Incomplete inquiries
A customer mentions bedrooms and a date, but leaves out location, home size or condition. The assistant helps you see what still needs asking.
For independent cleaning business owners
Organize customer requests and prepare replies for review. This demo focuses on residential cleaning inquiries and your approved service rules.
You decide the scope, price, availability and whether to send.
Pilot-stage service. No inbox connection is needed for the demo.

A focused starting point
This workflow is designed for small residential cleaning businesses that want help organizing incoming questions while keeping their existing inbox.
A customer mentions bedrooms and a date, but leaves out location, home size or condition. The assistant helps you see what still needs asking.
Your approved services, add-ons and exclusions guide a conservative reply. Anything unclear stays with the owner.
Try it alongside the tools you already use. The pilot measures whether it helps your business enough to keep using it.
If your existing software already handles inquiry replies well, this may not add enough value. The pilot tests that fit.
How it works
Start with fictional text in the demo. For a pilot, agree on permitted data and the AI path before using customer messages.
AI extracts service, property and requested-date details. The workbench shows customer wording and flags empty required fields.
A template uses your service rules to ask for missing information. Check the facts, edit the draft and copy it when ready.
The demo sends no messages, creates no bookings and offers no cleaning prices or available slots. Copying a reply does not send it.
A two-week paid pilot
The goal is a useful, repeatable workflow for your business. The pilot checks quality, handling time and the work required to support it.
Scope and pricing agreed before starting.
One business, one approved rule set and one chosen AI path. No automatic renewal. No promise of savings, new customers or profit.
Confirm services, area, add-ons and exclusions. Test the chosen AI path and agree on data handling.
Time your current replies and check extraction quality with representative examples.
You review every reply. Track mistakes, transfer time, waiting, edits and support, with a check-in around Day 7.
Review accuracy, time saved, total cost and realistic monthly value. Continue only by explicit agreement.
What the demo can show
An optional open language model can run on a compatible WebGPU device without an API key. First use downloads a large model; speed and compatibility vary.
Experimental: the local model missed details in testing. It has not passed the proposed pilot accuracy checks.
The workbench prepares a prompt to copy into an AI chat you already use. Paste the returned structured result into the workbench for review.
This shares the inquiry with that provider. Access and usage limits apply. The page cannot verify the origin of an imported result.
Explore the whole review flow immediately using a fictional deep-clean inquiry. The sample is a fixed fixture and does not run AI.
AI extracts facts in the live paths. Required-field checks, evidence grounding and reply wording are application logic.
Model output can be wrong, incomplete or misinterpret a customer's words. Quoted evidence is a review aid, not proof of correctness. A paid pilot starts only after the agreed AI path passes readiness and quality checks.
Before you try it
No. This demo works with pasted inquiries and copied drafts. It has no inbox or scheduling connection. Any future connection needs separate scope and approval.
No. This version asks for details and acknowledges the request. The owner checks scope, service area and availability before providing any quote or booking confirmation.
The workbench is mobile-friendly. The manual AI chat handoff works through copying and pasting. Local browser AI depends on your browser, hardware and memory, so it may be unavailable on your device.
An owner-approved service rule set, a chosen AI path that passes the quality checks, agreed handling of customer data and enough inquiries to measure. Pilot scope, pricing and payment terms are agreed before starting.
See it for yourself
Use the fictional sample first. Then try a fresh fictional inquiry with one of the actual AI paths.