Yes, tell customers when they’re chatting with AI
A short, visible disclosure builds trust, sets the right expectations, and helps customers decide how they want to continue the conversation.
Tell customers clearly when an AI assistant is responding. Identify it as AI at the beginning of the interaction, explain what it is there to help with, and provide a practical route to a person when the conversation requires judgment or individual attention. Disclosure is not an apology for using AI. It is a simple way to set expectations and make the experience more trustworthy.
- Identify the assistant as AI before or with its first substantive response.
- Explain its role in plain language instead of relying on a bot name or icon.
- Give customers a clear way to seek human help when the situation calls for it.
- Keep AI within a defined scope and monitor whether its answers remain accurate.
- Treat transparency as part of the customer experience, not as a legal disclaimer.
Why customers should know they are talking to AI
An AI assistant represents your business during the conversation, even though it is not a person. Customers therefore need to understand what is answering them before they decide what to ask, what information to share, or whether to rely on the response. Clear disclosure gives them that context without disrupting the experience.
AI disclosure is a clear statement that a customer is interacting with an artificial intelligence system rather than a human representative. The best disclosure also says what the assistant is designed to do and how the customer can obtain human help.
Do not assume that a human-sounding conversational style makes identification unnecessary. A friendly tone can improve an interaction, but it can also make the source of the response less obvious. Customers should not have to infer whether an assistant is automated from its writing style, response speed, avatar, or name.
Transparency also establishes the right standard for supervision. Disclosure does not answer the separate question of whether AI should talk to customers unsupervised. That decision depends on the assistant’s scope, the information it uses, the consequences of a wrong answer, and the escalation process behind it.
If a reasonable customer could believe the assistant is human, identify it as AI. Make the disclosure visible at the start rather than hiding it in terms, a help page, or an interface label that customers may overlook.
What good AI disclosure sounds like
Good disclosure is direct, brief, and useful. Start with “I’m an AI assistant” or an equally unambiguous statement. Follow it with the assistant’s purpose, such as helping visitors find information or understand available services. If human support is available, explain how to request it.
- Identify the system: state plainly that the customer is interacting with an AI assistant.
- Define the job: explain the kinds of questions or tasks the assistant handles.
- Set a boundary: say when the customer should seek human assistance instead.
- Offer the route forward: provide the available contact or escalation path.
- Repeat when context changes: identify the use of AI again if a person later hands the conversation back to an automated system.
“I’m an AI assistant. I can help you find information about this business and answer questions based on the information available to me. If you need individual assistance or would prefer to speak with a person, ask for human help.”
The wording can match your brand voice, but the meaning should never become vague. “Virtual concierge,” “digital guide,” or a character name may describe the experience without telling the customer that the system is AI. Use the plain term first; personality can follow.
Disclosure does not inherently make the interaction cold. Customers are more likely to judge the experience by whether it respects their time, answers the question, and gives them control. If your concern is broader than wording, examine whether customers like talking to AI assistants and how to keep AI from making a business feel less personal as separate design questions.
Disclosure is necessary, but it is not enough
Telling customers that an assistant is AI does not excuse weak answers, unclear boundaries, or missing escalation. Transparency is the opening condition for trust; accuracy and responsible operation must sustain it. A disclosed assistant that confidently supplies incorrect business information still creates a poor customer experience.
Accuracy deserves its own operating process. Determine how to check whether an AI tool gives accurate business answers, then test representative questions before relying on it publicly. Review the assistant again whenever important policies, offerings, or source information change.
You should also define unacceptable responses before launch. Decide what an AI assistant should never say to a customer, including where it must avoid unsupported certainty or stop and refer the conversation to a person. Clear boundaries are more effective than expecting customers to detect unreliable answers themselves.
Use stronger human oversight when an answer could materially affect a customer’s money, rights, safety, privacy, or access to a service. Disclosure tells customers what the system is; it does not remove the need for judgment about where automation belongs.
How to introduce AI without undermining customer confidence
Do not present disclosure as a warning that the customer is about to receive inferior service. Introduce the assistant around its job: helping people find information, navigate choices, or reach the right next step. That framing is honest and gives the customer a reason to use it.
- Clear identification
- Names the assistant as AI, describes its role, and offers a route to help. This is the strongest approach.
- Name without explanation
- Gives the bot a branded identity but leaves customers guessing whether it is human. Add explicit AI identification.
- Disclosure buried in policy text
- Documents the use of AI but fails to inform people at the moment of interaction. Move the message into the conversation.
- Repeated warnings
- Reminds customers so often that the experience becomes distracting. Disclose at the start and again when the interaction context changes.
A previous bad chatbot experience can make a business reluctant to try conversational support again. The right response is not to conceal automation; it is to improve the scope, knowledge, interface, and handoff. Consider what has changed since older chatbot experiences before deciding that every conversational system will produce the same result.
Plan for operating effort as well as the initial implementation. Reviewing conversations, updating source information, handling escalations, and maintaining the customer experience all require attention. Include these items when assessing the hidden costs of AI in a small business.
Do not design the assistant to imitate a human employee or let it make unsupported claims about its identity, experience, authority, or ability to complete a task. A customer should never discover only after a consequential exchange that the apparent person was an automated system.
A practical rollout process for customer-facing AI
Start with a narrow, useful job rather than attempting to automate every customer conversation. A focused assistant is easier to explain, test, and supervise. It also makes the disclosure more informative because you can tell customers exactly what the system is there to handle.
- Choose the customer problem: identify a repeatable need that conversational assistance can address.
- Set the scope: document the information the AI may use, the questions it should answer, and the situations it should escalate.
- Write the disclosure: identify the assistant as AI, state its role, and explain the human-help path.
- Test realistic questions: include ordinary wording, incomplete questions, ambiguity, and requests that fall outside the intended scope.
- Check failure behavior: make sure the assistant acknowledges uncertainty instead of inventing an answer.
- Launch with oversight: review interactions and correct gaps in the underlying information or boundaries.
- Maintain it: repeat testing when business information, customer needs, or the assistant’s responsibilities change.
Preventing fabricated responses is an operational priority, not merely a prompt-writing exercise. Use defined source information, constrain the assistant’s role, test edge cases, and establish escalation behavior. A fuller process for stopping an AI assistant from making things up should sit alongside your disclosure plan.
Customers should also be able to reach the business through a channel appropriate to the issue. Human help does not have to be instantaneous in every setting, but the route should be understandable and genuine. If you want to discuss an AI concierge and its customer-facing role, contact the OceSha team.
Where OceSha AI and Lumi fit
OceSha AI is the self-service creation platform of OceSha Ventures, and Lumi is its AI Concierge. Public-facing Lumi helps visitors, prospective customers, partners, and other users learn about OceSha without requiring them to search across multiple pages. It provides a conversational way to ask what they want to know or accomplish instead of first understanding every feature, menu, or workflow.
That role illustrates the right principle: identify the conversational system as AI and give it a defined purpose. Lumi is especially useful to people who want to understand OceSha before signing in. Authenticated Lumi can provide informational and how-to guidance about application pages, workflows, course creation, publishing, analytics, settings, subscriptions, and the OceSha side of integrations.
The OceSha AI self-service platform supports workflows in which user-supplied knowledge can contribute to a course, long-form video can become an episode and short clips, and content can become social material. Published material can attract visitors to an Authority Page, where those visitors can become leads, learners, or customers; analytics can then help the user understand the resulting activity.
OceSha AI can also work with organizations that want to make OceSha capabilities available to members, employees, customers, communities, or other defined audiences. Real courses and branded academy experiences can be explored through OceSha Academy. These are distinct from claiming that every broader OceSha Ventures offering is a self-service capability of the platform.
The organization behind the platform matters when evaluating who builds and operates the wider set of solutions. 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.
Specific external integrations, partner capabilities, and white-label functionality are not detailed here. If one of those areas is important to your intended customer experience, confirm that the required support is available before designing the rollout around it.
Define what your assistant should handle, disclose that it is AI, and build a dependable route to human help before putting it in front of customers.
Plan a transparent AI experienceFrequently asked questions
Does an AI assistant need to identify itself in every message?
Usually, a clear disclosure at the beginning is more useful than repeating the same statement in every reply. Identify the use of AI again if the interaction changes context, a human hands the conversation back to automation, or the customer could reasonably become confused about who or what is responding.
Is giving the assistant a name enough disclosure?
No. A name, avatar, or “virtual concierge” label does not necessarily tell customers that the system is artificial intelligence. Use the words “AI assistant” or another equally clear description.
Should customers be allowed to request a person?
Provide a practical human-help route when a question requires individual judgment, falls outside the assistant’s scope, or cannot be answered reliably. Explain how that route works rather than implying that a person is immediately available when that has not been established.
What information should the AI ask customers to provide?
Ask only for information relevant to the task. Avoid collecting sensitive details through an AI conversation when they are unnecessary, and explain the appropriate channel when a matter requires private or account-specific handling.
How often should customer-facing AI be reviewed?
Review it regularly and whenever its source information, responsibilities, or the business information it discusses changes. Conversation reviews should look for recurring unanswered questions, outdated responses, unsupported certainty, and failures to escalate.
What is Lumi?
Lumi is OceSha AI’s AI Concierge. Public-facing Lumi helps visitors and prospective customers understand OceSha conversationally, without requiring them to search through multiple pages. In authenticated platform contexts, it also provides informational and how-to guidance based on OceSha AI knowledge.
Tell customers when they are chatting with AI, and do it before the conversation becomes consequential. State that the assistant is AI, explain its purpose, and provide a clear route to human help. Then support that disclosure with accurate source information, defined boundaries, testing, oversight, and maintenance. Hiding automation may preserve the illusion of a human interaction briefly, but it creates avoidable risk when the truth becomes apparent. The strongest customer experience is not AI pretending to be a person; it is AI being useful, transparent, and properly supervised.
Rohan Hall
Rohan Hall is a technology entrepreneur, AI architect and author with four decades 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 builds and operates AI-first solutions — course creation, branded academies, AI assistants such as Lumi, and business intelligence — for businesses and organizations.