Pricing models every AI agency should test

Key takeaways

• Move beyond hourly billing. The most profitable AI agencies use value-based models like retainers, project fees, or performance pricing that reflect the business impact of their work, not just the time spent.
• Match the model to the service. Use retainers for ongoing AI management and optimisation, project fees for custom builds, and consider performance pricing for high-impact, measurable outcomes like lead generation.
• Protect your gross margin. Your pricing must cover not just team salaries but also the significant costs of AI tools, APIs, and cloud infrastructure, targeting a sustainable 50-60% gross margin.
• Test and iterate. No single model fits all clients. The best AI agency pricing strategy involves running controlled experiments with different models to see what drives the best profit and client satisfaction.
• Define scope with extreme clarity. AI projects are prone to scope creep. Successful pricing models include clear definitions of deliverables, data requirements, and assumptions to prevent margin erosion.

Why is pricing so critical for AI agencies?

Your AI agency pricing strategy directly determines your profitability and growth potential. Many AI agencies start by billing for their time, but this model caps your earnings and fails to capture the true business value you create.

AI services are unique. They often involve high upfront development costs, ongoing model training, and expensive infrastructure like API calls and cloud computing. If your pricing doesn't account for these specific costs, your margin disappears.

A strong pricing strategy aligns what you charge with the results you deliver. It turns you from a cost centre into a strategic partner. For AI agencies, getting this right is the difference between struggling to scale and building a profitable, sustainable business.

What are the core agency pricing structures to test?

The three core agency pricing structures to test are retainers, project-based fees, and performance-based models. Each suits different types of AI work and client relationships. The most successful agencies use a mix, not just one model.

A retainer is a fixed monthly fee for ongoing services. This is ideal for managing and optimising a live AI system, like a chatbot or a recommendation engine. It provides predictable revenue for you and predictable cost for your client.

Project-based billing models involve a fixed fee for a defined deliverable, like building a custom AI model or integrating an AI tool. This works well for one-off initiatives with a clear start and end point.

Performance pricing ties your fee to a specific business result, such as cost savings or revenue generated. This is high-risk but can be high-reward, aligning your incentives perfectly with the client's success.

How do you price an AI agency retainer profitably?

Price an AI retainer by calculating all your costs for delivering the service, then adding a healthy profit margin. Your fee must cover team time, software subscriptions, API usage, and infrastructure, not just labour.

First, estimate the monthly effort. How many hours will your team spend on maintenance, monitoring, and updates? Multiply this by your fully burdened hourly rate, which includes salary, benefits, and overheads.

Next, add your hard costs. This is where AI agencies differ. You must budget for OpenAI API calls, cloud server costs, or other third-party AI tool fees. These can be variable, so build in a buffer.

Aim for a gross margin of 50-60% on your retainer. If your all-in cost to deliver is £4,000 per month, you should charge between £8,000 and £10,000. This margin protects you when unexpected costs arise and funds your growth.

Always include a clear scope of work. Define what's included (e.g., weekly performance reports, monthly model retraining) and, crucially, what isn't (e.g., major new feature development). This prevents scope creep from destroying your profitability.

When should you use project-based billing models?

Use project-based billing models for AI work with a defined outcome and timeline, like developing a proof-of-concept or building a custom AI application. This model gives clients cost certainty for a specific investment.

To price a project, break it into phases: discovery, development, testing, and deployment. Estimate the time and costs for each phase, including a contingency for technical unknowns, which are common in AI projects.

Always base your quote on value, not just hours. If your AI solution will save the client £100,000 a year in manual work, charging £30,000 is reasonable and demonstrates value. Charging based on 500 hours of work does not.

Protect yourself with a detailed statement of work. Specify the data inputs required, the acceptance criteria for the model's performance, and who is responsible for data preparation. Many AI project delays come from unclear data requirements.

Consider a hybrid model. Charge a fixed project fee for the build, then transition the client to a retainer for ongoing management and support. This captures both the initial value and the long-term relationship.

Is retainer vs performance pricing right for AI services?

The choice between retainer vs performance pricing depends on how measurable and directly attributable the AI service's outcome is. Performance pricing works best for lead generation or sales optimisation, while retainers suit complex, foundational AI work.

Performance pricing means you get paid based on results. For example, you might charge a percentage of the cost savings your AI procurement tool identifies, or a fee per qualified lead generated by an AI marketing system.

This model aligns your incentives with the client's perfectly. However, it carries more risk for you. You need deep trust in the client's tracking systems and a clear, watertight agreement on how performance is measured.

Retainers are lower risk. You get paid for your expertise and ongoing effort, regardless of the specific business outcome in a given month. This is better for services like AI strategy or model maintenance, where impact is harder to isolate.

You can blend the two. A common hybrid is a lower base retainer to cover your core costs, plus a performance bonus for hitting agreed targets. This gives you revenue stability while still sharing in the upside you create.

What are the hidden costs in an AI agency pricing strategy?

The hidden costs in an AI agency pricing strategy are typically infrastructure, data, and iteration. Failing to price for these will crush your margins. They include cloud computing, API usage, data licensing, and the time needed for model tuning.

AI models cost money to run. Every API call to OpenAI, Google Gemini, or Anthropic has a price. Cloud servers for training models or hosting applications incur hourly fees. These are direct costs of goods sold (COGS) and must be passed through or baked into your price.

Data acquisition and preparation is another major cost. Clients often underestimate the state of their data. Cleaning, labelling, and structuring data for AI can consume 50% of a project's timeline. You must charge for this work.

Model iteration is not a bug, it's a feature. AI development is iterative. You will need to test, tweak, and retrain models. Your pricing must include buffers for these cycles, not assume a perfect first attempt.

Specialist accountants for AI agencies understand these unique cost structures. They can help you build pricing models that properly account for these variables, ensuring you don't win a project only to lose money on it.

How do you test and iterate your pricing models?

Test your pricing models by running small, controlled experiments with new clients or projects. Propose different agency pricing structures to similar clients and track which delivers the best profit margin, client satisfaction, and scope adherence.

Start with a hypothesis. For example, "For our AI automation service, a value-based project fee will yield 20% higher margin than an hourly rate." Then find a suitable project to test it on.

Track everything. Measure the actual time spent, the actual costs incurred (especially API/cloud), the final profit, and client feedback. Compare this to your other projects priced on different models.

Don't be afraid to change. If a model isn't working, adapt it. Perhaps your fixed-price projects always overrun because of data issues. The solution might be to add a separate data preparation phase with its own fee.

Use a tool like our financial planning template to model different pricing scenarios before you test them. This helps you forecast the impact on your cash flow and profitability.

What metrics should you track for pricing success?

Track gross margin, utilisation rate, client lifetime value, and project overrun frequency. These metrics tell you if your AI agency pricing strategy is working. Gross margin shows if you're covering costs, while utilisation shows if your team is efficiently deployed.

Gross margin is your revenue minus the direct costs of delivering the service (team cost, freelancers, AI tools). For AI agencies, a healthy gross margin target is 50-60%. This leaves enough to cover overheads and generate profit.

Utilisation rate is the percentage of your team's paid time that is billable to clients. Aim for 70-80%. If it's lower, your pricing may be too low or your scoping too loose. If it's 95%, you're likely understaffed and at risk of burnout.

Track the frequency and cost of project overruns. If your fixed-price projects consistently go over budget, your pricing or scoping is wrong. This metric is a direct report card on your project-based billing models.

Finally, measure client lifetime value. Are clients on retainers staying longer than project clients? Does performance pricing lead to more loyal clients? This long-term view helps you choose models that build sustainable relationships.

How do you present new pricing to clients?

Present new pricing to clients by focusing on value and alignment, not just cost. Frame the conversation around how the new model better serves their business goals and reduces their risk. Position yourself as a partner invested in their outcomes.

For a move to value-based pricing, explain that you're shifting from selling hours to selling results. "Our new model ties our fee to the business impact of the AI solution, so our success is directly linked to yours."

When introducing a retainer, highlight predictability. "A monthly retainer gives you priority access to our team and predictable budgeting, ensuring your AI systems are constantly optimised without surprise invoices."

Use case studies and data. Share anonymised examples (with permission) of how a similar pricing model delivered better results for another client. Quantify the improvement in efficiency, cost savings, or revenue.

Be prepared to negotiate, but know your walk-away point. Understand your minimum acceptable margin. If a client insists on an unprofitable hourly rate, they may not be the right partner for your agency's growth stage.

Getting your pricing strategy right is a major competitive advantage. It funds better talent, better tools, and better service. If you're navigating this transition and want expert guidance, our team specialises in the commercial challenges of scaling AI businesses.