An AI voice agent for solar companies: qualify leads before the site survey

An AI voice agent for solar companies earns its place in one spot: the gap between a lead form and a survey on the calendar. It calls the homeowner promptly, asks the few questions that decide whether a visit is worth it, and books the survey if the answers fit. Installers lose most money on wasted trips and slow follow-up, and both are problems a calling agent can reduce without pretending to be a salesperson.

Why solar leads are a good fit for AI calling

Solar leads have three features that make them workable:

1. They are usually inbound or semi-warm. The person filled in a form or responded to an ad.
2. The qualifying criteria are clear. Ownership, roof and energy use matter far more than charm.
3. The conversion event is concrete: a site survey or design call.

They also have a problem. A large share of leads are not viable, because the person rents, lives in a flat, has a heavily shaded roof or was only curious about prices. A human rep calling every lead spends most of the day hearing "not really".

Qualification that respects the caller

The agent should ask the minimum needed to decide whether to book a survey. A good set:

Question	What it tells you
Do you own the property?	Rules out renters quickly
Roughly what is your monthly electricity bill?	Rough size of the opportunity
What type of roof do you have, and does it face a sunny direction?	Early fit signal; callers often know roughly
Is there heavy shading from trees or buildings?	Another early fit signal
Is anyone else involved in the decision?	Tells you who must attend the survey
When would you like it installed?	Urgency

Do not interrogate. Two or three honest "I am not sure" answers are fine, because the survey exists to find out. The agent should explain why it is asking, for example: "so the surveyor knows what to look for".

For a deeper approach to question design and scoring, see AI lead qualification calls.

Handling the questions homeowners actually ask

Solar buyers ask a predictable set of things, and many of them can be answered from a well-written knowledge base:

• How long does installation take?
• What happens on cloudy days?
• Do you offer financing, and how do payments work?
• What warranties come with the panels and inverter?
• What does a survey involve, and is it free?
• How does a battery fit in?

Put your approved answers in the knowledge base. Be careful with claims about savings, payback periods and incentives, which change by region, tariff and time. If you include them, phrase them as general and tell the agent not to promise specific savings. A prompt line like "never state a specific saving or payback figure; offer a survey so a consultant can give an estimate" prevents trouble. Our guide to handling objections with an AI voice agent shows how to build an objection library for this.

Booking the site survey

When the lead qualifies, the agent offers survey slots from your Google Calendar. A few details improve the experience:

• Confirm the full address and postcode or ZIP, and repeat it back.
• Ask whether all decision makers can be present.
• Capture access notes such as pets, gated entry or parking.
• Send a confirmation, and consider a reminder the day before.

If your surveyors cover different territories, use separate calendars and tell the agent which area each covers. Details on the mechanics are in AI appointment setting.

Setting up an outbound campaign

In Purple Voice you create an agent, write the prompt and knowledge base, then create a campaign with your contacts. Set calling hours that match local expectations, since evenings often work well for homeowners but you must stay within legal windows. Use retries, up to five, with retry hours that vary across the day. Not-reached analytics on the campaign show you how many leads never picked up, which tells you whether to change times or channel.

Test before you launch. Use the test-call feature to ring your own phone, then play a skeptical homeowner, a hurried one and someone who asks about financing. The article on testing your AI voice agent has a checklist.

Calling lead lists: be careful

Solar has a reputation for aggressive telemarketing, and regulators notice. If your leads come from a third party or a "compare quotes" site, check what the person actually agreed to. A lead who consented to be contacted by one named company is different from one who ticked a broad box. Rules differ by country and state and change over time, so check current requirements. This is general information, not legal advice.

Good habits:

• Say who you are and that the caller is an AI assistant.
• Respect do-not-call requests instantly.
• Call within permitted hours.
• Keep a record of consent for each contact.

More on this in do-not-call lists and dialer hygiene.

What to do with the output

Every call produces a transcript and summary. Review the first fifty. You will find that homeowners say things you did not anticipate: a landlord owns the roof, a retired customer wants to talk about heat pumps, someone already signed with a competitor. Add each pattern to the prompt or knowledge base. The summary tells your sales team what to say when they follow up, so the lead does not feel they are starting again.

Hot calls can transfer to a person. A homeowner who says "I want to go ahead this week" is worth a live handoff.

Limits to keep in mind

• The agent cannot judge a roof from a description. Treat its notes as signals, not facts.
• It cannot give a legally binding price.
• Weather, incentives and tariffs change quickly. Review your knowledge base regularly.
• Elderly or vulnerable homeowners deserve extra care; give the agent an easy path to a human.

Measuring whether it pays

Compare two numbers over a month: surveys booked per hundred leads, and surveys that turn into signed proposals. If bookings rise but proposals do not, the agent may be booking people who do not fit, so tighten the qualifying questions. If bookings stay flat, check how many leads were never reached, using the not-reached analytics, and adjust calling hours or retry times. Also track surveyor time wasted on non-viable visits. For many installers that single figure, in hours and fuel, is the clearest sign the filtering is working. Use your own data, and change one thing at a time so you know what caused the difference.

Try it

Run a small pilot with the last month of leads that went cold. The Purple Voice Free plan includes one agent, one number, 100 contacts per campaign and 100 minutes a month with no card. Compare surveys booked against your usual rep outcomes before scaling. Start at voice.purplerocket.co, or see pricing for the plans above Free.