Pricing AI services: retainers, credits or outcomes?

AI makes some work so fast that billing by the hour collapses you end up punished for being efficient so a successful AI practice prices the outcome, not the effort. But “outcome pricing” is not the answer to everything; it only works when the result is measurable and within your control. The pragmatic playbook that works for most agencies and freelancers is this: start with packaged credits, add monthly retainers once trust is built, and use outcome pricing sparingly, on specific deals where you can genuinely measure and own the result. Here is why hourly billing breaks, how each model works, and how to decide including how to price your own work when your inputs have become cheap.

Why hourly billing falls apart with AI

The core problem is simple and it catches everyone who starts selling AI work: an AI-powered workflow can do in minutes what took a team member a day. If you charge by the hour, your revenue collapses as you get faster the exact opposite of how a healthy business should behave. Efficiency becomes a liability.

Hourly billing also breaks on two other fronts:

• It rewards the slow and penalises leverage. A provider who has built a great template that finishes in an hour makes less than one who fiddles for eight. Nothing about that makes sense to the client, and it actively discourages you from investing in better tooling.
• It anchors the client to the wrong thing. Buyers stop thinking about the value of the outcome and start scrutinising how your minutes were spent. That is a bad conversation to have, and it gets worse the faster your work gets.

This is not an edge case. It is the reason every agency doing serious AI work has had to change how it prices. The move away from hourly is a consequence of the technology, not a preference.

The three models

1. Monthly retainers

How it works: the client pays a flat monthly fee for ongoing access to you and your AI systems support, maintenance, improvements, and a baseline of deliverables. Best for: ongoing support and maintenance; the relationship where the client wants you “on call” and you want dependable, recurring revenue. Strengths: predictable income for you, predictable cost for the client, and a deepening relationship so you are not re-selling yourself every month. Retention beats acquisition; recurring revenue is far easier to build on than project-to-project. Watch out for: scope creep define clearly what is included, and price the retainer to cover not just the happy-path work but the maintenance AI actually creates: model changes, prompt updates, data cleanup, and the occasional “why did this break” session. The maintenance layer is what a retainer is really paying for.

2. Credits or packages

How it works: the client buys a bundle of deliverables in advance so many posts, so many automations, so many designs and redeems them as needed. Best for: content, design and defined automation builds where the scope is project-like and the deliverables are countable. Strengths: the easiest model to sell and understand; it front-loads revenue, gives the client predictability, and defines scope so there is no ambiguity about what “done” means. It is the natural starting point for a new practice. Watch out for: under-pricing out of embarrassment when your own costs are small. If producing the deliverable costs you almost nothing in dollar terms, that is not a reason to charge less price on the value and effort of the outcome, the scope of the deliverable, and the skill behind it. Cheap inputs should raise your margin, not lower your price.

3. Outcome-based pricing

How it works: you charge based on results leads generated, tasks automated successfully, revenue influenced, hours saved. Best for: where results are cleanly measurable and within your control (for example, a fixed number of qualified leads per month or an automation that demonstrably cuts the client’s workload). Strengths: the highest upside and the strongest client alignment you win when they win, which makes the pitch compelling and the relationship less adversarial. Watch out for: the two conditions are non-negotiable. If you cannot cleanly measure the outcome, or if the result depends mostly on the client’s own funnel, sales effort or timing, then outcome pricing gives away your upside for influence you do not have. It is a strategic weapon, not a default offer it only where you hold the levers.

How to decide – the sequence

Model	Best when	Main weakness
Monthly retainer	Ongoing support, predictable relationship	Scope creep; maintenance must be priced in
Credits / packages	Project-able deliverables (content, builds)	Easy to under-price if you forget value
Outcome-based	Metric is clear and you control it	Unmeasurable or client-controlled = bad deal

The sequence that works for most agencies and freelancers:

1. Start with packages. They are the easiest to sell and understand, they define scope, and they give you fresh revenue while you prove yourself. Get three or four package clients under your belt before anything else.
2. Add retainers once trust is built. Move your best package clients onto monthly recurring. It smooths cash flow, reduces re-selling, and deepens the relationship. The maintenance layer you discovered while delivering packages becomes exactly what the retainer is for.
3. Use outcome pricing sparingly. Offer it only on deals where you can measure the metric and own the levers that move it. That is where it is a powerful differentiator; everywhere else it is a risk you should hand back to the client in the form of a normal fixed or package price.

How to actually price when your costs are near zero

This is the question every AI agency hits. Three principles keep you from underpricing:

• Anchor to value, not production cost. The client is buying a result an automated process, published content, a working system. Price the value of that result and the scope and skill behind it, not what it costs you to produce.
• Use tiered offers to give yourself headroom. A light / standard / premium set lets the buyer self-select and raises your average sale. Most buyers do not choose the cheapest tier, and the tiers create an obvious “better value” option that your standard tier becomes.
• Keep a floor beneath your packages. Even if your inputs are cheap, the maintenance, the decisions, the customisation and the accountability are real work. A package that prices in the effort and accountability not just the raw generation is one you can actually stand behind when something breaks.

Selling the pricing change to clients

Switching from hourly to value or package pricing is a conversation, not just a decision and doing it badly is where most agencies lose trust. Three moves make it work:

• Lead with the outcome, not the mechanism. Do not sell “I’ve moved off hourly billing.” Sell “you get a defined deliverable, predictable cost, and I am accountable for it.” Clients accept a new model when they clearly understand what they are buying and why it is better for them.
• Put the value into words before the price. A client resists a £2,000 package until they see it replaces several hundred pounds of ad hoc hourly work and returns more. Frame the package against the alternative cost or the outcome value, not against your old hourly rate.
• Protect your new model from old habits. Scope creep arrives disguised as “just one more small thing.” A tight scope-in/scope-out boundary in the package keeps the value pricing honest and prevents the model from silently collapsing back into unbilled hours.

Done well, the pricing conversation becomes a positioning moment: you look like a modern, outcomes-focused provider rather than a vendor of hours which is exactly what a client trying to decide between you and a reseller is looking for.

Pricing mistakes to avoid

Beyond the models themselves, a set of recurring errors sinks agencies regardless of which model they choose. Knowing them is half the battle:

• Under-pricing out of uncertainty. If you do not know what to charge, you charge too little. Anchor to the value of the outcome and the scope, not to what feels comfortable, and test your prices with real prospects rather than guessing in a vacuum.
• Letting one bad client distort the model. A single massive client on outcome pricing, or one tiny client on an over-generous package, can quietly warp your whole approach. Price the offer for its market; exceptions are deliberate, not accidental.
• Quoting in effort when you mean value. “This will take me two hours” invites the hourly mindset right back in. Even on a package or retainer, describing what you deliver and the outcome it produces keeps the conversation where it belongs.
• Skipping the scope boundary. Every model needs a crisp line for what is included and what costs extra. Without it, scope creep converts predictable packages back into unbilled interventions and undermines the model’s whole point.
• Being unwilling to raise prices. Your packages and retainers should climb as your evidence and results grow. Raising prices for each new client (not just existing ones) is how a practice compounds; staying static is how it stays small.

Pricing is a system, not a single decision. A model you pick, a disciplined scope, and a practice of testing and raising prices together is what turns AI leverage into a real, growing revenue base.

The three models are not rivals; they are a ladder. Packages get you in and paid. Retainers make you recurring and reliable. Outcome pricing, deployed selectively, is the high-upside play for the deals where you genuinely control the result. Whichever you use, the principle is the same: price the outcome the client wants, not the hours you might save with AI. Do that, and your speed and leverage become margin instead of a clock which is exactly what choosing AI work over commoditised hourly work is supposed to buy you.

---

FAQ

Why can’t I just bill AI work by the hour? Because AI makes the work fast every hour you save becomes revenue you lose. Hourly billing penalises the efficiency that is your whole advantage. Price the value or the outcome instead of the effort.

What is the safest pricing model for a new AI agency? Credit packages. They are easy to understand, define scope, and generate revenue while you build trust. Add retainers and, later, outcome pricing as you gain client confidence and evidence.

When is outcome-based pricing a bad idea? When the result is not cleanly measurable or depends on factors outside your control the client’s funnel, sales effort or timing. In those cases outcome pricing gives away your upside for influence you don’t have.

How do I price AI work when my own costs are tiny? Don’t anchor on your production cost. Price on the value and effort of the outcome to the client, the scope of the deliverable, and the maintenance behind it. Cheap inputs should increase your margin, not lower your price.

What is the difference between a retainer and a package? A retainer is a recurring flat monthly fee for ongoing support and maintenance. A package (or credit bundle) is a one-off purchase of a defined set of deliverables in advance. Most practices start with packages and convert repeat clients to retainers once trust is built.