What should your AI receptionist never say? Prices, diagnoses and arrival promises

How this article was researched: review of Rosie AI’s published pricing and after-hours service documentation, re-verified October 16, 2026. No hands-on testing; no calls were made and no vendor staff were contacted.

Verdict: set three boundaries before you switch it on

An AI receptionist can answer calls around the clock, collect customer details and keep service requests moving when your crew is busy. But before you switch one on, we recommend keeping three things out of its answers: firm price quotes, technical diagnoses and promises about when a technician will arrive.

Those are TradeProStack’s editorial recommendations, based on our review of how Rosie AI lets businesses configure answers and train their receptionist. They are not vendor-published restrictions.

For a small HVAC, plumbing or electrical shop, the goal is straightforward: let the AI handle routine calls while keeping decisions that require a technician or owner with a person. Our Rosie AI review covers the full plan comparison; this guide covers what to fence off before activation.

TradeProStack is reader-supported. When you buy through links on our site, we may earn a commission. This doesn’t affect our editorial recommendations.

See Rosie AI’s current plans and pricing (affiliate link), then use the boundary checklist below before you activate it.

Why shouldn’t an AI receptionist quote service prices?

A caller asking “How much will it cost?” is often trying to decide whether to book an appointment. It’s a reasonable question, but the answer can depend on details that aren’t available during the call.

Consider an illustrative plumbing scenario. A customer reports a leaking pipe and asks for a total repair price. The receptionist has a general description of the problem, but nobody has inspected the pipe or confirmed what work is required. If the AI gives a specific price, the customer may reasonably treat that number as a quote, and the technician could then arrive to find the job needs different work.

Our recommendation is to keep unapproved price estimates out of the AI’s answers. The receptionist should collect enough information for the business to prepare an estimate and explain how the customer can get an accurate price. For example:

“I can take the details of the job and pass them to the team. They’ll confirm the pricing once they have enough information about the work.”

If your business publishes a standard service-call fee, you still need to decide whether the AI is permitted to repeat it. The source, wording and conditions of any approved fee should be documented.

Rosie’s Custom FAQ & answers feature lets owners teach the receptionist how to answer common questions on every plan. That provides a place to define the response you want callers to hear.

Rosie’s own after-hours FAQ addresses pricing capability directly, and the answer is tier-dependent: on the Growth plan, Rosie can provide pricing estimates; on the other plans she gives general pricing information based on what the business enters in its FAQs, and can also share service availability and send booking links. Our boundary recommendation doesn’t change with the tier. An estimate is only as good as the information behind it, and most shops are safer having the AI collect the details while the team confirms the price. The ability to configure an FAQ still shouldn’t be treated as a guarantee that every possible pricing question gets the intended response.

Why should technical diagnoses stay with a technician?

A caller may describe symptoms that sound familiar: an air conditioner isn’t cooling, a furnace won’t start or an electrical outlet has stopped working. The problem is that similar symptoms can have different causes, and without a proper assessment the receptionist doesn’t have enough information to say what component failed or what repair is needed.

Here’s an illustrative HVAC scenario. A homeowner calls because their air conditioner is running but the house isn’t cooling, and asks whether the unit needs refrigerant. A receptionist that answers, “Yes, that sounds like a refrigerant problem,” has effectively offered a diagnosis. The homeowner may now expect a particular repair before a technician has inspected the equipment.

Our recommended response is simpler:

“I can record what the system is doing and pass those details to the technician. They’ll need to assess the equipment before confirming what’s wrong.”

This still moves the service request forward. The business gets the symptoms, while the technician keeps responsibility for the diagnosis. If the goal of the call is to get the technician booked rather than to name the fault, our guide to booking the diagnostic instead of giving free advice covers the wording that works.

It’s also worth separating ordinary service questions from safety-related concerns. A customer asking why a furnace won’t start is different from a caller describing a potentially dangerous situation. The receptionist shouldn’t provide improvised safety instructions or attempt to decide whether a dangerous condition is safe.

For genuine emergencies, callers should contact emergency services. An AI receptionist is not an emergency response line.

Rosie states that its system can identify urgent or emergency situations during after-hours calls and automatically notify the business by text or email while reassuring the caller. That’s the vendor’s description, not a capability independently verified by TradeProStack, and it doesn’t establish that Rosie can diagnose equipment problems, assess physical danger or replace emergency services.

Why shouldn’t an AI receptionist promise arrival times?

For a home-service business, the calendar and the dispatch schedule aren’t always the same thing. A customer might be able to book an appointment without receiving a guaranteed technician arrival time, and a request for same-day service can depend on scheduling decisions that haven’t been made when the call comes in.

Consider an illustrative electrical service call. A homeowner asks whether someone can arrive within an hour. The receptionist collects the address and service details, but the owner hasn’t confirmed that a technician is available. If the AI says, “We’ll have someone there in an hour,” the customer has been given a commitment the business hasn’t authorized. A safer response:

“I can take your details and pass along the request for an appointment. The team will confirm when someone is available.”

Rosie’s published booking capabilities differ by plan. On Professional, Rosie can send appointment links by text, which is different from booking directly into a calendar. On Scale and Growth, Rosie can book appointments directly on the business’s calendar. Even with direct calendar booking, we recommend avoiding promises about technician arrival times unless the business has explicitly authorized that wording. An appointment placed on a calendar should be described accurately; what the receptionist shouldn’t do is turn an appointment request into an unsupported promise of immediate service.

This matters most for after-hours calls. A customer may want an urgent response, but the receptionist needs to distinguish between collecting an urgent request and committing the business to a dispatch time. Our after-hours triage guide breaks down how to sort those calls without a 24/7 receptionist.

How do you teach Rosie what it can and can’t say?

Rosie’s training model gives owners two main ways to provide business-specific information, and which one you can use depends on your plan. According to the Rosie pricing documentation (affiliate link), accessed October 16, 2026, Rosie learns from the business’s Google profile and website on all plans, and owners can also configure Custom FAQ & answers on every plan.

That public information deserves a review before activation. If your website contains old pricing, outdated service descriptions or policies you no longer follow, those details may conflict with the answers you intend the AI to give. Rosie says it learns from those sources, so checking their accuracy is a sensible first step.

For explicit boundaries, use the Custom FAQ & answers feature to define approved replies to common customer questions. The following wording is TradeProStack’s suggested configuration language, not Rosie-provided templates or independently tested responses.

The point is to make each boundary specific. “Be careful with prices” leaves too much room for interpretation; “Don’t provide a job price unless the business has explicitly approved the figure and wording” tells the receptionist what the owner wants to prevent. Include approved replacement wording too: when the AI is told to avoid an answer, it should have a clear next step to offer the caller.

Rosie’s published documentation doesn’t establish how reliably these instructions are followed across every possible conversation. Owners should review actual call records after activation and revise their configuration when they find wording that needs attention. If you’re weighing AI answering against a text-back setup instead, our missed-call text-back vs AI receptionist comparison lays out how to tell which one you need.

Do you need Rosie’s Growth plan for custom training?

Not necessarily. It depends on how much information the receptionist needs and how you intend to configure it. Custom FAQ & answers are available on all Rosie plans, which gives businesses on Professional and Scale a way to define responses to common questions about pricing, service requests and scheduling.

Growth adds custom agent training, which Rosie describes as the ability to upload training documents and policies. A small shop might only need FAQ responses explaining that estimates are confirmed by the team; a larger operation with written procedures covering different job categories may find document-based training more suitable, although the vendor documentation alone doesn’t demonstrate how the AI handles every policy exception.

One more Growth-tier distinction matters for this article: Rosie’s FAQ states that pricing estimates are a Growth-plan capability, while lower tiers stick to general pricing information from your FAQs. Our boundary recommendation doesn’t change with the tier, but it’s worth knowing where the vendor draws the line.

Here’s how the published plans compare (USD, re-verified October 16, 2026):

Professional also includes automatic spam detection and a website chat widget. Scale adds the ability to send texts to callers during calls. Growth includes white-glove onboarding and priority support. All three plans include automatic training from the business profile, 24/7 AI answering, English and Spanish support, email and text notifications for every call, and call summaries, transcripts and recordings. Rosie lists a 7-day free trial with all features and says customers can cancel anytime.

On overages: Rosie states there is no per-minute overage charge. You get heads-up emails as you approach your minute limit, and if you exceed it you are moved up to the next plan at that plan’s standard price.

For owners mainly concerned with setting boundaries, the key point is that you don’t need Growth to create custom FAQ answers. Growth becomes relevant if you want the document-upload training capability or the vendor’s pricing-estimate feature.

Compare Rosie’s current plans (affiliate link).

What should happen when the AI doesn’t know the answer?

Every receptionist needs a fallback response, including an AI receptionist. A caller might describe an unusual repair, ask about a policy that hasn’t been documented or request an arrival time the calendar doesn’t establish.

Rosie’s after-hours FAQ describes the default fallback: she lets the customer know she doesn’t have the answer, offers to have someone from the team follow up, and can direct the caller to the business website or a contact form. That’s the vendor’s stated fallback. We’d still recommend defining your preferred wording in the Custom FAQ tools, for example:

“I don’t have a confirmed answer for that. I can take your details and have the team follow up.”

This avoids inventing an answer while giving the customer a way to continue. Owners should review call transcripts to see how real conversations are handled and revise the fallback when needed.

How can the receptionist escalate a call without losing the booking?

Telling every uncertain caller to phone back defeats much of the purpose of having a receptionist. A better approach is to collect the service request while offering a human handoff when the situation calls for one.

Rosie’s Scale and Growth plans provide two relevant transfer features: warm transfers, where Rosie briefs the team before connecting the caller, and live call transfers, where Rosie connects the caller when needed. Growth also includes waterfall transfers, which try multiple phone numbers until someone answers. These give owners options for calls that need human involvement, although the published evidence doesn’t establish every configuration detail.

For an illustrative plumbing call, a customer might report a leak and ask whether a technician can come immediately. The receptionist could use wording such as:

“I can take the address and details of the leak. I’ll also try to connect you with someone who can confirm the service options.”

If the transfer succeeds, the team handles the scheduling question. If nobody is available, the business needs an approved message-taking and follow-up process. Rosie’s message-taking scenarios are another configuration surface: Professional includes 2 scenarios, Scale includes 5 and Growth includes unlimited. These are structured intake questions the business defines, such as service address and problem description. That is an illustrative use of the feature, not a claim that Rosie includes those specific questions by default.

The objective is to collect information the crew can use without letting the receptionist make decisions beyond its authority.

What should happen with urgent after-hours calls?

Urgent service calls need particular care because a caller may interpret reassurance as a commitment to send help. On its after-hours answering service page, Rosie states that it is programmed to recognize urgent or emergency situations during after-hours calls, and that when it identifies such a situation it automatically sends a text or email notification to the business and reassures the caller. That is vendor-published behaviour, described October 16, 2026; TradeProStack hasn’t independently verified its performance, and the documentation doesn’t establish the full detection logic.

For a trade business, we recommend keeping two separate instructions. First, the receptionist should avoid diagnosing the equipment or offering a safety determination. Second, it should avoid promising that a technician is already on the way unless the business has confirmed the dispatch. A suitable general response for an urgent service request:

“I’ll record the details and flag the request for the team. Someone will need to confirm the available service options.”

For a genuine emergency, the instruction must be clearer:

“If this is an emergency involving immediate danger, please contact emergency services.”

An AI receptionist is not an emergency response line, and its notification or transfer features shouldn’t be represented as a substitute for emergency assistance. Owners considering Rosie for urgent calls can review its after-hours answering service information (affiliate link) and decide what additional escalation arrangements their business needs. Rosie also confirms that the service can be enabled for after-hours and overflow calls only, via conditional call forwarding.

How do you check whether the receptionist is staying within its boundaries?

Rosie includes call summaries, transcripts and recordings on every plan, plus email and text notifications for every call. Those records give owners material to review after the receptionist starts handling customer conversations.

We recommend looking for specific problems rather than relying only on whether the call appeared to end successfully. Did the receptionist give an unapproved price? Did it suggest that a particular repair was required? Did it promise a technician arrival time the business hadn’t confirmed? The same review should check whether the receptionist collected enough information for the next step: an AI that avoids giving a quote but fails to record the service request still creates follow-up work.

For a small shop, a practical process is to check transcripts against the approved FAQ wording and revise the instructions when needed, and to make sure the published business information Rosie learns from stays current. If you want a concrete sense of what AI-vs-human answering sounds like in practice, our AI receptionist test-call breakdown walks through five test calls and what they reveal. These are recommended operating practices, not evidence of a particular error rate or a guarantee about Rosie’s performance.

What is still not published about Rosie’s pricing and behaviour?

Most of the important facts are on the record. A few things remain genuinely unknown in the documentation we reviewed:

• Emergency detection logic: Rosie describes urgent-call identification and owner notification, but the documentation doesn’t establish the full detection mechanism or independently verify performance.
• Canadian pricing: all prices in this article are USD; the reviewed pages didn’t establish CAD pricing or Canada-specific coverage.
• Instruction reliability: configurable FAQ answers don’t establish how the AI responds to every possible phrasing of a question, which is why transcript review matters.

For any of these, ask Rosie for the current documentation before making a purchase decision.

What’s the safest setup for a small trade business?

For a small HVAC, plumbing or electrical shop preparing to activate an AI receptionist, we’d start with a short set of approved answers. Keep job pricing with the people authorized to provide estimates, leave technical diagnoses with technicians, and make sure arrival times are confirmed through the business’s scheduling process.

Use Rosie’s Custom FAQ & answers feature to document those boundaries. If the business has more detailed written policies, Growth offers the additional ability to upload training documents. Scale is worth considering when direct calendar booking and live call transfers matter. Professional is the lower-cost entry point for AI answering, message taking and appointment links by text.

The key limitation is that configurable answers don’t establish how the AI will respond to every question. Owners still need to review transcripts and maintain accurate business information.

Based on Rosie’s vendor documentation reviewed October 16, 2026, its training and call-handling features provide practical ways to set up these boundaries. The exact performance of those instructions remains something each business needs to assess from its own call records. For the broader missed-call picture, our missed-call handling guide covers the four setups that actually book work.

Review Rosie AI’s current pricing and features (affiliate link).

Methodology: TradeProStack reviewed Rosie AI’s published pricing and after-hours service documentation, accessed October 16, 2026. This article provides editorial setup guidance based on those materials. We did not conduct hands-on testing or independently verify the vendor’s call-handling claims.

Source documentation: Rosie AI pricing and Rosie AI after-hours answering service. All prices are USD.

Related: Who answers when you’re on a job? Missed-call options for small trade crews