Your accounts, throughout
What runs where.
First-party code on your existing page posts the answers to Google Apps Script deployed under your own Workspace account.
Your page
Collects the visitor's answers on the site you already run.
Your Apps Script
Receives the submission under your Google Workspace account.
Your Sheet
Records the submission before an email is sent.
Your Gmail
Sends the notification from your domain to the inboxes you chose.
One company appears in that path: Google. It is the one you already chose for Workspace. Google receives the message under your Workspace agreement. No chat vendor is added, and no copy is routed to Ladus.
Why build it this way?
The intake is not a rebranded seat on a chat platform. You own the deployment, configuration and data; the software is licensed to you. No code path sends a submission or a copy outside the Workspace account it was built for, including to Ladus.
No added per-seat or per-conversation fees apply to the intake software. Your existing website and Google Workspace costs continue.
We considered a hosted chat widget and read its published terms before choosing this design. They did not support the data protections we were willing to promise, so we left that vendor out. This explains a software design choice, not the terms of your own agreement.
When someone submits
Checked.
Recorded.
Then notified.
Check the request
The submission is validated before it is written. Malformed or hostile input is rejected rather than stored. Recent submissions are checked so a double-tap can be recognized as a repeat, instead of producing a second request.
Save the record first
The script writes the row to the Sheet before sending the notification. A later mail failure cannot erase a saved record. Duplicate checks consider the full message and its length, so two long messages with the same beginning are not treated as identical.
Confirm what actually happened
The visitor receives a confirmation describing the outcome, rather than a generic thank-you that says the same thing whether the submission worked or failed.
The clock matters
The right expectation.
At the time they write.
The assistant uses your staffed hours and timezone, including timezone transition days. Inside those hours it says a person will reply. Outside them it sets a different expectation.
Wording that becomes ambiguous
“Tomorrow at 9 a.m.”
Written Sunday night but read Monday morning, “tomorrow” now sounds like Tuesday.
An explicit day
“Monday at 9 a.m. Eastern.”
The intended day remains clear when the message is read later.
Illustrative wording and hours. Messages that outlive the moment they are written use an explicit day; beyond six days, a date.
The assistant does not direct someone to an unstaffed phone after hours. A voicemail nobody will hear before Monday is another dead end.
Evidence has a limit
A saved record.
A received email.
Two different checks.
The logic is tested
Hundreds of automated checks are counted by the test gate, rather than an exact count maintained on this page. Fixes are tested by reverting them on a throwaway copy and requiring the relevant check to fail.
Checks run at fixed times on both sides of opening and closing boundaries. The behavior at 7:59, 8:00 and 8:01 is tested, not left to whichever time the suite happens to run.
The deployment still needs checking
Those checks do not prove a deployment. The runbook names the checks a person must perform by opening the real inbox and looking.
A notification in spam was delivered, but may never be noticed. That is why a delivery check cannot replace looking in the destination inbox and spam folder.
The saved spreadsheet row and the email notification are separate outcomes. The confirmation must describe what actually happened.
After delivery
It has to run
without us.
Software nobody but its builder can run is a liability. The response is a documented handover: an administrator runbook and rollback procedure written for somebody who is not an engineer.
- You own the data, deployment and configuration.
- You receive a perpetual license to run it on the site it was built for. Ladus keeps the underlying software.
- The instructions explain how to operate, change and turn it off.
- Fourteen days of corrections after launch are included at no charge.
Ongoing monitoring, hosting or support; reservation, booking or scheduling integrations; and additions to the scope after delivery are quoted separately.
From $4,500, with a fixed fee agreed in writing before work starts. Payment is on delivery. The price changes only if you change the scope in writing.
The boundaries
It routes to people.
People answer.
The assistant does not compose answers on your behalf. AI-written replies are a different product with a different risk profile: an answer can be confidently wrong, in your voice and on your domain.
It also cannot make a slow queue move faster. Somebody must act on what it captures. If nobody will, fix that first and keep your $4,500.
You need a Google Workspace account you control and an existing website that can accept a code snippet.
Ask the people building it
What would this look
like on your website?
Email your website, what visitors ask and who answers today. We will ask questions by email and tell you whether we can help. No call unless you want one.
Ladus AI Development, LLC is not a law firm, does not provide legal services, and no client-lawyer relationship or its protections arise from anything here.