Build an AI Agent That Quotes Every Lead in Minutes

The deal was not lost on price. It was lost on silence.

It is Friday, half past six. Maria runs a small catering company. An email lands: a couple wants a quote for a 60-person wedding in three weeks. Maria sees it, thinks "I will price that properly on Monday," and closes the laptop.

By Monday the couple has booked someone else.

The deal was not lost on price. It was lost on silence.

Every small business has a version of this. The plumber who quotes a bathroom two days late. The agency that lets a warm enquiry sit in an inbox over the weekend. The shop that replies to a WhatsApp "how much?" the next morning, after the customer has already paid a competitor.

You are not losing those jobs because your price is wrong. You are losing them because you were slow, and someone else was fast.

Here is how quoting actually works in most small businesses

Be honest about the current process. It usually looks like this.

• An enquiry arrives, by web form, WhatsApp, email, or a missed call.
• It waits. You are on a job, with a customer, or asleep.
• Hours later you read it, half of it, and mean to come back to it.
• You draft a quote when you finally get a quiet hour, often days later.
• You send it, then forget to follow up, because there is always a fresher fire.

Quoting is manual. Following up is manual. Remembering who never replied is manual.

So the slow, boring, high-value work of turning interest into money gets done last, or not at all. You are the bottleneck. Not because you are lazy, but because you are one person with two hands and a phone that never stops.

A chatbot answers. An agent finishes the job.

This is the part most owners get wrong, so read it slowly.

A chatbot talks. Someone asks "do you cater weddings?" and it says "yes, we do." Nice. The customer still has no price, and you still have no lead in your book.

An agent does the job. It reads the enquiry, works out what the person actually wants, checks your real prices, drafts a proper quote in your words, sends it, writes the customer down in your records, and comes back in two days to nudge the ones who went quiet.

Same technology underneath. Completely different outcome. One makes conversation. The other closes the loop.

Here is the small version, so it feels real. A message comes in: "Hi, roughly how much for catering, about 60 people, outdoor, end of the month?" The agent replies within a minute with a friendly note and a ballpark range from your actual pricing, asks the two questions it needs to firm it up (sit-down or buffet, any dietary needs), logs the enquiry, and schedules itself to check back on Wednesday if the couple has not replied. You were driving a van the whole time.

Why this is suddenly worth building

Speed has always mattered in sales. What is new is that speed used to need a night-shift human, and now it does not.

The numbers on lead speed are brutal and old enough to trust. In the Lead Response Management Study, led by Dr James Oldroyd, firms that contacted a web lead within five minutes rather than thirty were about 21 times more likely to qualify it. The odds of even reaching the person drop by more than ten times in the first hour alone.

The Harvard Business Review put the same finding another way in "The Short Life of Online Sales Leads": companies that tried to reach a lead within an hour were nearly seven times more likely to have a real conversation with a decision maker than those who waited just one hour longer, and 60 times more likely than those who waited a day.

You cannot personally hit a five-minute window at 6pm on a Friday. An agent can, every time, without complaint. That is the whole point.

"Won't it send a customer the wrong price?"

Good. That is the right fear, and it is the difference between a toy and a tool.

You do not let the agent invent prices. You give it your real price list as its single source of truth, and it quotes from that, the same way a trained new hire would. You set the rules: what it can quote on its own, and what it must flag to you first.

A clean way to run it is a confidence score. The agent reads each enquiry and scores how clear and how ready it is, from 0 to 100. Clear, standard jobs get a quote straight away. Vague or unusual ones, the 60-person wedding with a marquee and three allergies, get routed to you with a draft already written, so you approve in ten seconds instead of starting from a blank page.

Nothing forces you to hand over full control on day one. Most owners start with the agent drafting and a human sending, then loosen the reins once they trust it.

How to build it, step by step

You do not need to write code. Tools like n8n, paired with any capable model (Claude, Gemini, or GPT), let you wire this up as a visual flow. This is the same shape a creator on the channel First Step AI walked through in a recent build, a three-agent pipeline that qualifies, quotes, and follows up on its own. Here is that shape, translated into plain business steps.

1. Catch every enquiry in one place. Point your web form, WhatsApp, and email into a single entry point (a webhook, in n8n terms). No enquiry gets to hide in an inbox anymore.
2. Log it before you do anything else. Write the lead straight into a Google Sheet or your CRM: name, contact, what they asked, time received. This sheet becomes your memory. You will never again wonder who you forgot.
3. Give the agent your knowledge. Feed it your price list, your packages, and a few example quotes you are proud of. This is what turns a generic model into your business.
4. Let it qualify and score. The first agent reads the enquiry, matches it to the right product or package, and scores it 0 to 100 on how ready and how clear it is. Add a simple check here so a broken or spammy message never slips through.
5. Split the road in two. One rule: if the score is high, treat it as a hot lead; if it is low, treat it as one to nurture. Everything after this point follows one of two paths.
6. Hot lead: draft and send the quote. A second agent, whose only job is writing quotes, produces a clean personalised quote from your price list and sends it by email or WhatsApp within the minute. Then it updates the record to "quoted."
7. Quiet lead: wait, then nudge. For the ones who are not ready, the flow pauses (a wait step that holds its place even if your system restarts), then a third agent sends a warm, low-pressure follow-up in a different tone. Not "did you get my quote," but "still happy to help if the date is firming up." It updates the record either way.

Notice the design choice that makes this work: three small agents, each with one job, not one agent trying to do everything. The one that qualifies is not the one that writes quotes, and neither is the one that chases. One brain per job stays reliable. One brain doing three jobs gets confused and starts making things up.

Manual quoting versus an agent, side by side

This pairs naturally with two agents you may have read about here already: a support agent for the easy questions that come before a quote, and an agent that chases the late invoices after the job is done. Quote, deliver, get paid. The same idea, applied along the whole line.

What you have after this

In a week, every enquiry gets an answer in minutes, even the ones that arrive while you sleep.

In a month, you stop losing deals to silence, and you can see, in one sheet, exactly how many quotes went out and how many turned into work.

In six months, quoting is no longer the job you dread on a Sunday night. It runs. You spend your hours on the work only you can do, the cooking, the building, the actual craft, while the boring, decisive minutes of first contact are handled the instant they matter.

Same business. Same prices. Same you. A completely different machine underneath.

Your first step is smaller than you think: write down what you charge, in one place, clearly enough that a careful stranger could quote from it. That single document is the fuel your agent runs on, and building it will already make your quoting faster tomorrow, with or without the robot.