Start with your website
Scan supported public pages and review the generated knowledge. Check the wording, missing policies and product details before publishing answers your customers will rely on.
Useful AI starts with your knowledge, remembers eligible customer context and knows when your team should step in. Give Message AI the information, rules and specific jobs that make it useful to your company.
Customer-facing AI follows your workspace plan. Task agents have separate pricing.Your website, help articles and approved connections give Message AI a place to find the facts. You decide what becomes usable knowledge.
Scan supported public pages and review the generated knowledge. Check the wording, missing policies and product details before publishing answers your customers will rely on.
The onboarding interview asks about your business and follows up based on the answers. Short responses can become draft articles that you can review and improve.
Connect supported stores, documents or selected spreadsheet ranges. Preview the imported content and choose what belongs to this business before it becomes part of retrieval.
A returning customer should not get a brand-new greeting halfway through an ongoing issue. Useful memory comes with a clear understanding of whose information can be used.
Earlier turns and eligible customer history help AI follow the issue, preferences and prior promises. Linked chat, ticket and call context can support continuity across visits.
Supported order flows check email possession and matching order details before retrieving the order. Simply knowing a customer’s email address is not enough to request their private purchase information.
Knowledge, connected sources and company guidelines are scoped to the business. A rule taught in one workspace should not become another company’s policy.
Create focused helpers for the work your business repeats. Describe the task in plain language, choose the product and knowledge scope, and try it before activation.
Answer from knowledge, summarize a conversation, classify a request or extract action items. A focused instruction gives you a clearer result to evaluate than a vague request to handle everything.
Choose the supported product and business knowledge. Site-specific agents stay with that site’s conversations. Summaries and extraction results stay internal instead of changing connected customer systems.
Run a sample task before putting an agent to work. Review runs and usage, set a daily limit and pause an agent when you need to adjust its instructions.
The quality of the answer matters. So do the controls around it.
Give AI clear directions for tone, escalation and business policy. Internal coaching can produce a proposed guideline that an authorized person reviews and saves.
Useful completed conversations can suggest new questions and answers. Owners review the suggestions before they become published knowledge.
Agents can see estimated frustration, repeated requests for a human and manager requests. These are assistance signals for human judgment, not facts about a customer’s character.
Route a request to the team when a person is needed. A handoff should preserve the conversation and its context so the customer does not have to start over.
Ask internal AI to notify a teammate about a chat or callback. Include a conversation reference and a due time; acknowledgement or completion stops the relevant reminders.
Keep answers relevant to the business. Requests for unrelated tasks or private customer data should be handled without disclosing information or silently losing the customer’s support request.
Scan the business website or answer the onboarding interview. Add approved sources, then review and publish the material AI should use.
Choose where AI answers and where it asks for help. Define company guidance, human escalation and the permitted scope of connected data.
Try a returning customer, an unknown answer, an order lookup and a request for a human. Review outcomes with your team and refine the knowledge or guidance.
Customer-facing AI follows your workspace plan. Task agents have separate pricing.