An AI phone agent for the calls your cleaning company cannot answer mid-job
A cleaning company's phone rings hardest when nobody can pick it up. The crew is mid-house with gloves on; the owner-operator is scrubbing a bathroom or driving between jobs; the janitorial side works nights while clients call during the day. The caller on the other end is often a brand-new lead who wants service this week — and a lead who reaches voicemail can simply dial the next company on the list. For a business built on recurring visits, every one of those calls is not one job; it is potentially years of weekly or biweekly revenue deciding whether to leave a message.
ThunderPhone gives a residential or commercial cleaning company an AI phone agent that works in both directions: inbound coverage for the calls you receive, and outbound campaigns for the calls your team never has time to place — visit confirmations, cancellation backfill, win-back outreach to lapsed clients. It follows your instructions and searches your approved knowledge base without inventing company policy.
The agent does not quote jobs on its own. It never invents a price, promises a crew you have not committed, or waives a fee. You define what it may say, what it must collect, and when it must escalate — pricing language, scheduling rules, and complaint handling come from the company, and the agent executes them exactly.
What a cleaning-company phone agent can handle
New-lead intake for residential cleaning
A first call follows a script every owner knows by heart: name and contact details, address or neighborhood, home size, bedrooms and bathrooms, what kind of service — standard maintenance, deep clean, or move-in/move-out — how often, pets, and any product sensitivities. The agent can walk through that intake conversationally, answer approved questions about what each service level includes, and share only the ballpark language you have written — many companies quote a range on the phone and confirm the final price after seeing the home, and the agent can say exactly that. Then it books the next step your workflow uses: an in-home estimate, a firm appointment where your pricing rules allow one, or a callback from you. The structured lead record lands in your system rather than on a voicemail; anything outside the script gets a transfer or a callback instead of a guess.
Commercial quote requests and walkthrough scheduling
Commercial and janitorial leads ask different questions, and the agent can run that script too: building type and who uses it, approximate square footage, service frequency, restroom count, floor types, and any nonstandard areas the caller mentions. Because a serious commercial bid follows an on-site walkthrough, the agent's job is not to price the contract — it is to capture a complete, qualified inquiry and book the walkthrough into your calendar on the same call. A facilities manager who calls three vendors and gets one immediate, competent intake conversation remembers which company that was.
Booking and rescheduling recurring visits
Recurring clients call to move things constantly — a houseguest this Thursday, a holiday next month, an office closure. The agent can check availability and reschedule through a connected calendar or field-service API — via HTTP tools, MCP servers, or webhooks — apply the notice rules you configure, and state your cancellation policy as written rather than negotiating it. When a change frees a slot, that opening is captured as structured data your team (or an outbound campaign) can fill instead of eating the gap.
Day-before confirmations, with access details
For a cleaning company, a confirmation call is not just about attendance — it is about getting in the door. An outbound campaign can call tomorrow's schedule, confirm each visit, and run through the access checklist you define: will someone be home, is there a lockbox or garage code arrangement on file, are the pets secured, where should the crew park. A cancellation captured tonight is drive time saved tomorrow; an access answer captured tonight is a lockout that never happens. When a call reaches voicemail, voicemail detection can leave your configured reminder or hang up and let the campaign retry later, on the retry policy you set. Each outcome — confirmed, rescheduled, unreachable — is delivered to your systems by webhook before the crew is dispatched.
Cancellation backfill and the waitlist
When a client cancels, the slot is perishable. An outbound campaign can call your waitlist or your flexible recurring clients, offer the opening, and book the first taker — the kind of fast, repetitive calling a human office never gets to before the slot expires. The same mechanism works in reverse for clients who asked to be called when an earlier day opens up.
Complaint and quality-callback intake
The calls that decide whether an account renews are the unhappy ones: a missed area, an incomplete task list, a crew that arrived late. The agent can take that call with the composure the moment needs — acknowledge, apologize in the terms you approve, and collect a structured record: which visit, which rooms or tasks, what the client wants done. Then it follows your severity map: file the complaint by webhook for your stated resolution process, offer the re-clean policy you have written, or transfer a commercial account contact straight to the owner. What it never does is argue, promise compensation you have not authorized, or let the complaint die in a voicemail box.
Win-back and reactivation campaigns
Every established company has a list of lapsed clients — the biweekly customer who paused a year ago and never came back. An outbound campaign can work that list with a script your team approves: a polite check-in, an offer you have authorized, and a booking on the same call for anyone ready to restart. Clients who ask to be left alone are logged as exactly that, and the campaign respects it. Reactivation is multi-touch, low-hit-rate calling — precisely the work a phone agent does not get tired of and a two-person office never gets to.
After-hours and urgent-request capture
Cleaning-company hours and caller hours rarely match: residential leads call evenings and weekends, janitorial clients call during the business day while your crews sleep, and the occasional urgent request — a flooded break room, a same-day turnover — arrives at 10 p.m. After hours, the agent books and reschedules like it does at noon, and for urgent requests it follows the path you configure: collect the details and dispatch a message to the on-call person by webhook, or warm-transfer callers who meet your urgency criteria. It never promises a crew or a response time on its own — availability commitments are yours to make.
Multilingual client communication
ThunderPhone supports 47 voice languages. Set a primary language and additional languages for your client base, and the agent can switch mid-call when a caller changes to one of the selected languages — no separate line, no callback protocol, no waiting for the one bilingual team member to be off a jobsite. The same intake, scheduling, and escalation rules apply in every configured language, and every call produces a transcript the team can review.
Overflow, and calls that hit someone else's phone tree
Many companies do not need the agent to answer every call — they need it to answer the calls the office cannot reach: the morning rush, the hours when the whole team is on site, vacations. As overflow coverage it picks up where the humans leave off and takes structured messages for the rest. On the outbound side, when a call reaches another business's automation — a property manager's front desk, a building's switchboard — the agent can wait on hold at a flat 2¢-per-minute rate and navigate keypad menus before handing off to your staff.
Sample cleaning-company call flow
Agent: Thank you for calling Brightside Cleaning. Are you calling about new service, an existing appointment, or something else?
Caller: New service. How much do you charge for a three-bedroom house?
Agent: Happy to help with that. Pricing depends on the home and the type of cleaning, so the office confirms the exact price after a quick estimate — but I can take the details now and get you scheduled. Is this a regular cleaning, a deep clean, or a move-in or move-out?
Caller: Probably a deep clean first, then every two weeks after that.
Agent: That is our most common way to start. How many bathrooms, and do you have any pets?
Caller: Two bathrooms, one dog.
Agent: Noted. I have estimate visits available Thursday at 4:00 p.m. or Saturday at 10:00 a.m. — which works better?
Caller: Saturday.
Agent: You are set for Saturday at 10:00 a.m. May I have your name, address, and the best number for the crew lead to reach you?
If the caller had reported a problem with a recent visit, the agent would instead run the company's complaint script — which visit, which areas, what resolution the client wants — and file it for the stated follow-up process or transfer to the owner. Every branch, including the pricing language and the estimate-first rule, comes from the company's script.
Features that matter in a cleaning business
GDPR and HIPAA compliance posture
ThunderPhone's public compliance posture is GDPR and HIPAA compliant. Your company still controls the workflow: limit the agent to necessary information — client addresses, access notes, and alarm details deserve deliberate handling — attach only approved documents, choose recording settings deliberately, and define where call data may be sent.
Scheduling and field-service connections
Agents can use connected apps, HTTP APIs, or MCP tools during a call. That makes it possible to check crew availability, book an estimate or a recurring visit, log a cancellation, or file a complaint task when the receiving system exposes the required action. Webhook deliveries are HMAC-signed and retried automatically, so a briefly unavailable endpoint does not lose a booking.
Knowledge grounded in your documents
Attach your service descriptions, what-is-included checklists, pricing-range language, cancellation and re-clean policies, and service-area details — as text or PDF/DOCX uploads (up to 5 MB per text file and 50 MB per PDF or DOCX), or imported directly from a page on your website. The agent can search those documents mid-call instead of relying on a generic answer, and each agent uses only the documents you scope to it.
Your numbers, your telephony
Bring your company's existing numbers through a supported VoIP connection or SIP configuration; verified numbers handle inbound and outbound calling. A web voice widget can put the same agent on your website, so a visitor comparing companies can ask about service and book an estimate by voice without dialing.
Human handoff
Cold transfers are available for immediate routing. Warm transfers are available on eligible production VoIP or SIP numbers that can place an outbound leg: the caller waits while the agent privately briefs the recipient, then connects the call only if the recipient accepts — the right shape for putting an unhappy commercial account directly through to the owner.
Testing before clients hear it
Use a browser mic test for quick iteration, then run AI-caller simulations for the scenarios that worry you: a price-shopper who pushes for a firm number, an angry missed-areas complaint, a lapsed client on a win-back call, an after-hours urgent request. Simulations create transcripts and grading so you can tighten the prompt before deployment, and graded scenarios can be kept as a regression suite so a prompt change never quietly breaks the no-invented-prices rule. When two scripts seem equally good, an A/B experiment on live traffic settles it. Production calls produce transcripts, recordings where enabled, and grading you can review.
Guardrails the company controls
- No invented prices, ever. The agent shares only the pricing language you approve and books the estimate or walkthrough your workflow requires. It never produces a number your documents do not contain.
- No unauthorized promises. Crew availability, response times, discounts, refunds, and fee waivers are commitments only your team makes; the agent collects and routes instead.
- Approved answers only. Policy answers come from the prompt and scoped knowledge base; anything unanswered routes to a person instead of a guess.
- Escalation is configured, not improvised. Complaint severity paths, after-hours routing, and transfer targets are set by the company and exercised in testing before launch.
What would this cost each month?
Suppose a company handles 800 AI-assisted calls per month, averaging 3 minutes each:
800 calls × 3 minutes = 2,400 minutes per month
- Spark:
2,400 × $0.02 = $48/month - Bolt:
2,400 × $0.05 = $120/month - Storm:
2,400 × $0.09 = $216/month
That is base engine usage, not a quote. Premium voice or language paths can add up to 3¢ per minute, and phone-number or carrier charges depend on the telephony setup. Hold time bills a flat 2¢ per minute; voicemail and phone-menu minutes bill at the engine rate. Optional add-ons such as verbal acknowledgements, call supervision, and long prompts carry their own published surcharges. The agent builder shows the current all-in rate before calls are placed. There is no subscription — usage is pay as you go.
Frequently asked questions
Can a ThunderPhone agent book cleaning appointments?
Yes, when it is connected to a calendar or scheduling API that permits the action. The agent can distinguish job types — estimate visit, one-time deep clean, recurring service, commercial walkthrough — gather the required fields, check available times, book or reschedule, and confirm the result to the caller. The connection can be an HTTP API, an MCP server, or a webhook into your own systems.
Can it give callers a price?
Only the pricing language you approve. Many companies quote a range on the phone and confirm the final price after an estimate visit or walkthrough; the agent can state your ranges, explain what affects the price, and book the estimate — but it never invents a number or commits to one your rules reserve for a human.
Can it handle both my residential and commercial lines?
Yes. Residential intake, commercial quote qualification, and existing-client service can run as different branches of one agent's script or as separate agents on separate numbers, each with its own prompt, knowledge documents, and escalation rules.
What happens when an unhappy client calls?
The agent runs your complaint script: acknowledge in the terms you approve, collect a structured record of the visit and the issue, and follow your severity map — file it by webhook for your resolution process, offer your written re-clean policy, or transfer to a person. It does not argue with the caller or authorize compensation on its own.
Can it make outbound calls to confirm visits or win back old clients?
Yes. Outbound campaigns can confirm tomorrow's schedule and access details, backfill cancelled slots from a waitlist, and work a lapsed-client list with your approved script. Voicemail detection can leave a configured message or hang up, and the campaign retries unreachable clients on the schedule you set. Every call produces a transcript and an outcome your systems can consume by webhook.
Can one agent speak both English and Spanish?
Yes. Configure one as the primary language and the other as an additional language. The agent can switch mid-call when the caller changes language, provided the selected voice supports both. The same applies across ThunderPhone's 47 supported languages.
How do we make sure it sounds right before clients hear it?
Test in the browser first, then run AI-caller simulations of the scenarios that worry you — a caller demanding a firm price, a complaint about a missed room, a late-night urgent request. Simulated calls are graded against your criteria and can be re-run as a regression suite whenever the prompt or knowledge base changes.