AI Agent for Tradespeople: Answer Leads While Working

Tuesday, 11:20. A plumber from Rishon LeZion is under a kitchen sink in Holon, wrench in hand. The phone rings in his pocket and he cannot answer. Twenty minutes later: two missed calls and a WhatsApp message, "hi, I have a leak in the kitchen, do you work in this area?". He replies at six in the evening, and she has already booked someone else, because a person with a leak calls three tradespeople and takes the first who answers. The advertising worked; what broke is operational. For a tradesperson, the person selling and the person on the job are the same person.

Why a good tradesperson loses more leads

The paradox: the fuller your calendar, the less available you are for new work, so the tradesperson most in demand misses the most. A business with an office has someone at a screen; here the sales department is a phone in a pocket, with dirty hands, up a ladder. Three costs follow:

• A missed call does not call back by itself. An urgent fault does not wait; the customer moves to the next Google result or the next name in the neighbourhood group. On the link between response time and closing odds, see the five-minute rule.
• The big job hides among the small ones. A 15,000 ILS bathroom renovation sits in the same unread list as "how much to unblock a drain", and without triage the order of handling is set by whoever wrote last.
• There is no list. WhatsApp is not a lead management system. A quote sent on Tuesday with no answer disappears into the scroll.

The four questions that decide whether an inquiry is worth the drive

An AI agent for tradespeople does not need to hold a conversation. It needs four facts, pulled out in three or four messages rather than a form:

1. What the job is. A blockage, a leak, an air-conditioner install, a full renovation. The type sets the visit length and the price range.
2. Where. City, neighbourhood, exact address. Your service area is a rule set in advance: if a small call-out 40 minutes away does not pay, the agent says so politely instead of you discovering it in the van.
3. Urgency. Now, this week, or planning for next month: three completely different tracks.
4. Scope and constraints. Floor and lift, access and parking, tenant or owner, one photo if possible. These decide which tools to load.

The fifth item is not a question but a capability: answering immediately, at 22:40 too and while you are inside a wall. The agent gives no binding quote for work it has not seen; it quotes ranges from your price list plus the call-out fee and says the firm number follows an inspection. So the customer gets an answer without you committing to anything.

Three ways to answer while you are on a job

Return calls in the evening. Costs nothing, but most of these inquiries are urgent, and by eight some are already closed with a competitor.

A human answering service. Someone takes details politely, but does not know your price list, whether you work in Modiin, or your calendar. You get a name and a number and still have to call.

An AI agent on WhatsApp and the website. Answers in seconds, knows the price list and service area from a live knowledge base, qualifies, books the visit and opens a card. Anything needing professional judgement is handed to you with the conversation history. Deliberately hybrid: the agent takes the volume, you take the trade.

Voice notes, photos and addresses: what actually happens

Customers of tradespeople do not type paragraphs. They send a 40-second recording: "there is a leak under the sink, I put a bucket there, can you come today?". Our agents understand and process voice messages sent on WhatsApp or Messenger and carry the flow on from there, so nobody is forced into a form.

The same in reverse: the agent sends and receives images, files and documents inside the conversation. It asks for a photo of the leak or the electrical panel, and that photo stays on the conversation card, so you know what is waiting and which tools to load. Some cases close with no drive at all. Language matters in Israel too: many customers write in Russian or English, and the agent replies in all three on the same channel. The scheduling, confirmation and reminder layer is covered by the AI office manager.

Booking the visit, reminders and chasing quotes

Calendar booking is where a smart agent separates from a bot that only collects details, and where a mistake costs money. The calendar runs on rules you set: slot length per job type, travel time between calls, areas by day and a daily cap on call-outs. So two call-outs never land in the same hour in Herzliya and Ashdod.

The customer gets a confirmation email with the time and what to prepare (clear the cupboard under the sink, leave a parking space) and a reminder the day before. That reminder is worth more here than anywhere else, because "nobody was home" is a whole drive wiped off the day; see the guide to cutting no-shows.

Then the part that adds revenue: quotes that never closed. Three days on, the agent asks whether the quote is still relevant; a renovation postponed until after the holidays returns to the list on the agreed date. Every plan includes unlimited follow-ups and workflows with no trigger meters, so a quote-chasing sequence adds nothing to the bill.

How to work out whether it pays for itself

Two numbers you already know: average revenue per job, and how many inquiries a week you never return. Say an average call-out brings in 600 ILS and two a week fall between the cracks: roughly 5,000 ILS a month driving off to a competitor, and even if only a third would have closed, that is about 1,700 ILS, more than the Pro plan. For a renovation contractor or an air-conditioning installer, where one job is a few thousand shekels, one extra job a month is enough. These are example numbers, not market data: put your own in.

The other side is a saving nobody talks about: cancelled drives. An out-of-area inquiry filtered before it became a trip, a visit that did not evaporate because a reminder went out, a fault solved from a photo. That qualification, scheduling and follow-up logic is what the sales agent covers.

What to prepare before setup

Setup is done by our team, but the material comes from you, and this stage decides how good the agent is:

• A price list or price ranges by job type, plus the call-out or diagnostic fee.
• Your service area and what falls outside it, including a travel surcharge if you have one.
• Working hours and emergency availability, including holiday eves and Saturdays.
• The calendar: slot lengths, travel times and a daily cap on call-outs.
• 10-15 recurring questions with your answers, in your words.
• The handoff rule: what is urgent, what is complicated, who is an existing customer.

The setup fee is one-time from 490 ILS depending on scope, and setup takes up to 14 business days from the moment we receive all your material; the subscription starts counting only when the agent goes live. The agent does not replace you at the trade: it takes the layer between the inquiry and the visit, the one currently falling between jobs.

What it costs

Solo costs 490 ILS per month: one agent on one channel, 2,000 AI messages (about 300 conversations), a knowledge base of up to ten documents, up to 20 media files and WhaleBiz CRM. Pro costs 990 ILS per month: up to three channels, 5,000 AI messages (about 750 conversations), an unlimited extended knowledge base, up to 100 media files and in-chat payments. Custom is built to need, for example a crew of technicians with an existing system to integrate.

A one-person business whose inquiries all arrive on WhatsApp starts on Solo; anyone also receiving them from a website, Facebook or Instagram needs Pro. On top of the plan there is a one-time setup fee from 490 ILS, and prices exclude VAT. Worth knowing in advance: Meta charges separately only for outbound business-initiated messages, not for ones the customer initiated, so answering incoming inquiries is the cheap part. Full detail is on the pricing page.

Frequently asked questions

Will the agent quote a final price before I have seen the job?

No, unless you decide otherwise. In most setups the agent quotes ranges from your price list plus the call-out fee, and explains that a binding quote follows an inspection or photos. Fixed-price work, such as a diagnostic visit, can be configured as an exact price. It is a rule written during setup, not the agent guessing.

My customers send voice notes instead of typing. Does that work?

Yes. The agents understand and process voice messages customers send on WhatsApp or Messenger, and carry the flow on from there: one clarifying question, a booked visit, a card opened. The agent also sends and receives images and documents, so it can ask for a photo of the leak or the electrical panel before anyone drives out.

How do I stop two call-outs landing in the same hour?

The calendar runs on rules you set during setup: slot length per job type, travel time between calls, areas by day of the week and a daily cap on call-outs. The agent offers only what is free inside those rules.

What does this cost for a one-person business?

Solo costs 490 ILS per month: one agent on one channel, 2,000 AI messages (about 300 conversations) and WhaleBiz CRM. If inquiries also arrive from your website and social pages, Pro at 990 ILS adds up to three channels and 5,000 AI messages. There is also a one-time setup fee from 490 ILS, prices exclude VAT, and the subscription starts counting only when the agent goes live.

What happens when a customer insists on talking to me rather than an AI agent?

It is handed to you with the full history: what the customer described, which address was given and what the agent already answered. The model is deliberately hybrid: the agent takes the volume, the qualification and the scheduling, and you step into cases that need professional judgement.