Transparency

How our AI characters work

Plain answers about who and what you are talking to when you chat with Claire, Vivian, Harper, Grace or Audrey, how their voices are made, what is remembered, and who decides what.

Effective November 2, 2026. This notice is part of our Terms of Service and Privacy notice.

1. Summary

Claire, Vivian, Harper, Grace and Audrey are artificial intelligence. They are not people. Their names, faces, backstories and voices are AI-generated. They answer questions and prepare requests; every booking, charge, refund, hiring step, damage claim and complaint is decided by a human being at Clarté. Their voices are AI-generated and are produced on our own servers, with no third party involved in making the voice. You can type or speak, and you can ask for a person at any time. Memory is off unless you turn it on.

2. Definitions and interpretation

2.1 Defined terms. In this notice: “Clarté”, “we” and “us” mean LIKE A KING INC., a New York corporation doing business in California as Clarté; “AI Characters” means the five fictional, AI-generated personas named in Section 3; “Chat” means the text and voice conversation feature on the website and in the Clarté app; “Model Provider” means the company whose language model generates the text of replies, as described in Section 9; “Voice” means the AI-generated speech used for spoken replies; “Memory” means information that the Chat keeps between conversations, as described in Section 11; “Confirmation Button” means a button in the Chat that you must press before anything is sent to our team or recorded; “Decision” means any determination about a booking, price, charge, refund, credit, claim, complaint, application, hiring step or the terms of service; “Personal Information” has the meaning in the California Consumer Privacy Act of 2018 as amended (Cal. Civ. Code § 1798.100 and following) (the “CCPA”); and “Sister Brand” means FiftyField, an apparel brand that is related to Clarté through common ownership.

2.2 Interpretation. Headings are for convenience only. “Including” means “including without limitation”. A reference to a law includes it as amended or replaced. Where this notice and the Privacy notice address the same subject, the Privacy notice controls on your rights and this notice controls on the description of the AI features. This notice is to be read in favor of transparency where it is ambiguous.

3. Who they are

3.1 The five AI Characters. The AI Characters are fictional. They are AI-generated personas created for Clarté. None of them is a real person, a Clarté employee, a contractor or a representative of a real individual, and none performs cleaning or any other service in a home.

3.2 Handoff. The AI Characters can hand a conversation to one another when another desk is better suited, and they tell you when they do. A handoff is a change of topic, not a change of nature: all five are AI.

3.3 Shared likenesses with our Sister Brand. The appearance of each AI Character is shared with a fictional character in the Sister Brand FiftyField. The shared looks are a deliberate creative choice by the common owner of both brands. They do not mean that the Sister Brand’s characters are employees of Clarté or that Clarté’s AI Characters work for the Sister Brand. Each brand uses its characters for its own communications.

3.4 No real person is depicted. The faces, bodies and voices of the AI Characters are generated and are not copied from, trained on, or intended to resemble any identifiable real person. If you believe that an AI Character resembles you or someone you know, tell us through the contact form and we shall review it promptly.

3.5 They say so. Each AI Character tells you at the start of a conversation that it is an AI, and the Chat shows an “AI” label on every character’s name. We do this whether or not the law requires it.

4. What they do

4.1 Tasks. The AI Characters answer questions about our services, prepare quotes using our posted pricing formula, explain our policies, help you fill in a booking or a request, and prepare requests for a person to act on. They can read information you have given in the conversation and, if you are signed in, certain information in your account that is needed to help you.

4.2 Confirmation Buttons. The Chat can prepare a consultation request, a callback request, a job application, an issue report or feedback. Nothing is sent to our team until you press a Confirmation Button that shows you what will be sent. If you do not press it, nothing is sent.

4.3 What a person does. Every booking, charge, refund, credit, hiring step, damage claim and complaint is decided by a person at Clarté. A person reviews every request that an AI Character files. An AI Character never charges a card and never changes the price that our posted formula produces.

4.4 Allowed actions are limited. The Chat can use only a short list of tools that we have approved. Each tool has defined inputs. The Chat cannot run arbitrary actions, cannot access other clients’ information, and cannot change settings of the systems that run the business. Every use of a tool is written to an audit log (Section 10).

5. What they are not

5.1 No professional advice. They give no legal, tax, medical, insurance or financial advice. If you ask for it, they say so and suggest that you consult a qualified professional.

5.2 No binding promises. Nothing an AI Character says binds Clarté beyond our posted Terms, the Cancellation and refund policy and our posted prices. They cannot promise a refund, a credit, a discount or an exception; a person decides those. If an AI Character states something that conflicts with the posted terms, the posted terms control, and we shall correct the error and, where you relied on it to your detriment in a way the law protects, we shall consider the matter fairly.

5.3 No emergencies. The Chat is not an emergency service. In an emergency call 911.

5.4 No claims of humanity. The AI Characters do not claim to have a body, a home, a family or personal experiences, and they do not claim to be human. They may use a warm manner of speaking, but they do not pretend to be people.

5.5 No impersonation. They do not impersonate any real person, Clarté employee or public figure.

6. Images and likenesses

6.1 Labeled. Portraits and lifestyle images of the AI Characters are AI-generated and are labeled as such where they appear.

6.2 Not our team or your home. They do not show our real team or any real client home. We do not use them as testimonials, as “our cleaner”, in before-and-after photographs, or in anything that suggests a real person performed or endorsed a service.

6.3 No fake reviews. We do not publish reviews or endorsements written by or attributed to an AI Character, and we do not use AI to create reviews, testimonials or ratings (see the Federal Trade Commission’s rule on consumer reviews and testimonials, 16 C.F.R. part 465).

6.4 Shared looks. See Section 3.3 on the shared looks with our Sister Brand.

6.5 Your images. We do not use your photograph to create or train an AI Character, and we do not generate images of you.

7. Voice

7.1 The Voice is AI-generated. When an AI Character speaks, the voice you hear is generated by a text-to-speech system. It is not a recording of a human being speaking the words, and it is not the voice of a real person. Each AI Character has one fixed voice, chosen from a voice set that we run ourselves.

7.2 Made on our own servers. The Voice is produced by software running on our own servers. We do not send the text of a reply, or any audio, to a third-party voice service in order to make the Voice. No third-party company is involved in producing it.

7.3 What is spoken. Before text is turned into speech, we remove what looks like a card number, Social Security number, password or gate or alarm code, link addresses and formatting marks, and replace removed items with a short phrase such as “something private”. Spoken replies are limited in length.

7.4 Audio is not stored, except a short cache. We do not record, keep or listen to audio of your voice through the Voice feature, and we do not store the audio of spoken replies against your name or account. The only audio we keep is a short-lived cache of generated reply audio, mainly Claire’s, so that a reply that is requested again can be played without being generated again. The cache is keyed by a one-way hash of the voice and the text, not by your name, account or device; it contains no recording of you; it is limited in size and the oldest entries are deleted first to make room; and it can be emptied at any time. You can ask us to delete any cached audio through the contact form.

7.5 Your microphone. If you use the microphone button in the Chat, speech recognition runs in your own browser, on your device. Your browser or its vendor may process your speech to turn it into text (for example, a browser may send audio to its vendor’s speech service), under the browser vendor’s own privacy policy and not ours. We receive only the text that your browser produces, and only when you send it. Clarté does not receive or keep the audio of your speech. If you prefer, type instead; the microphone is always optional and is used only when you press the button and allow it.

7.6 Read aloud and captions. Spoken replies are optional and off by default. When a reply is spoken, a caption of the words is displayed. The Chat always displays the written reply.

7.7 Telephone. The Voice is a feature of the Chat. It is not used to place telephone calls. If we ever place a call that uses an AI-generated voice, the Text messages and calls notice applies and we shall say at the start of the call that the voice is artificial.

7.8 Recording laws. California law generally prohibits recording a confidential communication without the consent of all parties (Cal. Penal Code § 632). The Chat does not record your voice. If you use the microphone, you are choosing to speak to the Chat; do not speak other people’s private information into it.

8. No automated decisions about you; no AI in hiring

8.1 Decisions are human. An AI Character does not make a Decision. It may gather the facts you give, prepare a request and explain a policy, and then a person at Clarté decides. We do not use the AI Characters, or any other automated system, to make a decision that produces legal or similarly significant effects on you, such as denying service, setting a price different from our posted formula, refusing a refund, or ending a relationship, without a human reviewing the facts and being able to change the outcome.

8.2 Hiring. No AI Character, and no automated decision system, makes, ranks, scores, screens or influences a hiring, promotion, discipline, scheduling-assignment or termination decision. Grace, the AI Character for careers, answers questions and can prepare an application for you to confirm; a person reads every application and every hiring decision is made by people, in compliance with the California Fair Employment and Housing Act and its regulations on automated-decision systems (Cal. Code Regs. tit. 2, § 11008.1 and following), the Fair Credit Reporting Act and the Fair Chance Act. Consumer-report and background-check steps are described in our employment documents and are not carried out by an AI Character.

8.3 Pricing. Quotes are computed from our posted pricing formula. The AI Characters show the result; they cannot change the formula or grant a discount.

8.4 Profiling. We do not use AI to profile you for the purpose of denying you a service, and we do not use AI to infer your protected characteristics.

8.5 Your right to a human. You can ask for a person at any time by pressing “Talk to a person” (Section 17). You may also ask a person to review any AI-assisted outcome.

8.6 Review of new rules. If California or federal regulations on automated decision-making technology require additional notices or opt-outs for a use that we make of such technology, we shall give them before we make that use. Today we make no such use.

9. Your data and the Model Provider

9.1 What we process. To answer you, we process the messages you type or send by voice (as text), the information you give in the conversation, and, if you are signed in, the account information that is needed to help you. We store the conversation to provide the service and as described in the Privacy notice.

9.2 The Model Provider. The text of your messages is sent to our language-model provider (currently NVIDIA, through its API) only to generate a reply. Under our arrangement with the Model Provider, it processes your messages as our service provider and may not use them to advertise to you. It processes them under its own terms and in its own systems, which are outside our servers. The Voice is not generated by the Model Provider or any third party (Section 7.2).

9.3 What we remove first. Before a message is stored or sent to the Model Provider, we remove what looks like a card number, a Social Security number, a password, or a gate or alarm code, and we ask you to use the secure access-notes field in your account instead. Pattern removal is not perfect, so please do not put such information in the Chat.

9.4 Where it lives. The conversation is kept in your own browser tab (session storage) only if you allow Preferences in the Consent Center. It ends when you close the tab. The cookies notice lists every item.

9.5 Staff review. Staff may review requests and conversations that were escalated to a person, and the audit log (Section 10), for quality, safety and to resolve your request. Staff access is limited to those who need it and is logged.

9.6 Retention. Escalated conversations are kept for 24 months; the audit log for four years; both are deleted after those periods. Conversations that are not escalated are not kept on our servers after the tab closes unless you have turned on Memory. See the Privacy notice, Section 4, for details.

9.7 Sale and sharing. We do not sell your Personal Information and we do not share it for cross-context behavioral advertising. The Model Provider is a service provider and not a third party under the CCPA.

9.8 Your rights. You have the rights to know, delete, correct and limit that the Privacy notice describes. You can exercise them through the Consent Center form or the contact form.

10. Safeguards and human oversight

10.1 Limited tools. Each AI Character may use only the tools its desk needs. A tool that is not on the list is refused, and the refusal is logged.

10.2 Confirmation. Any action that creates a record or contacts our team requires a Confirmation Button that you press (Section 4.2).

10.3 Audit log. We keep a log of which AI Character used which tool, whether it succeeded, and whether you confirmed it. We use the log to detect errors and misuse and to review quality.

10.4 Security measures. We use filters to reduce the risk that someone manipulates an AI Character through a malicious message or link (for example, links in replies are checked), rate limits to prevent abuse, and separation between the Chat and the systems that hold bookings and payments, so that the Chat cannot move money or read other clients’ records.

10.5 Human review. A person reviews requests the AI files, reviews a sample of conversations for quality, and responds to every report of a problem.

10.6 Testing. We test the AI Characters before we change them, including their honesty about being AI, their handling of sensitive information, and their refusal to give professional advice.

11. Memory

11.1 Opt-in only. Memory is off unless you turn it on. If you do not, the AI Characters do not remember you from one conversation to the next.

11.2 Signed-in clients. If you are signed in as a client and choose to let the AI Characters remember you, we keep a memory profile: preferences you state (for example, fragrance-free products), a short summary of notes about your household, and a one-line summary of your last five conversations. You can view, edit and delete the profile under “What Claire remembers” in your account. Signed-in memory is kept until you delete it or close your account.

11.3 Visitors who are not signed in. A visitor who is not signed in is remembered only if the visitor ticks “Remember this conversation on this device”. That sets a random-identifier cookie (aim) and keeps a memory profile for 30 days, after which it is deleted automatically.

11.4 Forgetting. You can say “forget me” to any AI Character at any time, untick the box, or use the delete control in your account. We then delete the memory profile at once.

11.5 What Memory is not. Memory is used only to help you in later conversations. It is not used to set prices, to decide whether to serve you, or for marketing.

11.6 Sensitive information. Memory does not hold payment details or access codes. Please do not ask an AI Character to remember them.

12. Accuracy and limits

12.1 AI can be wrong. A language model produces plausible text and can make mistakes, including about prices, availability, policies and law. Treat what the AI Characters say as a helpful starting point and check anything important with the written terms or with a person.

12.2 Source of truth. The booking confirmation in your account, the posted prices and the posted policies control over anything said in the Chat.

12.3 Correction. If you tell us that an AI Character was wrong, we shall correct the record, and where the error affected you we shall review it fairly with a person.

12.4 Content limits. The AI Characters decline to produce content that is unlawful, harassing or unsafe, and may end a conversation that is abusive.

12.5 Language. The AI Characters are designed to work in English. Translation by an AI may contain errors.

13. Laws we follow

• California Business and Professions Code sections 17940 to 17943 (bot disclosure): we tell you plainly, at the start of every conversation, that you are talking to an AI, and we do not use an AI Character to mislead you about its artificial nature in order to incentivize a purchase or influence a vote.
• Section 5 of the Federal Trade Commission Act (15 U.S.C. § 45) and the FTC’s guidance on claims about AI: we describe what the AI does accurately and make no exaggerated claims about it.
• The FTC’s rule on consumer reviews and testimonials (16 C.F.R. part 465): no fake or AI-generated reviews or testimonials.
• The CCPA as amended by the California Privacy Rights Act, and its regulations: notices at collection, limits on use, service-provider contracts, and your rights.
• The California Fair Employment and Housing Act and the Civil Rights Council’s regulations on automated-decision systems (Cal. Code Regs. tit. 2, § 11008.1 and following): no AI makes or influences an employment decision here.
• The California Invasion of Privacy Act (Cal. Penal Code § 630 and following): the Chat does not record confidential communications and does not wiretap.
• Children’s privacy law, including COPPA (15 U.S.C. § 6501 and following): the Chat is for adults (Section 14).
• Other states’ AI laws, including Colorado’s artificial intelligence act, to the extent they apply to us. We shall follow state AI laws as they take effect.

Nothing in this notice limits a right that these laws give you.

14. Children

The Chat and the Services are for adults aged 18 and over. We do not knowingly collect Personal Information from children. If you think a child has used the Chat, tell us through the contact form and we shall delete the conversation.

15. Accessibility of the Chat

The Chat can be used by typing or speaking, by keyboard, with a screen reader and with spoken replies and captions. Everything the Chat does can be done without it. See the Accessibility statement, Section 6.

16. Worked examples

Example 1: Asking for a refund. You tell Audrey that a visit was poor and you want your money back. Audrey tells you she is an AI, asks what happened, shows you what will be sent to the team, and files a request after you press Confirm. She does not promise a refund. A person reviews your request against the Cancellation and refund policy and answers you in writing.

Example 2: A quote. You ask Claire for a price for a three-bedroom house. She applies the posted formula and shows the result. She cannot offer a discount. If you ask for one, she says a person decides and offers to file the request.

Example 3: Applying for a job. You tell Grace you want to join the team. She explains the role, prepares an application for you to confirm, and files it after you press the button. A person reads it. No AI ranks it or screens it.

Example 4: You put a gate code in the Chat. You type a code. The Chat removes it before it is stored or sent to the Model Provider and asks you to use the secure access-notes field. Because pattern removal is not perfect, you should change the code if you are worried.

Example 5: You turn on the speaker. You switch on “Read aloud”. Claire’s reply is converted to speech on our servers and played, with a caption. We do not keep a recording of you. A short-lived copy of the generated audio may be kept in the reply cache.

Example 6: You use the microphone. You press the microphone and say your question. Your browser turns your speech into text, possibly with help from your browser vendor. The words appear in the message box. You review them and press Send. We receive only the text.

Example 7: Memory. You are signed in and agree to let Claire remember that you prefer unscented products. Next month she mentions it. You open “What Claire remembers”, delete the note, and it is gone.

Example 8: An error. An AI Character tells you a Sunday visit costs nothing extra, but the posted price list includes a Sunday charge. The posted price controls. You tell us, a person reviews it, and where you relied on the statement before you booked, we consider the matter fairly and may honor the stated price.

17. How to reach a person; reporting; escalation

17.1 Talk to a person. Press Talk to a person in any conversation, or use the contact form. A person on the Clarté team reviews every request an AI Character files. We aim to reply within two Business Days.

17.2 Reporting a problem. Tell us through the contact form about any statement by an AI Character that you believe is wrong, unsafe, offensive or misleading, and about any concern about the images or voice. Include the time and, if you can, the words.

17.3 Escalation ladder. (1) The contact form, reviewed by a person within ten Business Days; (2) a written request for reconsideration through the same form, reviewed by a different person within a further ten Business Days; (3) the California Attorney General, the California Privacy Protection Agency or the Federal Trade Commission; and (4) the dispute resolution procedure in the Terms of Service. You may skip any step.

17.4 No retaliation. We do not penalize you for asking for a person or for reporting a problem.

18. Interaction with other documents; survival

18.1 Order of precedence. This notice is incorporated into the Terms of Service. If it conflicts with the Terms of Service, the provision that is more protective of you controls on matters of AI transparency and your rights.

18.2 Disputes. Disputes are resolved as the Terms of Service provide in their section on dispute resolution and arbitration, including your right to opt out and to use small-claims court. Nothing in this notice waives any right that cannot be waived.

18.3 Severability. If any provision is held unenforceable, it is limited to the extent necessary and the rest remains in effect.

18.4 Survival. Sections 9, 11, 17 and 18 continue to apply to information we hold after you stop using the Chat.

18.5 Changes. If we change how the AI Characters work in a way that affects your privacy or the accuracy of this notice, we shall update it before the change and, where consent is needed, ask for it.

19. Version history

20. Contact

Press Talk to a person in any chat, or use the contact form. We do not publish an address, telephone number or email address on this website.