What to Charge for an AI Chatbot or Receptionist Setup

Charging for an AI chatbot or receptionist is not simply a matter of adding a markup to software costs. The client is paying for discovery, configuration, testing, integration and the unglamorous work of making the thing reliable.

Tickd is an independent resource. Nothing here is financial advice or a promise of income. Results vary widely and many people earn nothing. Do your own research.

This guide gives you pricing anchors, not supposed industry averages. Your quote should reflect the scope, risk and value of the actual project. If you are still deciding what service to offer, start with AI automation for local business and AI Automation for Small Businesses: Where to Start.

What's realistic

For a beginner, the realistic first job is a narrowly defined assistant: answering common questions, collecting lead details, routing enquiries or handling calls outside opening hours. It is not an all-knowing digital employee that replaces reception staff.

Small projects may take longer than expected because business information is scattered across websites, PDFs, inboxes and people's heads. Phone receptionists add further work: call flows, pronunciation, background noise, transfers, voicemail, consent messages and failed-call handling.

A useful starting framework for setup fees is:

• £500–£1,500 for a basic website chatbot using an established platform, limited content and one simple lead-capture workflow.
• £1,500–£4,000 for a more involved chatbot with multiple workflows, integrations, structured testing and staff handover.
• £2,000–£6,000 or more for an AI phone receptionist involving call routing, booking systems, custom integrations and careful fallback handling.

These are editorial pricing anchors, not reported market averages or promises of what clients will pay. A tightly scoped project can fall below them; complex integrations, regulated information or multiple locations can push a quote higher. Beginners should resist quoting complex deployments until they can test them properly.

For broader context on turning technical ability into a defined service, read How to Become an AI Consultant for One Industry.

What the setup fee should cover

List deliverables rather than selling a mysterious bundle of “AI automation”. A sensible setup scope might include:

• A discovery call and written requirements.
• Review and organisation of approved business information.
• Conversation or call-flow design.
• Platform configuration and prompt development.
• One or two agreed integrations.
• Testing against realistic questions and failure cases.
• Human handover and escalation rules.
• Installation on the website or connection to the phone workflow.
• Basic staff training and documentation.
• A defined post-launch correction period.

You can use Tickd's Agent Builder to map the agent's role, tools, boundaries and escalation behaviour before committing to a platform. That planning step often exposes vague requirements before they become unpaid work.

State what is excluded too. New integrations, additional locations, multilingual support, major knowledge-base rewrites and unlimited revisions should not quietly sneak into the original fee.

Separate setup, usage and maintenance

One price should not disguise three different cost categories.

Setup

This is your one-off project charge for planning, building, testing and launching. Quote it against named deliverables and assumptions.

Platform and usage costs

Chatbot platforms may charge by message, conversation or seat. Voice systems can involve phone numbers, call minutes, transcription and model usage. These costs should usually be paid directly by the client or recharged transparently.

Do not advertise “unlimited AI calls” unless the underlying service genuinely supports that arrangement and you have accounted for heavy usage. A busy campaign can make a cheap-looking fixed fee rather less charming.

Ongoing support

After launch, the client may need content changes, conversation reviews, prompt adjustments and integration monitoring. Offer either:

• An hourly support rate.
• A monthly plan with a clear allowance.
• Prepaid blocks of support time.

Define response times and unused allowances. A support plan is not a promise that the system will never fail; it defines what you will do when something needs attention.

Build your quote from effort and risk

Estimate each stage separately rather than guessing a number that sounds affordable:

• Discovery and process mapping.
• Content preparation.
• Build and integration work.
• Testing and corrections.
• Deployment and training.
• Project management and contingency.

Apply a higher contingency where you depend on undocumented software, third-party access or client-supplied data. If you cannot inspect an integration before quoting, make the estimate conditional or charge for a paid discovery phase.

Risk matters as much as time. A chatbot that collects contact details is relatively contained. A phone agent that books urgent appointments, discusses sensitive information or represents several branches needs stricter testing and human escalation. Charging more does not remove that risk, so decline work you cannot deliver responsibly.

A simple pricing example

Illustrative example — not a real result

Suppose a local clinic wants a website chatbot that answers approved administrative questions, collects enquiry details and sends them to an existing inbox. You estimate:

• 4 hours for discovery and content review.
• 10 hours for configuration and conversation design.
• 6 hours for testing, revisions and launch.
• 5 hours of contingency for client changes and platform quirks.

That is 25 hours. At a hypothetical project rate of £60 per hour, the setup calculation is £1,500. Third-party platform charges are separate, and the quote includes one integration, one revision round and 14 days for correcting agreed-scope defects.

The point is not that £1,500 is the correct price. The useful bit is showing how the figure connects to work, assumptions and boundaries. If the client adds appointment booking or phone support, revise the scope rather than absorbing it.

How to present the proposal

Keep the proposal plain enough that a business owner can understand it without an AI glossary open in another tab.

Include:

• The business problem and intended outcome.
• Exactly what the assistant will and will not do.
• Deliverables and project stages.
• Client responsibilities, including content approval and access.
• Setup fee and payment schedule.
• Estimated third-party costs and who pays them.
• Revision, support and cancellation terms.
• Privacy, data retention and human-escalation considerations.
• Acceptance criteria for launch.

Avoid selling certainty. Say the system is designed to answer from approved information, not that it “eliminates errors”. Do not claim it will replace staff or produce a particular number of bookings.

A short paid pilot can suit uncertain projects. Limit it to one channel, one workflow and measurable technical criteria such as successful routing or accurate retrieval from approved material. A pilot tests feasibility; it does not guarantee commercial results.

Common pricing mistakes

• Quoting before seeing the client's systems and content.
• Charging only for time spent inside the chatbot builder.
• Bundling unlimited edits into a fixed setup fee.
• Paying all usage costs yourself without a cap.
• Ignoring testing for malicious, irrelevant or ambiguous inputs.
• Forgetting staff training and human handover.
• Promising sales, savings or staff replacement that you cannot verify.
• Taking on sensitive workflows without suitable privacy and legal checks.

Be particularly wary of business advice built around instant riches and copied templates. “Make $10K a Month With AI” Claims: Spot the Hype explains why revenue claims often leave out costs, failed outreach and delivery work. The wider Make Money With AI hub takes the same reality-first approach.

Who this is NOT for

This service is not a good fit if you want a hands-off income stream, dislike client support or expect one template to work for every business. Local companies have different policies, software and tolerance for automation.

It is also unsuitable if you are unwilling to test edge cases, document limitations or tell a client when human handling is safer. Do not deploy an autonomous receptionist for medical emergencies, legal decisions or other high-stakes situations without appropriate professional oversight and safeguards.

Finally, avoid clients who demand deceptive impersonation, hidden call recording, fabricated reviews or misleading claims that callers are speaking to a human. Walk away rather than trying to price around unethical requirements.

FAQ

Should I charge hourly or a fixed fee?

A fixed fee works well when the deliverables and revision limits are clear. Hourly pricing is safer for discovery, uncertain integrations and open-ended changes. You can combine them: paid discovery first, followed by a fixed implementation quote.

Should the client pay the AI platform costs?

Usually, yes. Having the client own the relevant accounts makes usage and billing clearer and reduces awkwardness if the relationship ends. If you recharge costs, explain the calculation and any management fee in writing.

Do I need a monthly maintenance plan?

Not every project needs one, but the client must know who handles outdated information, platform changes and failures. Offer maintenance as an option with defined tasks, response times and limits rather than an unexplained recurring charge.

Can I charge more for an AI phone receptionist?

Often, because voice projects usually require more testing, telephony configuration and failure handling than a simple text chatbot. Price the actual complexity, call volume assumptions and integrations—not the fashionable label attached to the service.