Yes—AI can answer customer questions accurately when its knowledge is deliberately constrained. Lumi’s website concierge answers from the business’s approved public knowledge and provides no knowledge-based answer before that material is approved. It supports text chat in 12 verified languages, while website discovery can collect content from up to 25 public pages for review. Businesses should still test important, ambiguous, multilingual, and policy-sensitive questions before relying on any AI concierge.
- Customer Lumi answers from approved public knowledge belonging to the business, rather than responding from unrestricted general information.
- Discovered website content enters the knowledge workspace as pending material; it must be approved before Customer Lumi uses it.
- Text chat has been verified in 12 languages, but businesses should test their own terminology, policies, and mixed-language questions.
- Lumi records conversations so staff can identify unanswered questions, confusing wording, and gaps in approved coverage.
- Connected SHOPLINE stores can provide synchronized catalog information, but Lumi does not answer order-status questions.
Reliable AI answers begin with a controlled source of truth
The central question is not whether an AI model knows a great deal. It is whether the customer-facing assistant is restricted to information the business has selected and approved. Customer Lumi, the website-facing part of Lumi—the AI Concierge for your website, answers from that business’s approved public knowledge. It provides no knowledge-based answers before content has been approved. That is the foundation for reducing unsupported answers.
This approach is more dependable than asking a general-purpose assistant to infer a company’s services, policies, availability, or product details. The wider discipline is covered in helping customers get reliable business answers: decide what visitors are allowed to receive, organize it clearly, approve it, and then test how the concierge handles real wording.
How Lumi turns business information into customer answers
Lumi gives businesses a private workspace for knowledge management, conversations, leads, first-party analytics called Customer Intelligence, team management, billing, and Business Lumi, a private AI assistant. Customer Lumi is the separate visitor-facing concierge delivered through the business’s website. This distinction keeps public answers focused on approved public material while staff work privately with business information and customer interactions.
- Discover — Lumi’s website discovery reads up to 25 public pages and organizes the material into pending knowledge items. This has been verified across six businesses.
- Review — Staff inspect the discovered material, correct stale or ambiguous wording, and decide what belongs in public answers.
- Approve — Only approved public knowledge becomes available to Customer Lumi. Pending material is not automatically presented as an approved answer.
- Publish — The business adds one script to load Customer Lumi as a floating, mini-card, or embedded website experience. The domain must be registered and enabled; wildcard domains are not supported.
- Observe — Conversations are written to the business workspace, giving staff material they can review for knowledge gaps, unclear policies, or questions that need a human response.
- Improve — Staff update and approve knowledge, then retest the questions that exposed weak or incomplete coverage.
Website discovery accelerates preparation, but it is not a substitute for review. Public pages can contain expired offers, duplicated explanations, or language that makes sense in context but becomes ambiguous when quoted alone. Use approved-information welcome-question preparation to shape useful starting questions without treating every sentence on a website as ready for an AI answer.
What Lumi can answer—and where it should hand off
Customer Lumi supports working text chat through a mini-card launcher, a floating interface, or an embedded experience. Its text answers have been verified in 12 languages. Automatic multilingual suggested questions help visitors begin, while rich cards can present products or offerings selected through references in approved knowledge or a connected catalog. Images in public cards come only from business-owned sources such as its website, merchant uploads, or SHOPLINE; they are not AI-generated.
Current customer-facing capabilities
For a specialist business, breadth is less valuable than defensible coverage. Test whether the approved material actually addresses the questions customers bring, using a process such as assessing approved answers for specialist consultancy questions. Where staff repeatedly encounter missing context, capturing customer preparation obstacles for staff review turns those conversations into a practical knowledge-improvement queue.
Accuracy requires testing language, policies, and voice separately
A correct source can still produce a weak customer experience if the question is vague, uses unfamiliar terminology, combines languages, or refers to an exception that the approved material never explains. Testing must therefore cover both factual support and the ways real visitors phrase questions. Start with routine wording, then test shorthand, misspellings, follow-ups, and requests that should be handed to staff.
Voice needs its own test plan. Voice input converts speech to text and has been verified in English, Traditional Chinese, Japanese, and Spanish. Accepted audio is WAV at 16 kHz with a 20 MB cap. Occasional Traditional Chinese mishearing has been observed. Voice output has been verified in the same four languages, with one voice configured per business and language controlled through instruction.
What each test proves
Freshness deserves particular attention in retail and service businesses. Keep unapproved campaigns away from customer answers by separating draft retail promotions from public knowledge. For sensitive policies, discreet text interaction can also be useful; see how to ask store-policy questions discreetly using text.
Conversation records create an accuracy improvement loop
No initial knowledge set anticipates every customer question. The advantage of a managed concierge is that conversations are stored for the business, creating direct evidence of what visitors ask and where approved content needs improvement. Staff can look for unanswered questions, repeated follow-ups, ambiguous terminology, and requests that consistently need a human response.
- Review recurring questions and identify which ones were answered clearly from approved knowledge.
- Separate a missing fact from a poorly worded fact; adding more content will not fix contradictory wording.
- Update the relevant knowledge item rather than creating several overlapping versions of the same policy.
- Approve the revised public information before expecting Customer Lumi to use it.
- Retest the original question, a paraphrased version, and an exception case.
- Continue routing account-specific, order-specific, or unsupported questions to staff.
Lead capture can support that handoff. When enabled by the business, the in-chat form collects a visitor’s name, email, company, phone, and challenge. Email notifications can go to as many as five addresses. Optional high-intent SMS depends on a configured provider. This gives visitors a next step when a reliable answer requires personal follow-up instead of an improvised response.
Lumi can also open the business’s scheduling URL and record a booking-start event. That event means the visitor opened the booking path, not that an appointment was completed. Confirmed booking data through Calendly or Google Calendar is partial, comes from an earlier OceSha Ventures system, and has no verified self-service setup for current Lumi merchants. If calendar connectivity is essential, confirm support for the business’s setup with the Lumi team.
For businesses in Orange County, Los Angeles, and San Diego, CA
In these busy Southern California markets, service providers, retailers, and specialist businesses often face customers comparing several options online before making contact. Questions about availability, pricing, service areas, policies, and multilingual support can quickly expose outdated or unclear website content. The first practical step is to review the public pages customers rely on most, then approve only current, defensible information for the concierge to use. Before launch, test common questions alongside ambiguous wording, policy exceptions, and realistic Spanish-language or mixed-language queries. Conversation records can then reveal where the approved knowledge needs clarification or expansion.
Who stands behind Lumi and what businesses should evaluate
Lumi is the AI Concierge in the OceSha ecosystem and is also the AI Concierge for OceSha.ai. It is a product rather than a separate legal entity or seller; billing is handled by OceSha Inc., and the brand relationship is shown as Powered by OceSha. OceSha Ventures builds and operates AI-first solutions spanning course creation, branded academies, AI assistants such as Lumi, and business intelligence for businesses and organizations.
When evaluating Lumi, do not stop at a generic demonstration. Bring the questions that create risk for the business: policy exceptions, specialist terminology, multilingual wording, product variants, draft-versus-current information, and requests that should go to staff. Check the answer against approved material, inspect whether the source is current, and confirm that unsupported requests receive an appropriate handoff.
Lumi supports subscriptions and trials, with monthly AI credits and optional credit top-ups. A tested top-up provides 100 credits for $12. Because voice and other usage consume credits, evaluate the expected mix of text, speech, and visitor volume rather than judging the concierge only by its interface.
Accuracy comes from controlling what the AI is allowed to use, reviewing that knowledge before publication, and testing the questions customers actually ask.
Evaluate Lumi for your websiteFrequently asked questions
Does Lumi use unapproved website content in customer answers?
No. Website discovery organizes public-page content into pending knowledge items. Customer Lumi provides knowledge-based answers only after that material has been approved as public knowledge.
Can Lumi answer in languages other than English?
Yes. Text chat has been verified in 12 languages. Voice input and output have been verified in English, Traditional Chinese, Japanese, and Spanish. Businesses should test their own terminology and mixed-language questions.
Can Lumi answer questions about products in an online store?
Yes, when the business connects a SHOPLINE store. Lumi can search the synchronized catalog, help select a variant, show items, and manage the real store cart. It does not answer order-status questions and has no dedicated product-comparison tool.
Can staff see what customers asked?
Yes. Conversations and messages are written to the business’s private workspace. Staff can use them to find unanswered questions, improve approved knowledge, and identify situations that need human follow-up.
Does Lumi confirm that a visitor completed a booking?
Opening a scheduling link creates a booking-start event, not proof of a completed appointment. Confirmed booking support through Calendly or Google Calendar is partial and has not been verified as a self-service setup for current Lumi merchants.
Can Lumi collect contact details when it cannot complete the conversation?
Yes, if the business enables lead capture. The in-chat form supports name, email, company, phone, and challenge. Lead emails can be sent to up to five addresses; optional SMS requires a configured provider.
The bottom line
AI can answer customer questions accurately when the business controls the source material and treats accuracy as an ongoing operating process. Lumi follows that model: Customer Lumi answers from approved public knowledge, stores conversations for review, and supports structured handoffs when the answer should come from staff. The right approach is not to trust fluent output automatically. Approve current information, test real and multilingual questions, review gaps, and keep draft, private, order-specific, or unsupported material outside public answers.
Rohan Hall
Founder of OceSha Ventures · AI architect and author
Rohan Hall is a technology entrepreneur, AI architect and author with years of technology experience, now focused on practical AI across business, education, government and global impact. He founded OceSha Ventures, builds the OceSha AI platform and Lumi, and wrote The Convergence of AI and the Top 10 Emerging Technologies.
OceSha Ventures — OceSha Ventures builds and operates AI-first solutions — course creation, branded academies, AI assistants such as Lumi, and business intelligence — for businesses and organizations.
Sources
About this page
It is based on Lumi's current approved product information. Product details come from Lumi's current product capabilities audit (Oct 2026).