You’ve built your first AI automation offer. The tech works, the demo goes well, and then the business owner asks the one question that stops most agency owners cold: “So what does this cost me?”
Silence. A shrug. A number pulled out of thin air.
We’ve watched that moment kill more deals than bad tech ever has. Pricing AI services isn’t guesswork, it’s a repeatable framework. In this guide we’ll cover how to charge clients for AI services across six proven pricing models, how to diagnose a business before you quote, and how to anchor your price against a return the client already understands.
Why Pricing Is the Real Bottleneck for AI Agencies
Most people building an AI automation agency get stuck not because their build is broken, but because they can’t confidently answer “how much should I charge?”
Once your pricing is nailed, the rest of the sales process gets easier. Clients move from “let me think about it” to “when do we start.” AI is no longer a buzzword, it’s a tool that streamlines tasks, increases efficiency and improves customer engagement, and clients can see that value clearly. That makes it far easier to negotiate favourable terms and lock in long-term partnerships, provided you know what to ask for.
Diagnose Before You Price: The Doctor's Diagnosis Method
Never quote a price before you understand the business. We call this the doctor’s diagnosis, and it’s the most overlooked step in pricing AI services.
On a discovery call, ask these questions before you say a single number:
- How many leads do you generate each month?
- How many of those leads turn into booked appointments?
- What’s your average close rate?
- What’s a customer worth to you in gross profit?
- What’s your gross profit per sale?
- How large is your database of old or dormant leads?
You’re absorbing information here, not pitching. Write it down. Only move to pricing once you understand the patient in front of you.
These numbers do two things. They tell you which of the models below actually fits, and they give you the anchor you’ll price against later. An agency that quotes before asking these questions is guessing, and clients can tell.
Start With an Offer That Removes Risk
Before you price anything, you need an offer that takes the risk off the client’s shoulders. The one we use with students is simple: turn a business’s old, dead leads into booked appointments or sales using AI and SMS, on a pure performance basis. If they don’t get paid, we don’t get paid.
No upfront fees, no long contracts, no tech jargon. That performance-first structure is exactly why AI automation agencies are winning deals that traditional agencies can’t touch, and it makes every pricing conversation below far easier to have.
1. Performance-Based Pricing
Performance-based pricing is the most attractive model for AI automation agencies because it directly links compensation to results. You charge only when specific outcomes are achieved, which reduces the client’s risk and incentivises you to deliver.
Say you’re working with a mortgage company, using AI to increase lead conversion. Performance-based pricing might look like:
- Per appointment: you get paid each time the AI secures a lead appointment.
- Per show-up: you earn a fee when the lead actually attends.
- Per sale: your highest fee comes when the lead converts.
Depending on the industry, that can run from $50 per appointment up to $3,000 per sale.
Working out your per-appointment price
Here’s the maths in practice. A client generates 100 leads a month and has 5,000 dormant leads sitting in their database. Each sale is worth $3,000 in gross profit, and they close roughly 1 in 10 appointments.
At $120 per booked appointment, the client pays $1,200 to make $3,000 from a single sale. That’s a return that’s very difficult to argue with, and you’re not asking them to gamble anything upfront.
A useful rule of thumb when you’re setting a per-lead or per-appointment number: divide the average order value by three, then apply the client’s conversion percentage.
Profit share (backend) pricing
If you’d rather take a cut of revenue than a flat fee, offer a 50/50 profit share on sales generated from the client’s database. Using the same $3,000 deal, that’s $1,500 per sale, and the client pays nothing until cash actually lands in their account.
Profit share works best where the sales cycle is short and attribution is clean, such as ecommerce. Per-appointment works better when the cycle is longer or harder to track.
Tracking and transparency
Clients will ask how you’ll track and report performance. Transparency is what wins here:
- CRM access: request access so you can monitor progress directly.
- Tagging systems: set up tags or custom fields so both parties see conversions in real time.
- Reporting: for smaller clients without a robust CRM, a simple daily report is enough.
Agree on verifiable anchor points before you start, so there’s no argument later about when payment is due. Aligning your pricing to their own CRM data builds trust and turns the client from a customer into a partner.
2. The Anchor Point Method
This isn’t a separate pricing model so much as the technique that makes every model above land.
Use a number the client already understands, their cost per appointment, their customer value, their current cost per acquisition, and frame your price against it. When a client says “I pay $12 to make $3,000,” the maths does the selling for you.
Anchoring protects your margin while making your offer feel like an obvious return rather than another line on the expense sheet. Without an anchor, your price is just a number they’re comparing to zero.
3. Upfront Drawdown
An upfront drawdown facility combines performance pricing with an upfront payment that covers your initial costs.
The client prepays for a set number of outcomes, usually appointments, which you deliver over an agreed period. If you estimate 100 appointments at $100 each, you take $10,000 upfront and draw it down as you deliver. If you fall short of the quota, you refund the difference.
Clients get a performance structure with a clear cap on spend. You get working capital from day one, which matters when you’re paying for SMS and AI usage before a single sale closes.
4. SaaS Model With Performance Add-Ons
For clients who prefer a predictable flat rate, a SaaS model works well. They pay a monthly fee to license your AI tool or chatbot, and you deliver the automation without carrying the variable cost of a pure performance deal.
A SaaS model might include:
- Monthly subscription fee: a flat rate, for example $497, for access to the tool.
- Performance-based add-ons: a percentage of each successful lead or sale on top.
This hybrid suits clients who want ongoing support and room to scale. Agencies running it well have reached $20,000 a month by stacking performance commissions on top of the base fee.
5. Retainer Model
A retainer gives you consistent monthly revenue and suits clients with complex or large-scale AI needs. You take a fixed monthly fee to support the business across multiple areas, from lead generation to customer support automation.
The important question is when to introduce it. A retainer works when the client already has a strong database, an established sales team, and a genuinely good offer. Layer in a maintenance fee or per-asset charges as you build out more automations.
What it shouldn’t be is your opening position with a client who has never worked with an AI agency. Prove the value on performance first, then move the relationship to recurring revenue once the results are undeniable.
6. Project-Based Fees and Setup Costs
For one-time builds, a project fee is the right shape. Setting up a chatbot, integrating AI with a CRM, or configuring document collection automation all fit here, often with an optional maintenance retainer attached afterwards.
Performance pricing doesn’t mean working for free upfront either. You can add a setup fee to any of the models above. In practice these run from $500 to $10,000, and some agencies have charged $20,000 for larger, more complex builds.
Alternatively, pass SMS and AI usage costs directly to the client, or bake them into your per-appointment or backend rate. Pick one and be explicit about it in writing, because unrecovered usage costs are what quietly turn a profitable deal into a break-even one.
7. High-Impact Problem-Solving Fee
When you’re solving a problem that generates significant value, price against that value rather than against your hours.
A large mortgage company might save millions a year by automating document processing. Negotiating a fee based on their realised savings positions you as indispensable rather than as a vendor, and deals structured this way often lead to exclusivity agreements or long-term partnerships.
This is the model with the highest ceiling and it requires the most confidence. You need the diagnosis from earlier in this guide to make the case credibly.
| Model | Best for | Client risk | Your cash flow speed |
|---|---|---|---|
| Per appointment | Beginners, no case studies yet | Very low | Fast, paid per booking |
| Profit share (50/50) | Clients wary of upfront costs | Very low | Fast, once the sale closes |
| Upfront drawdown | Clients who want a capped spend | Low | Immediate, full prepayment |
| SaaS + performance | Clients wanting predictable tooling | Low | Predictable, recurring |
| Setup fee + backend | Larger builds, complex automations | Low to medium | Immediate partial payment |
| Monthly retainer | Established clients wanting ongoing work | Medium | Predictable, recurring |
| High-impact fee | Enterprise cost-saving projects | Medium | Negotiated, often staged |
Always Start With a Test Deal
You might worry about a client understating results to avoid paying you. That’s exactly why every deal should begin with a small test run before you commit long term.
A test shows you how the client communicates, how their sales team actually handles the leads you send, and whether they’re transparent about outcomes. If you see red flags, blame-shifting, missed follow-ups, vague reporting, walk away. A small test protects you from carrying risk with an unreliable partner, and it costs you a fraction of what a bad six-month deal does.
There’s a second benefit if you’re early. A test deal on performance terms removes the client’s risk entirely, and the results become your first case study.
Why Your Margins Are Better Than a PPC Agency's
AI and SMS delivery costs are a fraction of what paid advertising costs, which is the structural reason you can afford to take performance risk that a traditional PPC agency can’t.
SMS open rates sit at around 98% according to the Mobile Marketing Association, and reaching a dormant database costs cents per contact rather than the rising cost per click on Meta and Google. That gap is your margin, and it’s what lets you say “if they don’t get paid, we don’t get paid” and still run a profitable business.
There Is No Single Right Way to Charge
Some clients want performance-only deals. Others are used to paying upfront and prefer it that way. Some like a monthly retainer, and some want to dip a toe in first.
Your job isn’t to force one model onto everyone. It’s to ask the diagnostic questions, listen to the answers, and match the structure to what that specific business needs. The agencies that grow fastest are the ones flexible enough to run three or four of these models across different clients at the same time.
Common Questions About Charging for AI Services
What if I don’t have case studies yet?
Start with a small test deal on a performance basis. It removes the client’s risk, and the results become your first case study.
How do I know whether per appointment or profit share is better?
Use per-appointment pricing when the sales cycle is longer or harder to attribute. Use profit share when the cycle is short and attribution is clean, such as ecommerce.
Should I charge a setup fee on a performance deal?
You can, and most established agencies do. Anywhere from $500 to $10,000 is normal depending on build complexity. It covers your costs and filters out clients who aren’t serious.
How do I stop a client under-reporting results?
Get CRM access, agree on tagging and anchor points before you start, and run a test deal first. If a client won’t give you visibility into their own numbers, that’s your answer.
Final Thoughts on How to Charge Clients for AI Services
There’s no one-size-fits-all pricing model for AI services, but every option here gives you a way to align your revenue with the value you create.
Diagnose the business properly, choose the model that fits their numbers, anchor your price against a return they already understand, and start with a small test. Do that and you remove almost all of the friction that stops businesses saying yes.
With AI automation driving growth across industries, your expertise is more valuable than it has ever been. Price it accordingly.
For the full framework, including the exact scripts we use on client calls and how we handle pricing objections, take a look at The Instant AI Agency.
What pricing model are you using right now, per appointment, profit share, or something else? Drop a comment and let us know what’s worked.
