This article is about the business of selling chatbots to other businesses: what clients pay for, how you find them, how delivery and support work, and whether the economics hold up. For the hands-on build, see our no-code WhatsApp chatbot guide; for the "bot vs hiring a person" question, see AI customer support tools vs a freelance VA. This page assumes you can build one and asks whether you can sell one.
What businesses actually pay for
Not "a chatbot." They pay to stop losing enquiries after hours, to cut the volume of repetitive questions their staff answer, to qualify leads before a human calls, or to book appointments without phone tag. The chatbot is the mechanism; the outcome is what's on the invoice. If you can't tie your build to one of those outcomes for a specific business, there's nothing to sell.
Good-fit clients
- Steady volume of the same questions (hours, location, pricing, availability, "do you do X").
- Enquiries arrive outside staffed hours or faster than staff can answer.
- A booking or quote step that's currently manual.
- Not a tech company — they won't build it themselves — but organised enough to give you the answer content.
- Examples of the shape: clinics and salons, property/lettings, trades with a quote process, tutoring, small e-commerce with pre-sale questions.
Poor fit: businesses whose questions are all unique or emotional, very low enquiry volume, or no one internally who can own the content and sign off changes.
Discovery questions before you quote
- What are the five questions customers ask most? Roughly how many a week?
- What happens now when one comes in after hours?
- Where should the bot live — website, WhatsApp, Instagram DMs, Facebook?
- What should it never try to answer, and where should it hand off to a person?
- Who on your side will keep the answers up to date?
- Do you have opening hours, price list, FAQ, booking link in a document already?
- What does an after-hours enquiry that never gets answered typically cost you?
Simple service packages
| Package | Scope | Billing shape |
|---|---|---|
| Starter FAQ bot | One channel, ~15–20 answers, business-hours message + human handoff | One-off setup fee |
| Bot + booking/lead capture | Above, plus a booking-link flow or a qualifying form that emails the lead in | Higher setup fee |
| Care plan (recommended add-on) | Monthly: answer updates, monitoring, a transcript review, small changes | Recurring monthly fee |
Set your own numbers against your local market and the value of the outcome — this article deliberately gives no universal rate. Quote the setup and the monthly separately so the client sees what's one-off and what's ongoing.
Setup vs ongoing
Setup (one-off): discovery, writing/importing answers, building the flow, styling, testing, connecting the channel, training the client's contact. Ongoing (monthly): updating answers when prices/hours/staff change, checking transcripts for questions the bot missed, adjusting flows, applying platform changes, being reachable when something breaks. The ongoing work is real — a bot with stale answers is worse than no bot — which is why a care plan matters more than a bigger setup fee.
Tool and platform categories
- No-code chatbot builders: Chatfuel, ManyChat, Landbot, Voiceflow, Tidio. Pick one and get good at it rather than spreading across all.
- Messaging channels: WhatsApp Business Platform, Facebook/Instagram, or a website widget. WhatsApp involves a separate approval and messaging cost.
- Content help: ChatGPT / Claude to turn the client's raw notes into clean answer text — you check every line.
Verify current pricing, free-tier limits, and messaging fees on each provider's own page before you quote; they change and vary by region.
Implementation workflow
- Discovery call, collect the answer content.
- Draft the flow on paper: entry points, main menu, the 15–20 answers, the handoff.
- Build in the chosen tool with the client's real data.
- Write natural answer copy; remove templated stiffness.
- Define handoff rules: unknown question, angry tone, explicit "talk to a human", anything about money or complaints.
- Test with a script of real questions plus deliberate edge cases and misspellings.
- Connect the live channel; do a soft launch.
- Train the client's named contact on updating answers and reading transcripts.
- Review after week one; fix the gaps the transcripts reveal.
Human handoff and testing
Every bot needs a clean exit to a person. Decide where handoffs go (a shared inbox, an email, a WhatsApp number), what the bot says when it hands off, and what hours a human is actually available. In QA, check: does it loop when confused? Does it claim to have done something it hasn't (booked, refunded)? Does it answer money/complaint questions it shouldn't? Does it break on a one-word or misspelled message?
Support responsibilities and recurring costs
On a care plan you're on the hook for: keeping answers current, monitoring for missed questions, and responding when the bot or channel breaks. Your own recurring costs: the builder subscription, any per-message channel fees, and your time. Price the monthly so it covers the worst realistic month, not the quiet one.
Illustrative economics
Made-up numbers to show the structure — not a claim about typical earnings.
| Item | Illustrative figure |
|---|---|
| Client need | Salon losing after-hours booking enquiries |
| Setup fee (one-off) | $X |
| Care plan (monthly) | $Y / month |
| Your monthly costs (builder + channel fees) | −$Z / month |
| Gross difference per client per month | $Y − $Z |
The model only works if ($Y − $Z), multiplied by however many clients you can realistically support, is worth your time — and if clients stay long enough for the setup effort to pay back. Fill in real figures from your market and chosen tools before deciding.
Risks and when not to sell a chatbot
- Client churns after setup, so the effort never earns back — mitigate with a minimum term or a setup fee that stands alone.
- Client won't maintain the content and blames the bot when answers go stale.
- Platform pricing or policy changes and breaks your margin or your flow.
- The business's questions turn out to be too varied for a bot to handle well.
- Expectations mismatch — they wanted "AI that handles everything", you built an FAQ router.
Don't sell one when enquiry volume is low, when no one internal will own it, when the questions genuinely need human judgement, or when a simple change (an FAQ page, an auto-reply, updated Google Business hours) would solve the problem for free.
A realistic first few months
Month one is mostly building a solid niche demo with realistic sample data. Month two is outreach and pitching, with more no's than yes's — non-tech businesses want to see a live demo before paying, so this is direct, near-in-person selling. A first paying client, if it comes, tends to land in month two or three for consistent effort, not week one. Treat each rejection as information about the pitch, the price, or the target, and adjust — then price it as a real service.