Where Is My Order?
Most calls are the same question — where is it, how much is owed, has the paperwork been approved. All of it is reading a value off a screen.
Your team becomes a screen-reading service
Pain Points
Everything customers want to know is already in your systems — and the only way in is to wait for a staff member to read it off a screen.
Most calls are the same question — where is it, how much is owed, has the paperwork been approved. All of it is reading a value off a screen.
Your team becomes a screen-reading service
One answer means opening two or three systems — ERP, the courier portal, a finance spreadsheet — while the customer waits on hold.
Longer calls, thinner patience
A traditional auto-attendant can only play back fixed recordings; it cannot pull one customer’s data to answer.
Callers press through the menu and still queue for a person
Customers want to check status at night or on a holiday. The data is already there, with nobody to open it.
The data is ready but unreachable
How it works
Connect the API once; from then on every call gets the same data your staff see on screen.
The customer calls your existing number and asks in natural speech — no menu to press.
The AI asks for the identity details you specify, such as an order number plus the last 4 digits of the phone number.
It calls your API in real time and reads only the fields you have allowed.
Values from your system become words the customer understands, with the time of the last update.
Every query is written to the Audit Log, and calls beyond scope are transferred to a person.
Customer Call
Thai / English · 24/7
Cloud PBX / SIP Trunk
Your existing number
Your Systems
ERP · CRM · OMS · TMS
Access Rules
Identity · fields · limits
Integration Agent
Real-time API query
Live Answer
Value + last updated
Audit Log Entry
Who asked what, when
Escalate to Staff
Out of scope or system down
The AI keeps no advance copy of your customer data — it queries your system only when a real call asks for it.
Solution
Integration Agent is a voice AI agent that connects to your business systems through an API to read customer-specific data and answer it over the phone in real time. Unlike AI Knowledge Base, which answers from static documents, the answers here change with the live data in your system every second — parcel status, outstanding balance, or stock on hand.
IT’s first question is not “is the AI any good at answering” but “if anyone can call in and ask, how do you stop a leak?” — these are the control layers we build in.
Define the identity conditions required before data opens up — an order number with the last 4 digits of the phone number, part of a national ID, or an SMS OTP.
Choose which fields may be spoken aloud and which must never be disclosed. Anything outside the allowlist never reaches the AI at all.
Reading data is the default. Writing back has to be enabled per operation with an explicitly defined scope.
Mask part of a sensitive value when it is spoken — account and card numbers reveal only the last 4 digits for confirmation.
Cap how many queries a single call or caller number can make, preventing repeated guessing to harvest data.
If the source system is down or slower than your timeout, the AI does not guess the answer — it tells the customer and transfers to a person.
IT and internal audit can review which call requested what, whether identity was verified, and which fields the system disclosed — in line with PDPA accountability for personal data.
Key Features
Built so an IT team can connect it without tearing out the existing stack, and keep full control of scope.
Queries your system during the call, with no copy of your data kept on the AI side.
Benefit: Answers always match the screen your staff see
For one question that spans several systems, the AI calls each in turn and merges them into a single answer.
Benefit: No holding while staff open several screens
Turns raw values such as IN_TRANSIT or a status code into a sentence the customer understands.
Benefit: No back-end changes needed to make the AI make sense
Let the AI write back into your system per operation — changing a delivery address, or opening a complaint case.
Benefit: Finish the job on the call, with no handoff to a data-entry team
Set how long to wait for the source system; past that, or if it is down, the AI tells the customer and transfers immediately.
Benefit: A back-end problem never leaves a caller stranded
Connects to the Yeastar P-Series Cloud PBX and SIP Trunk you already use, keeping your numbers and lines.
Benefit: Fast to start, with no number change to announce
Integration
If your system has an API or exposes a callable endpoint, in principle it can connect to Integration Agent.
The primary connection method. JSON supported, authenticated with an API key, and restricted by IP allowlist.
Read document status, purchase orders, and customer data from any ERP that exposes an API.
Check case status, repair jobs, and the contact history of that specific customer.
Order status, payment, and stock on hand from your sales and warehouse systems.
Delivery status, the latest scan point, and the estimated arrival time from your courier system.
Outstanding balances, due dates, and invoice status from your accounting or billing system.
Where no API exists, read from a permission-limited view or read replica [scope to be confirmed].
Connect through the middleware your organization already runs, so core system endpoints stay off the public internet.
Security & Privacy
Integration Agent works directly with customer data, so “keep nothing we do not need” is the default design.
We do not pull your customer database in advance. Data is retrieved only while a call is in progress.
The service and its activity logs are hosted on AWS Bangkok (ap-southeast-7), aligned with PDPA and public-sector procurement policy.
Your customer and conversation data serves your organization only — it is never used to train models or shared across customers.
Voice is encrypted with TLS/SRTP, API calls run over HTTPS, and stored data is encrypted with AES-256 [protocols to be confirmed].
The keys and tokens used to call your API are stored separately and encrypted, and you can revoke them from your side at any time.
Every query is recorded with its identity-check result and the fields disclosed, so auditors can review it afterwards.
Why Call On Cloud
The practical questions IT and procurement teams ask us before they decide.
“Do we have to expose our core system API to the internet?”
Connect through middleware, a permission-limited view, or a channel closed off by allowlist. There is no need to expose the whole core system externally.
“Our IT team is small — how much do we have to build?”
Our team designs the connection, maps system values into spoken language, and tests alongside your team. Your side supplies API documentation and access rights.
“If the AI gives a wrong answer, how would anyone know?”
Every answer cites a value pulled from the live system, and the Audit Log lets you review afterwards what was fetched, when, and under which identity check.
“What if it books or fetches the wrong thing while we start out?”
We connect to a sandbox first, run live test calls with your team (UAT), and tune until it meets the criteria you set — then connect to production.
“Our back end goes down fairly often — what happens then?”
When the source system does not answer, the AI does not guess. It tells the customer plainly and transfers to a person with a summary of the conversation.
“Who do we contact if something breaks at night?”
A local team is on duty around the clock by phone, LINE OA, and email — no waiting on a ticket across time zones.
Use Cases

Logistics
Most calls are “where is my parcel”. The AI verifies identity with the tracking number and phone number, then reads the status and estimated arrival from the courier system. Calls with a real problem go to the customer care team.
The team gets time back for cases that need fixing

Retail & B2B Sales
Shops and dealers call to ask whether an order is approved, when it ships, and whether a model is in stock. The AI answers from OMS and the warehouse system instantly, around the clock.
Sales stops being the middleman for stock checks

Billing & Utilities
Customers call about outstanding balances, due dates, and the latest payment received. The AI verifies identity first, then states the amount with account numbers partly masked per your rules.
Fewer month-end calls into accounting
Plans & Pricing
Pricing is charged per AI concurrent line.
THB 1,000/AIconcurrentline/month
Starts at THB 2,000/month (2-line minimum) · 3-month minimum term
Pricing follows the number of “AI concurrent lines” you need — one line means the AI can hold one call at a time. Add lines to take more calls simultaneously.
Use one on its own, or all of them together in a single account.
Answers customer questions from your company documents — products, pricing, and policy
Answers live data from your systems through an API — order status, outstanding balance
Every Agent runs on your existing numbers and Cloud PBX, on one bill with one support team.
| AI concurrent lines | Monthly fee | Minutes included | Suited to |
|---|---|---|---|
| 2 lines (starter) | THB 2,000 | 400 minutes | Businesses with lighter call volume, starting out |
| 3 lines | THB 3,000 | 600 minutes | Shops or offices with overlapping calls at peak |
| 5 lines | THB 5,000 | 1,000 minutes | Multi-branch or multi-team businesses |
| 10 lines or more | Contact Sales | By line count | Large organizations · call centres |
[To be confirmed: whether unused minutes roll over · monthly billing or 3 months in advance · whether initial design and integration is charged separately]
FAQ
AI Knowledge Base answers from the documents your company uploads — static information such as policies and product specifications. Integration Agent answers from live data in your systems, which changes daily: order status, outstanding balance, and stock on hand. Most organizations run both together.
Mainly a callable endpoint (REST API or webhook), its documentation, and access rights limited to the data actually needed. If your system has no API, our team will help assess alternatives, such as reading from a view or going through the middleware your organization already runs.
The AI will not open customer-specific data until the identity conditions you define are met, and it can disclose only the fields on your allowlist. Anything outside that list never reaches the AI. On top of that, values are partly masked, the number of queries per call is capped, and every query is written to the Audit Log.
No. The service queries your API only while a call is in progress and keeps no advance copy of your database. What is stored is the Audit Log of queries and the conversation records, hosted on AWS Bangkok with a configurable retention period.
You can set a timeout per endpoint. Past that, or if the system does not answer, the AI does not guess — it tells the customer using the wording you define and transfers to a person with a summary of the conversation.
Read-only is the default. If you want write-back — changing a delivery address or opening a case — it is enabled per operation with an explicit scope and stricter identity conditions, and every action is written to the Audit Log.
Typically about 7–14 business days, covering connection design, identity and field-allowlist rules, mapping system values into speech, and live test calls with our team before connecting to production. The timeline depends mainly on how ready your API is.
Pricing is per AI concurrent line at THB 1,000 per line per month, including 400 conversation minutes per line per month. Additional usage is THB 3 per minute. The minimum is 2 lines (THB 2,000 per month with 800 minutes) on a 3-month minimum term.