Answers grounded in your materials, with a handoff to a person when something needs checking. AI customer support from recurring inquiries to operational records.
Behind every answer,
operating standards you can verify.
Questions covered by your materials receive an immediate answer. Inquiries that need your team’s review are forwarded with the customer’s contact details.
01Find answers in your uploaded materials
02Show the source alongside each answer
03Leave inquiries for your team when needed
These sample conversations illustrate the features.
Actual answers depend on your materials and settings.
When will my order arrive? How do refunds work? What are your opening hours? A person types the same answers every time.
Shipping, refunds, opening hours, stock. The answers are already defined, yet every inquiry still needs a person.
Example: if half of 1,000 monthly inquiries are questions like these, AI could handle 500 of them with the right materials.
Inquiries received after work wait until the next business day. In the meantime, customers look elsewhere.
When that time goes into basic policy questions, important complaints and sales conversations get pushed back.
Example: if one support agent costs KRW 2,500,000–3,000,000/month and handles 60 inquiries a day, the cost is around KRW 2,000 per inquiry.
From a separate support link to a conversation window within your website. Choose the connection that fits where customers reach you.
FAQs, manuals, shipping and refund policies, and your website address. Together with the prepared materials, define what AI may answer and which topics go to your team.Preview the knowledge intake form used for setup →
Configure your company name, tone, materials and answer scope. Check answers against representative questions, then start operating.
Supports popups opened from an existing menu, separate windows, an in-page conversation window, links and QR codes, and a floating button. Review support records in your admin page.
Enter your monthly inquiry volume and handling time per inquiry to see an illustrative estimate of staff time freed up and our service cost.
Illustration · This is an estimate of working time based on the percentage you enter, not a cash saving in labor costs. VAT and time spent maintaining materials are excluded. Our monthly cost includes the base fee and charges above the included allowance (standard setup KRW 0; separately quoted material preparation, customization or integration work is added if commissioned).
Answers in 9 languages, grounded in your materials and in your tone. Questions it cannot handle are directed to your team, and every conversation is recorded in your admin page.
We provide your registered materials to the AI and configure it to answer within that scope, showing source section numbers with answers. If there is no source or the topic is excluded, it directs the customer to your team.
Polite, friendly or concise. You choose the agent’s name and tone.
One set of Korean materials supports replies in English, Japanese, Traditional Chinese, Spanish, Portuguese, German, French and Italian, in the language of the inquiry.
For inquiries AI cannot handle, customers leave their contact details and inquiry, which are forwarded to your team’s email.
Inquiry count, handling rate, common questions and a list of unanswered questions. Input for improving next month’s materials (initially prepared by a staff member).
Usage above the included conversation allowance is deducted from your prepaid balance. We notify you when 50, 80 and 95% of a top-up has been used. If no included conversations remain and the balance is too low to start a new session, customers can leave an inquiry for your team instead of starting a new AI conversation.
Receive an invitation at your admin email address and set a password using the password setup link. Browse full transcripts by date range and unresolved-by-AI filter, download CSV files, and review usage and balance.
AI provider outage handling, handoffs to your team, usage records and alerts are managed together. Before adopting the service, you can review how issues are handled and what operations cover.
| Item to check | How it operates | What you can verify |
|---|---|---|
| AI provider outage handling | We use three providers: Anthropic, Google and OpenAI. If the primary provider times out or fails, we try the configured backup model. Providers and processing conditions are disclosed in the processor list. | Reduces the chance that a single provider outage stops answers. |
| Receiving inquiries during outages | We temporarily pause providers that fail repeatedly and check for recovery. If every candidate fails, we display a retry message and do not bill failed response turns. The route for leaving an inquiry for your team works independently of AI answers. | The inquiry submission route works separately, including during outages. |
| Materials and answer scope | We provide your registered materials to the AI and configure it to answer within that scope, showing source section numbers with answers. If there is no source or the topic is excluded, it directs the customer to your team. Errors can occur, so you define a “do not answer” list for topics where mistakes are costly, such as pricing and refunds. | Answer scope and conditions for team review are defined separately for pricing, refunds and policies. |
| Human handoff | Questions it cannot handle are directed to your team. When customers leave their contact details and inquiry, they are forwarded to your team’s email. The submission route works separately even if the bot stops. | You can also review handed-off inquiries in the admin page. |
| Usage and prepaid balance | Conversation counts and top-up/deduction histories are recorded separately for each client. Only usage above the included conversation allowance is deducted from the prepaid balance, with notifications at 50, 80 and 95% of a top-up used. | Monthly conversation counts and deduction history appear directly in the admin page. |
| Data storage and processing | Data is separated by client-specific paths and stored in the Seoul region. Answer generation uses overseas AI provider APIs (currently Anthropic by default, Google and OpenAI during outages; all in the United States). Conversation records are retained for 90 days by default; adoption inquiries retain only an IP hash, not the original IP address. Personal data is not used for training. | Details of outsourced processing and international transfers are stated in the data processing agreement, processor list and privacy policy, and finalized before adoption. |
| Monitoring and alerts | Every 5 minutes, we check failed answers, provider status, usage limits, abnormal costs and queued emails, and alert the operator. Inquiry acknowledgement and handoff emails are retried automatically if sending fails. | The person responsible for operations receives early signs of outages and acts on them. |
| Operating cost control | We set a daily AI processing cost limit for each client and restrict abnormal usage. The same client’s records and limits continue to be managed when providers or models change. | Usage is checked against the per-conversation rate and balance communicated to the client. |
| Reviewing and exporting records | We send an invitation to your admin email address with a password setup link. Find conversations by date range or unresolved-by-AI status, and review full transcripts, usage and balance. | Download full transcripts as session-by-session CSV files and materials as .md files at any time. Bulk exports are provided as files on request. |
VAT excluded. Prepaid top-ups currently use bank transfers in increments of KRW 100,000 and are credited after payment is confirmed. We issue tax invoices; top-up and deduction histories are available in the admin page.More about payment methods →
300 conversations included per month
1,500 conversations included per month
8,000 conversations included per month
| Billing method | What is counted |
|---|---|
| Token billing | You pay based on the amount of text (tokens) the AI reads and writes. |
| Message or turn billing | You pay based on the number of messages or question-and-answer turns exchanged. |
| Per-resolution billing | You pay based on the number of conversations classified as “resolved.” |
| Flat rate per conversation (ours) | 30 minutes or 20 exchanges count as one unit; only units above the allowance cost KRW 300. Prepayment keeps usage within the amount topped up. |
After 30 days of actual operation, you decide whether to enter the full contract. For the first 3 clients, the base fee is waived for 3 months from service start (D1), including the pilot period, on the agreed plan. Included conversations, overage charges, material improvement scope, and acceptance/termination criteria are confirmed in the quote and agreement before starting. If you stop, the remaining prepaid top-up balance is refunded. Separately quoted setup work is nonrefundable after work starts and fully refundable before it starts.
We describe both features in normal operation and responses when problems arise. Scope of responsibility, cost, data and cancellation are also set out in the contract documents.
AI guidance can contain errors. Questions without a source are directed to your team, and pricing, refund and contract confirmations require your team’s approval. If you report a wrong answer or policy change, we update the materials and answer scope and check the answers again. The scope of responsibility is stated in the terms of service.
Only conversations above those included in the monthly base fee cost KRW 300 each. With prepayment, you are not charged beyond the amount topped up. Once the monthly allowance is used and the prepaid balance is too low to start a new session, we receive inquiries for your team instead of starting new AI conversations.
Provider outages are bypassed through automatic failover. If all providers stop, the support page only displays guidance for contacting your team. Technical support is available on weekdays, 10:00–18:00 KST. For emergencies (the entire support page unavailable), email support@alienpeoples.com with [Urgent] in the subject. Our initial response target is within 4 business hours, or the next business day for nights and weekends. The company’s representative is personally responsible for operations; AI and handoff emails continue working during their absence. There is no numerical uptime SLA commitment yet.
Stored in the Seoul region, separated by client, and retained for 90 days by default. Answer generation uses overseas AI provider APIs (currently Anthropic by default, Google and OpenAI during outages; all in the United States). Personal data is not used for training. Processing conditions are stated in the data processing agreement, processor list and privacy policy.
Cancellation is monthly with no fixed-term commitment. It takes effect at the end of the month requested, and unused top-up balances are refunded within 10 business days from that date. Download full transcripts by session as CSV files and materials as .md files from the admin page; bulk exports are provided as files on request. Separately quoted setup work is nonrefundable after work starts (fully refundable before it starts). There are no other penalties or additional fees.
AlienPeoples. Our B2B customer support service is recruiting its first clients and has no public case studies yet. Operations are led by the company’s representative (15 years in software · 12 years at a telecom company), and we run 3 consumer apps on the same server.
Preview outsourced processing, international transfers, pilot evaluation and operating standards in HTML samples. The documents below are drafts before legal review; each is finalized before contracting to reflect the client’s scope and conditions.
We set up the standard service with your prepared materials and representative questions, check the answers, then start operating. Schedules for system integration or additional customization are agreed separately according to scope.
We configure it to answer within your registered materials and show source section numbers with answers. If there is no source, it directs the customer to your team. Errors can still occur, so you define a “do not answer” list for topics where mistakes are costly, such as pricing and refunds, and your team makes the final confirmation.
A separate support page link can connect to an existing website button, KakaoTalk Channel or QR code. We also support popups opened from existing menus, separate windows, an in-page conversation window and a floating support button. Scope and schedule for internal system integration or custom features are agreed separately.
It is sent only to generate answers, through enterprise AI APIs (currently Anthropic by default, Google and OpenAI during outages—3 providers disclosed in the processor list). Personal data is not used for training. The server masks patterns for resident registration numbers, card and account numbers, passwords, phone numbers and email addresses before storage or transmission. Conversation records are deleted after the default 90-day retention period. Details are in the privacy policy.
The service is monthly. Request cancellation in the admin page; it takes effect at the end of that month, with no billing from the following month, and unused top-up balances refunded within 10 business days of cancellation. Separately quoted setup work is nonrefundable after work starts (fully refundable before it starts).
Prepare FAQs, product/service descriptions, shipping/refund/after-sales policies and your website address. We first review formats such as PDFs, documents and screenshots; if separate preparation is needed, we explain its scope and cost.
Yes. Place the support page link in the Channel’s “Chat” entry. Existing channels such as Channel Talk and telephone remain in place, with AI handling inquiries first.
It replies in English, Japanese, Traditional Chinese, Spanish, Portuguese, German, French or Italian to customers using those languages. One Korean set of materials is enough; we review technical terms together during setup.
Prepaid top-ups currently use bank transfers in increments of KRW 100,000. We credit them after payment confirmation and issue tax invoices. Card payments and automatic top-ups are in preparation; availability will be announced separately. When card payments are offered, the card sales receipt will serve as evidence, and a tax invoice will not also be issued for the same transaction.
Yes. An invitation is sent to your admin email address, where you set a password using the link in the email. In the admin page, browse full transcripts using date-range and unresolved-by-AI filters, and download them as CSV files.
You do not need to send materials yet. Just tell us your company and approximate inquiry volume. We will reply within 3 business days and include the knowledge intake form with the acknowledgement email.