Let your AI agent qualify new enquiries before you call them back
Most small businesses lose more time on the first phone call than on the job itself. You ring back a "hi, how much?" message, find out the customer lives forty miles away, wants it done tomorrow and has a budget that would not cover the materials, and twenty minutes have gone. An AI agent can do that first conversation in the chat instead. It asks the three or four questions you would ask, in your tone, tags the enquiry and gives you a short list of people worth calling, with everything you need on one line. This guide shows how to choose the questions, set the agent up step by step, what a good call-back list looks like and where a person should take over.
What qualifying means for a small business
Qualifying sounds like a sales-team word, but every owner does it. When a message comes in, you want to know whether you can do the job, whether you can do it when the customer needs it, and whether the size of the job is roughly right for you. Everything else can wait for the call.
The agent's job is to find out those few things politely and quickly, while the customer is still in the chat. It is not there to turn people away. A customer who is not a fit today should still leave the chat feeling helped, with a pointer to the right kind of business if you have one.
Done well, this changes the shape of your day. Instead of ringing ten people to find the four who are ready, you ring the four, and you already know what they need when they answer.
Choose your questions before you write anything
Write down the questions you always end up asking on the first call. Most businesses need only three or four. Here is what that looks like in a few different trades. Names and details are examples.
| Business | Must know before you call | Nice to know |
|---|---|---|
| Plumber or heating engineer | Postcode, what the problem is, whether water is leaking now | Make of boiler, a photo |
| Wedding photographer | Date, venue or area, hours of coverage | Rough budget, how they found you |
| Cleaning company | Area, size of the home or office, one-off or regular | Pets, parking, preferred day |
| Consultant or agency on LinkedIn | Company, what they want help with, when they want to start | Budget range, who else is involved |
| Kitchen fitter | Area, whether the room is measured, when they want work to start | Style, appliances included or not |
| Tutor | Subject, level or year group, online or in person | Exam date, days that suit |
Keep the "must know" column short. Every extra question is another chance for the customer to stop replying. Anything in the "nice to know" column can be asked on the call, or offered as optional ("if you have a photo, that helps").
Setting it up, step by step
- Write your three or four must-know questions in plain words, the way you would say them.
- Decide what makes a good fit. For a plumber that might be "inside our area, not a new installation". Write it as one or two lines.
- Decide your tags. Three are enough: ready to call, needs more info and not a fit. Spaces lets your agent tag a conversation, which sends nothing to the customer.
- Ask your AI agent to add the questions, the fit rules and the tags to your business brief, the note in Spaces it reads before replying. Ask it to change only that part, so your hours and prices stay as they are.
- Hold replies as drafts for the first few days and read what the agent would have said.
- Send yourself three test enquiries from a friend's phone or a second account: one perfect fit, one outside your area and one vague "how much?". Check each reply and each tag.
- Once the drafts are right, let the agent send on its own, and ask it for the ready-to-call list at set times each day.
If you use a chat app such as ChatGPT, Claude or Grok, it only acts when you ask it something in a chat, so it will not watch your messages on its own. An always-on agent such as Claude Code, Codex, OpenClaw or Hermes can run continuously and pick up each message as it arrives.
A sample note for your AI agent
This is the kind of note a heating engineer might add to the brief. Swap the questions and the area for your own.
New enquiries: before I call anyone back, find out:
1. Their postcode (we cover HX, HD and the west of LS).
2. What the problem is, in their words.
3. Whether water is leaking or there is no heating or hot water right now.
Ask one or two questions at a time, never a list of five.
Thank them first and keep each message to two short sentences.
If they give a postcode outside our area, say so kindly and suggest
they search for a Gas Safe engineer nearer to them.
Tags: "ready to call" when you have all three answers and they are
in our area. "needs more info" if they stop replying. "not a fit" if
they are outside the area or want a brand-new system fitted.
Never give a price for a repair. Say: "Jo will call you to talk it
through. When is a good time today?"
If they smell gas, tell them to call the National Gas Emergency
number straight away and hand the chat to Jo.Notice the last line. A qualifying note should always say what happens with anything urgent or unsafe, because the agent will meet those messages too.
Example: a heating enquiry on WhatsApp
Names and details are examples.
| Who | Message |
|---|---|
| Customer | Hi how much to look at a boiler |
| Agent | Hi, thanks for getting in touch. Could you tell me your postcode and what the boiler is doing? |
| Customer | HX3, it keeps losing pressure and the heating cuts out |
| Agent | Thanks, that is in our area. Is there any water leaking at the moment, or are you without heating or hot water right now? |
| Customer | No leak, hot water is fine, heating is on and off |
| Agent | Thanks, that helps. Jo will call you to talk it through and book a visit. When is a good time to ring today? |
| Customer | After 4 is best |
| Agent | Lovely, I will let Jo know to call after 4. |
Four short replies, and Jo has the postcode, the fault, the urgency and a time to ring. The agent tags the chat "ready to call" and never mentions a price.
Example: a LinkedIn enquiry for a consultant
Names and details are examples.
| Who | Message |
|---|---|
| Prospect | Hi Sam, saw your post on pricing. Do you work with software companies? |
| Agent | Hi Anna, thanks for reaching out. Yes, Sam works with software companies. What are you hoping to change about your pricing? |
| Prospect | We have three plans and nobody buys the middle one. Want to fix it before our next launch |
| Agent | That is a common one. When is the launch, roughly? |
| Prospect | March |
| Agent | Thanks. Sam usually starts with a 30-minute call. Would next week suit you, and is there anyone else from your team who should join? |
On LinkedIn the tone is more formal and the questions are about the company rather than a postcode, but the pattern is the same: what they need, when they need it, and who decides. The agent leaves budget for the call, which suits most consultants.
What the call-back list should look like
Ask your agent for the list in a fixed format, so you can read it in thirty seconds between jobs. For example:
Show me every conversation tagged "ready to call" from today, as:
name | channel | what they need | where | how urgent | best time to call
Then list anything tagged "needs more info" with the question they did
not answer.A good list fits on one phone screen. If yours is longer, either the day was busy, which is good news, or the agent is tagging too many people as ready, which usually means your fit rules need one more line.
When the agent should stop asking and hand over
- The customer asks to speak to a person, or asks for a call straight away.
- Anything urgent or unsafe, such as a leak, a gas smell, a locked-out tenant or an injury.
- A complaint about past work, even if it arrives as a new enquiry.
- A large or unusual job that does not fit your normal questions.
- The customer is clearly annoyed by the questions. One apology and a hand-over is better than a third question.
Spaces lets your agent hand a conversation to a person on your team. The agent then stays out of that chat, with no replies, scheduled messages or routines there, until it is handed back. Our guide on handing a chat from your AI agent to a person covers the wording.
The honest limits
An AI agent can only work with what the customer tells it. It cannot check that a postcode is real, that a budget is genuine or that the person is who they say they are. It can misjudge a vague message, so a weekly look at the "not a fit" tag is worth doing, to make sure nobody good was turned away.
Spaces does not make phone calls. The agent gets the customer ready for your call; you or your team still ring them. It also only answers people who wrote to you, in the chat they started. It is not a tool for finding strangers to qualify.
Finally, platforms restrict automation on personal accounts, and WhatsApp, Instagram and LinkedIn each have their own rules. Spaces paces every message and keeps each account inside safe sending limits, which lowers the risk, but no tool can promise a platform will never restrict an account.
Common mistakes
- Asking every question in the first message. It reads like a form, and many people stop there.
- Making the agent sound like a gatekeeper. The customer should feel helped, never screened.
- Letting the agent quote for work you always price on site.
- Forgetting what happens to urgent messages, so a leak gets the same three questions as a dripping tap.
- Never reading the "not a fit" tag. A few of those will be wrong in the first weeks.
- Calling back the next day. The point of qualifying in the chat is that you can ring the ready ones while they are still thinking about it.
Who you message
Summing up
Pick three or four questions, write them into the brief with one or two fit rules and three tags, test with a few fake enquiries, then let the agent run the first conversation. Ask for the ready-to-call list at the same times each day and ring those people first. You will spend your phone time on customers who are ready, and everyone else still gets a quick, friendly answer.
Pricing
FAQ
How many questions should the agent ask?
Three or four at most, and one or two at a time. If you need more detail, ask for it on the call. Long lists of questions are the most common reason customers stop replying.
Will customers mind answering questions from an AI agent?
Most people are happy to answer a couple of quick questions when it means a faster, better call back. Keep the tone warm and explain why you are asking, such as "so Jo can tell you if we can help". Some owners also tell customers they are talking to an AI, which our guide on that question covers.
Should the agent ask about budget?
It depends on your business. For consultants and wedding suppliers a rough range can save everyone time, so a gentle question works. For trades it often lands badly in a chat, so leave it for the call.
What happens to people who are not a fit?
The agent tells them kindly and, where you have one, points them somewhere better. It tags the chat "not a fit" so you can check those now and then and make sure the rules are not too strict.
Can the agent book the call for me?
It can ask when suits the customer and pass that on with the call-back list. Spaces does not connect to calendars or booking systems, so if you want the agent to put the call in your diary, it needs its own calendar tool set up separately.
Does this work on email as well as chats?
Yes. On email the agent can ask the same questions in a short reply on the same thread. Emails tend to arrive with more detail already, so it often needs only one follow-up question.
What if the customer answers with a voice note or a photo?
Spaces lets your agent fetch photos, voice notes and files from a message. Whether it can understand a photo or a voice note depends on the AI agent you use, so test this with your own agent and tell it what to do if it cannot, such as asking for a short written answer.
Do I need a card to try this?
No card is needed to start. The Free plan has 1 account and 25 actions a day until you add a card, then 50 a day. You stay on the Free plan. Pay as you go is only for when you want it.