AI sales agent for small business: what it does and what it costs
The jobs an AI sales agent takes off a small team, how pricing usually works, what Onpolar.ai costs in credits, and a simple way to check whether it pays off.
Published 9 min read

On this page
An AI sales agent for a small business is software that does the repetitive parts of selling for you: it answers new leads within minutes, follows up, writes to the people on your own list and books meetings in your calendar. You pay for the work it does rather than for another salary; with Onpolar.ai, that means prepaid credits from $49, with no subscription.
Key takeaways
- Four jobs, one AI agent. Answering leads fast, following up, writing to your own contacts and booking meetings.
- Two common ways to pay. Most tools charge a monthly subscription or bill for usage. Onpolar.ai sells prepaid credit bundles in US dollars, from $49 for 500 credits, valid for 12 months, with nothing to renew.
- A credit is a small unit of work. An email costs 3 credits, a chat answer 1, reading a reply 1 and a booked meeting 2.
- The payoff test is simple. Compare a month of credits with what one extra meeting is worth to you.
- Start small. Buy the Try bundle and stay in Manual mode, where you approve every email before it goes out.
What an AI sales agent does for a small business
In a small business, leads arrive while you are busy, follow-ups slip, and the list of people you meant to contact never gets shorter. An AI sales agent takes the parts that repeat and does them every day. If the idea is new to you, start with what an AI sales agent is.
It answers people who come to you
People who come to you are your inbound leads: website visitors, contact forms and emails. The agent answers your website chat on the spot, at night and on weekends too, in the visitor’s language, and replies to forms and emails within minutes. It uses only what it learned about your business and asks the questions you would ask (need, timing, team size), so you know the fit before you get on a call.
It follows up until there’s an answer
Most small teams answer the first message and forget the second. The agent follows up politely on a schedule you set, and stops the moment someone replies or asks not to be contacted. Automated lead follow-up with AI goes into this in more detail.
It writes to people you reach out to
People you reach out to are your outbound contacts: past customers, people you met at events, distributors you would like to work with. You add them or import a spreadsheet. The agent ranks them by how much they look like your best customers, reads up on each company and writes a personal first email in your tone, in any of 45 languages. It can also draft LinkedIn messages that you copy and send yourself; it never logs in to LinkedIn for you.
It books the meeting in your calendar
When someone wants to talk, the agent offers free times from your connected Google or Microsoft calendar and sends the invite with a Google Meet or Teams link, or sends your own booking link if you prefer. Each meeting comes with a short brief: who they are, what they need and what was said.
What it doesn’t do yet
Onpolar.ai writes to contacts you bring; finding new companies and contacts for you is coming soon, and so are CRM logging and WhatsApp. Today it works through website chat, Telegram, contact forms and email.
Four small-business examples
These examples are illustrative, not customer stories.
- A marketing agency of six people. New work comes through the contact form, and nobody owns the inbox in busy weeks. The agent answers every inquiry within minutes, asks about scope and timing, and books first calls in the founder’s calendar.
- A two-partner consultancy. The partners have a few hundred past contacts and conference business cards they never followed up on. They import the list; the agent writes to each person with a reason to talk, follows up twice and hands over the ones who reply with interest.
- A manufacturer that exports. It wants distributors in two new countries and already knows who they are. The agent writes to each one in their own language, during their working hours, and books calls with the interested ones.
- A professional-services firm, for example an accounting practice. Inquiries arrive in the evening, after the office closes. The website chat answers on the spot, checks what the visitor needs and offers a time for a first consultation.
The use cases page shows six situations like these, for people who come to you and people you reach out to.
How AI sales agent pricing usually works
AI sales tools are usually priced in one of three ways. None is right for everyone; what matters is how steady your volume is.
| Pricing model | How you pay | Works well when | Watch out for |
|---|---|---|---|
| Monthly subscription | A fixed fee every month, often per plan or per seat | Your volume is steady and high | Paying in quiet months; renewals you forget to cancel |
| Usage-based billing | A bill afterward for what was used | You want to pay only for real work | Bills that are hard to predict |
| Prepaid credits | You buy credits up front and spend them as the agent works | Volume goes up and down, or you want to start small | Credits that expire; buying more than you will use |
Onpolar.ai uses prepaid credits. A bundle is paid once, there is nothing to renew or cancel, and the credits stay valid for 12 months, with the ones that expire first spent first. Every bundle is the full agent; a bigger bundle costs less per credit and lets you send from more mailboxes.
| Bundle | Price (paid once) | Credits | Per 1,000 credits | Sending mailboxes |
|---|---|---|---|---|
| Try | $49 | 500 | $98 | 1 |
| Start | $199 | 2,500 | $79.60 | 2 |
| Grow | $399 | 6,000 | $66.50 | 5 |
| Scale | $990 | 18,000 | $55 | 15 |
Larger volumes (Enterprise) start at $2,490 for 50,000 credits or more, invoiced. Prices are in US dollars and exclude taxes. Credits are prepaid and non-refundable, except where the law requires a refund.
What a credit pays for
Everything the agent does with AI costs credits. The common actions:
- An email written (a first email, a follow-up or an answer): 3 credits
- A reply, inquiry or form message read and sorted: 1 credit
- A chat answer on your website: 1 credit
- A meeting booked, with its brief: 2 credits
- A LinkedIn note drafted: 3 credits
- Learning your business from your website or your answers: 10 credits
- Ranking your contacts by fit: 1 credit for every 25 contacts
Adding and importing contacts, connecting mailboxes, the daily summary and unsubscribe links cost nothing. An email is charged when it is written, so a draft you skip still counts. If your credits run out, the agent pauses until you top up; replies wait and nothing is lost. The full list is on the pricing page.
How to tell whether an AI sales agent pays off
You don’t need a forecast, only a rough monthly cost and a rough value for one meeting. Here is a worked example for a small consultancy. Every number in it is made up: replace each one with your own.
- Estimate a month of work. Count the inquiries you get, the chat messages on your site and how many people from your list you want to contact.
- Turn it into credits with the prices above.
| A month in the example | How many | Credits each | Credits |
|---|---|---|---|
| Inquiries by form or email, read and sorted | 30 | 1 | 30 |
| Answers to those inquiries | 30 | 3 | 90 |
| Website chat answers | 100 | 1 | 100 |
| People from your list, up to 3 emails each | 50 | 9 | 450 |
| Replies to those emails, read and answered | 10 | 4 | 40 |
| Meetings booked | 5 | 2 | 10 |
| Total | 720 |
Add 10 credits once, when the agent learns your business, and 2 to rank the 50 contacts.
- Turn credits into dollars. At the Start bundle’s price ($79.60 per 1,000 credits), 720 credits are about $57 for the month. The bundle’s 2,500 credits would last about three and a half months at this pace.
- Put a value on a meeting. In the example, a new client is worth $2,000 in their first year, and one first meeting in four becomes a client. So each first meeting is worth about $500.
- Find your break-even point. Divide the monthly cost by the value of a meeting: $57 ÷ $500 is about 0.11. In this example, one meeting you wouldn’t otherwise have had, roughly every nine months, covers the credits.
- Count the hours, honestly. Say you spend four hours a week answering, chasing and writing to contacts, and an hour of your time is worth $60: that is about $1,000 a month. You won’t get all of it back, because you still check the agent’s work at the start and take the calls. Count only the hours you expect to save.
Two costs sit outside the credits: your time in the first weeks, while you approve, and any extra mailbox you add for outreach, which your email provider bills as usual. In an example like this, the credits are small next to one client, so the real test is quality: are the emails good enough to send, and do the people you write to want what you sell? A small, supervised first month answers both.
What to watch out for
Your email reputation
Cold email sent too fast from a new mailbox lands in spam and can drag your main address down with it. Onpolar.ai sends from your own mailbox, starts a new one with a few emails a day and raises the number over the first weeks, adds an opt-out to every outreach email, and pauses a campaign on its own if too many emails bounce or too many people opt out. Many businesses use an extra mailbox for outreach so the main company address keeps its good name. Ask whoever manages your domain for SPF, DKIM and DMARC, the records that show your emails really come from you; the app checks them and says what is missing.
Approval at the start
Start in Manual mode, where nothing goes out until you approve it, so check in a few times a day. Read the first few dozen emails, correct what the agent knows in plain words, and switch to Autopilot when you are happy; answers then still wait 5 minutes, so you can stop one. The first email answer to someone who used your contact form or chat always waits for you to check the address, because anyone can type someone else’s email into a form.
Honesty about AI
People mind being misled more than they mind help from AI. The website chat introduces itself as an AI assistant, and when someone asks by email, the agent says that your AI assistant wrote the reply and a real person sees the conversation. When a conversation needs you, it hands it over with the full history.
Data and privacy basics
- Write only to business contacts you have the right to contact. The agent never writes to an address that opted out, bounced or is on your exclusion list.
- As the sender, you follow the law where your recipients are: in the US that includes the CAN-SPAM Act, in Europe the GDPR, and some countries require consent before a first business email. The app shows a caution for some countries; it is a hint, not legal advice.
- If you add the chat or contact form to your website, tell visitors how their data is used and link to your own privacy policy.
How to start small with an AI sales agent
- Buy the Try bundle. $49 for 500 credits: the full agent with one sending mailbox.
- Add your website, or answer a few questions. No website is fine: a few short questions about what you sell, who buys it and how buyers reach you work just as well.
- Check what it learned. Correct anything in your own words.
- Connect your email and calendar. Sign in with the Google or Microsoft account you already use, or connect any other mailbox with its mail server details.
- Stay on Manual. Add the chat or contact form to your site and import a small list, say 20 to 50 people you know well.
- Review after two weeks. Look at what was answered, who replied and what got booked. Then switch to Autopilot, move to a bigger bundle, or stop; there is nothing to cancel.
For each step in detail, see how to set up an AI sales agent.
Try an AI sales agent on a small scale
The quickest way to know whether an AI sales agent fits your business is one small, supervised month: add your website below, or answer a few questions instead, and see what it would write.



