One shared workbook, one identity gate, one place cases are opened. This pack answers "where is my order", "can I return this" and "something arrived broken" in your website chat - and it proves who it is talking to before it says a word about an order. Identity is checked against rules you write in a sheet, not a model's judgement, and what reaches the agent is already masked: a partial email, the last digits of a phone number, and only the order fields you listed as safe to share. A failed check reveals nothing at all, not even whether that email has an account. Anything needing a person becomes a case - opened once, deduplicated by customer, order and problem type, so a retried click or an impatient customer produces one case and not three.
Answers your phone and manages appointments by voice - booking, moving, cancelling and confirming - against your real Google Calendar, 24 hours a day. It opens with the disclosure you write, reads every name, date and time back before acting on it, and checks live availability before agreeing to anything. It cannot invent a free time: availability and any alternatives are worked out in code from your diary and your working hours, and the agent may only repeat what that returns. Anything it cannot finish on the call becomes a callback request with the real reason attached.
Answers your customers in your website chat using only the answers you have written down — and says so plainly when something is not written down, instead of inventing it. It walks people through your own troubleshooting steps, stops the moment your own rules say a problem needs a person, and hands your team a ticket carrying everything the customer already tried.
Guides a prospect on your website from a first question to a warm handover: it runs your own discovery questions, recommends only from the products you have written down, and explains why that one fits what the customer told it. It never invents a price and it never offers a discount - any pricing ask goes to a named person instead. Every lead is written to your sheet with the reason for the recommendation, and your sales team is emailed the full brief.
Handles standard return requests in your website chat, using your own written rules rather than a model's judgement. It checks the customer against the order, applies the right policy for that product category and sales channel, and tells them plainly what they are entitled to — then records the return and emails your team. Refunds above a limit you set can never be promised by the agent; they go to a person.
Answers "where is my order?" in your website chat, at any hour. It checks the customer against your own order records before it says anything at all, describes every parcel of a split delivery separately with its own tracking link, and turns a late, lost, damaged or undeliverable parcel into a case your team is emailed about — instead of a promise nobody can keep.
Lets customers book, move and cancel their own appointments in the chat on your website. It reads your real Google Calendar, offers only times that are genuinely free, checks the slot again at the moment of booking so two people cannot take the same one, and puts the appointment straight into your calendar with the customer's details on it. It never invents a time and never asks for payment details.
Answers the people who arrive in your website chat. It replies only from answers you have written down, says plainly when something is not written down rather than guessing, and turns anything it cannot answer — and anyone who wants a person — into a recorded contact plus an email to the right team, carrying the whole conversation.
Answers the customers who write in after your office closes. Routine questions get a reply drawn only from information you approved, anything it cannot answer honestly becomes a next-day callback with a ticket, and genuine emergencies page whoever is on your on-call rota — chasing them until somebody acknowledges, then escalating to a manager.
Every missed call is a customer about to phone a competitor. This agent texts them back within minutes, finds out what they needed, and does not let the request drop — a callback stays open until a human confirms they made it. It checks the opt-out list, the line type and your quiet hours before it sends anything.
Contacts the customer using the phone number provided at launch to confirm their upcoming service appointment. Checks access details and handles confirmation, rescheduling, or cancellation requests, then sends a summary notification of the call to your sales team — reducing no-shows and keeping your schedule accurate without manual follow-up calls.
Qualifies inbound service enquiries by collecting the customer's requirements, job details, location, preferred timing, contact information, and buying intent — asking one question at a time and skipping anything they've already provided. Once the conversation is complete, the qualified lead is logged and sent to your team for follow-up.
Cuts no-shows by reminding every customer about their upcoming appointment by email, collecting a confirm / rebook / cancel answer, updating your Google Calendar, and alerting your team the moment a slot frees up or a reply needs a human.
Removes the back-and-forth of interview scheduling. When a recruiter advances a candidate, the agent checks every interviewer's real free/busy — never the contents of their calendars — and emails the candidate a numbered list of open times. The moment they pick one it re-checks availability, books the Google Calendar event with every participant invited (and a Meet link for video), and sends the preparation notes you wrote for that job. Anything it cannot do safely — a calendar it cannot see, a slot taken while the candidate deliberated, a candidate who needs different times — goes to your recruiter with the full story instead of being guessed.
AI recruiting assistant is the front door for candidates. Answers questions about your open roles from information you provide, takes applications by email with consent recorded, tells people where their application stands, and hands anything sensitive to a human. It records each application so the screening and qualification agents can pick it up — it does not assess anyone itself.
Runs a structured competency assessment on candidates who have already passed screening. Asks about each competency you define, scores the answers against your own guide, and attaches the candidate's own words as evidence for every score. The recommendation from automated candidate pre-qualification systems goes to a recruiter to accept or override — the agent never makes the final call.
Screens applicants against the requirements you set, in your own words, and emails each one a short set of job-related questions. Answers are scored against rules you keep in a spreadsheet (not invented by the agent) and every pass, fail or borderline case is written back with the reason. Anything AI candidate screening agent is not sure about goes to a recruiter instead of being decided.
Qualifies general cleaning enquiries by collecting service requirements, property information, location, preferred timing, contact details, and buying intent — asking one question at a time and skipping anything the customer already provided. Once the conversation is complete, the qualified lead is logged and sent to your team for follow-up.
Receives a cleaning job from Google Calendar and checks cleaner availability, skills, service area, and existing assignments in Google Sheets. Selects a suitable cleaner, sends an Email notification, records the assignment, and alerts operations if nobody is available — automating job dispatch without manual scheduling.
Starts from a booked Google Calendar event and automatically sends the customer a confirmation follow-up. Notifies your sales team about the booking and follow-up status, so every cleaning appointment on your calendar gets confirmed without manual outreach.
