls to identify repeat questions, failed answers, and cases requiring human handoff. Improve the knowledge source or routing rule based on real support patterns rather than adding vague instructions.
6. Publish clear customer information
Explain that the business uses an AI assistant, how customers can reach a human, and where the relevant privacy information is available. Review local legal and data-protection requirements for the business and its customers.
Frequently asked questions
What should I put in the first knowledge source?
Begin with the policies and facts your team repeats most often, then review and expand from real conversations.
Should an AI chatbot answer every question?
No. Define when it should hand off to a person, particularly for sensitive decisions or account-specific requests.
How often should I review it?
Review it whenever policies, products, offers, delivery arrangements, or customer questions change.
PlugginAI provides the website, inbox, and secure integration foundation. Businesses remain responsible for provider-account approval, their own policies, and accurate customer information.