In this article (13 sections)
In short: an AI sales agent is software that answers every person who asks about your business within minutes, asks the questions your best salesperson would ask, decides whether the lead is worth pursuing, and offers a meeting slot on your calendar. It doesn't replace your sales team. It handles the first conversation, which is where most opportunities get lost to delay. In an audit published by Harvard Business Review of 2,241 companies, 23% never responded to an inquiry sent through their website.
This article explains what an agent like this does, what you should not ask it to do, and how it gets built without your team having to learn anything new.
The problem: the first reply comes too late
An inquiry that arrives through your website, an ad or WhatsApp has a short life. Someone who asked for information at 9:00 PM has probably asked two other companies too. The one that replies first, and clearly, starts ahead.
The study by Oldroyd, McElheran and Elkington, published in Harvard Business Review in 2011, measured this with a simple method: they sent a test inquiry to thousands of companies and timed the reply. According to summaries of the article, the results are uncomfortable:
- About a third replied within the first hour.
- Nearly a quarter took more than a day.
- Nearly a quarter never replied.
The study is from 2011 and was conducted with US companies, but the mechanism hasn't changed: interest cools quickly. In a small or mid-sized business, the bottleneck is usually the same. The owner or the salesperson is busy running the day's operations, and new inquiries wait.
What a sales agent does, step by step
A well-built agent covers four tasks that someone currently does by hand.
- It responds right away, at any hour, on the channel the person used: web form, WhatsApp, email or the website chat.
- It qualifies. It asks three to five questions you define: what they need, when, company size, rough budget. The answers are organized into a record.
- It decides the next step using rules you set. A contact who fits your ideal customer moves to a meeting. One who doesn't gets a polite reply, a useful resource or a referral.
- It books. It checks your calendar, proposes open slots, confirms, sends the invitation and sets a reminder.
At the end, your team gets a summary of each conversation: who the person is, what they asked, what they answered, their score and when the meeting is.
What it should not do
- It doesn't negotiate prices or promise timelines. Those decisions belong to a person.
- It doesn't make up information. It answers only with what you loaded: services, ranges, frequently asked questions. If it doesn't know, it says so and hands the conversation to a person.
- It doesn't close the sale. Its job ends when there is a meeting with someone worth talking to.
What to define before building it
The technology is the shortest part. What decides whether the agent works are the business rules, and it helps to write them down first.
Decision | Question to answer | Example |
|---|---|---|
Ideal customer | Who is a yes and who is a no? | Companies with more than 10 employees and a specific operational problem |
Qualification questions | What do you need to know to decide? | Need, urgency, size, rough budget |
Criteria to move forward | Which answers unlock a meeting? | A clear need and a decision within 90 days |
Who attends | Which person takes the meeting? | The sales lead or the owner |
When it goes to a person | Which cases doesn't the agent handle? | Complaints, current customers, unusual requests |
Channels | Where do inquiries come from? | Web form, WhatsApp, ads |
If you can't answer the row on criteria to move forward, the problem isn't technical. It's your sales process, and it's worth fixing first.
Where the agent lives
The agent connects to the tools you already use. The usual pieces are:
- WhatsApp, through Meta's official platform, which lets a custom program send and receive messages. If you're choosing a provider, WhatsApp Business API: how to choose a provider explains the differences.
- Your calendar (Google Calendar or Outlook), to read availability and create meetings.
- Your contact base: a CRM, a Google Sheets spreadsheet or a custom system where each person's record lives.
- A language model, which understands what the person writes and drafts the reply within the limits you set.
In business terms, "connected" means the agent reads and writes to those tools for you, so nobody copies data from one screen to another.
How to measure whether it works
Before you build it, set three numbers to compare against:
- Time to first reply, today and after the agent.
- Percentage of inquiries that become meetings. This is the metric that matters most.
- Meetings that actually happen, meaning the ones the person attends out of the ones booked.
A fourth signal is qualitative: read twenty random conversations every week during the first month. That's where the missing questions and the answers that need adjusting show up.
Cost and timelines
A sales agent with qualification, booking and summaries, connected to WhatsApp and your calendar, is priced like Atomix's other AI agents for business: from USD 3,000 for the agent, plus monthly maintenance from USD 300, with a free working demo in 24 to 48 hours before we quote, and under a month to production for most projects. Separately, you pay the third-party services the agent uses, such as language model usage and WhatsApp messages, which depend on volume.
For other ways companies use agents beyond sales, read AI agents for business: what they actually automate.
Frequently asked questions
Can an AI agent book meetings without mistakes?
Yes, as long as it reads your real calendar and confirms with the person before booking. Mistakes happen when the agent works with hand-copied times or badly defined time zones, which the design avoids from the start. If a case is unclear, the conversation goes to a person.
Will my customers notice they're talking to an AI?
We recommend saying so. A line at the start, such as "I'm the company's assistant. If you'd prefer to speak with a person, just let me know," protects trust and doesn't reduce response speed. What matters is that the conversation is useful and fast.
Does it replace my salesperson?
No. It takes over the work of replying, filtering and coordinating times, and hands over people who are already qualified with their context. Your salesperson's time goes to meetings and closing.
What happens to the data of the people who write in?
It's stored in your own database, not in the agent's, and you should say so in your privacy policy. Also define how long conversations are kept and who can access them.
How long does it take to go live?
With the business rules defined, a demo with your questions and your calendar can be ready in one or two days, and the full system in under a month. Most of the time goes into adjusting replies using real conversations.
If you want to see how it would work with your own qualification questions, you can book a call and we'll review your case.