The four AI agent billing models are time-based (charge per agent-hour), outcome-based (charge per deliverable), hybrid (retainer plus usage fees), and value-based (charge based on value delivered to the client).
A consulting firm bills a client £200/hour for human work. Their AI agent does the same work in 8 minutes. Do they bill 8 minutes at £200/hour — giving the client a £26.67 invoice for work worth thousands? Bill the full hour — raising ethical questions about transparency? Or rethink the entire model? Every professional services firm deploying AI agents faces this question. And there is no single correct answer. The right billing model depends on the task type, client relationship, competitive position, and the firm’s pricing philosophy. This article compares the four main models with worked examples and a decision framework.
Key Takeaway: Most PS firms will use multiple billing models — hybrid for ongoing engagements, outcome-based for defined deliverables, and value-based for strategic work.
How Does Time-Based AI Agent Billing Work?
Time-based billing charges clients per agent-hour (or agent-minute) at a published rate. It is the closest model to traditional professional services billing.
Setting the Rate
Start with the agent’s cost per hour. A mid-tier coding agent might cost £5.00/hour in model inference, tools, and infrastructure. Apply a 3-5x multiplier (the same range professional services firms use on human salaries). At 4x, the client rate is £20/hour.
Compare that to the human equivalent. A senior developer bills at £200/hour. The AI agent bills at £20/hour. The client pays 90% less for agent time. The firm maintains a 4x margin.
Typical Rate Ranges
Agent rates vary by task complexity and model tier:
| Task Complexity | Agent Cost/Hour | Client Rate/Hour | Human Equivalent/Hour |
|---|---|---|---|
| Simple (formatting, extraction) | £2-4 | £10-20 | £80-120 |
| Standard (research, analysis) | £4-8 | £20-40 | £150-250 |
| Complex (reasoning, creative) | £8-15 | £40-75 | £200-500 |
Pros
Time-based billing is familiar. Clients understand hourly rates. Invoicing is straightforward — hours multiplied by rate. Tracking is simple — log agent-hours per client. There is no need to define deliverables or measure value outcomes.
Cons
The model penalises efficiency. If your agent gets faster (through better prompts, model upgrades, or workflow optimisation), it completes tasks in fewer hours — and you earn less revenue. A coding agent that improves from 30 minutes per task to 10 minutes per task cuts your billable hours by two-thirds.
It is also hard to justify when tasks take seconds. A data extraction agent that processes a spreadsheet in 8 seconds generates a bill of £0.04 at £20/hour. The invoice administration costs more than the charge.
When to Use
Time-based billing works for routine, predictable tasks with consistent duration. It suits clients who insist on hourly billing because their procurement processes require it. It also works well for internal chargebacks between departments where simplicity matters more than margin optimisation.
Worked Example
A legal firm deploys a document review agent billed at £40/hour of active processing time. The agent reviews 200 documents in 4 hours. The client pays £160. A junior lawyer would have taken 40 hours at £120/hour (£4,800). The client saves 97%. The firm earns a 4x margin on agent costs.
How Does Outcome-Based AI Agent Billing Work?
Outcome-based billing charges per deliverable — per report, per analysis, per document set, per processed dataset. The client pays for what they receive, not how long it took.
Setting the Price
Calculate the average cost per deliverable across your last 20-50 completions. Add overhead, quality assurance costs, and margin. Set a per-deliverable price that covers the range of cost variability.
A market research report costs an average of £4.50 in agent compute (tokens, tools, infrastructure). Human quality review adds £12.00 (15 minutes of a senior analyst’s time). Total cost: £16.50. At a 4x margin, the per-report price is £66.00. Rounded for client simplicity: £75 per report.
Typical Rate Ranges
| Deliverable | Average Agent Cost | QA Cost | Client Price |
|---|---|---|---|
| Competitive analysis report | £4.50 | £12.00 | £75 |
| Contract review summary | £6.00 | £20.00 | £100 |
| Data extraction package | £1.50 | £5.00 | £30 |
| Market entry brief | £8.00 | £25.00 | £150 |
| Technical specification draft | £10.00 | £30.00 | £175 |
Pros
Outcome-based billing aligns price with client value. Clients know exactly what they will pay before work begins. Agent efficiency benefits the firm — if the agent gets faster, the cost drops but the price stays the same, increasing margin. It encourages investment in agent performance because every improvement flows directly to profitability.
Cons
This model requires clear deliverable definitions. If a “competitive analysis report” is not precisely scoped, clients may expect 50 pages when the agent delivers 10. Variable quality is hard to price — a brilliant report and a mediocre one cost the client the same. Scope creep is a constant risk on loosely defined deliverables.
When to Use
Outcome-based billing suits well-defined deliverables with consistent scope. It works when the output is standardised enough to price predictably — research reports, document reviews, data extraction packages. Clients who want cost predictability prefer this model.
Worked Example
A consulting firm offers AI-generated competitive analyses at £75 per report. A client orders 20 reports per month. Monthly revenue: £1,500. Monthly cost: 20 x £16.50 = £330. Monthly margin: £1,170 (78%). As the agent improves and average cost per report drops to £10.00, the margin rises to 87% with no price change.
How Does Hybrid AI Agent Billing Work?
Hybrid billing combines a fixed retainer with variable usage charges. The retainer covers fixed costs and agent availability. The variable component scales with actual usage.
Setting the Structure
The retainer covers: infrastructure costs, base agent capacity, account management, monitoring, and support. Set it high enough to cover fixed costs in a low-usage month.
The variable component covers: per-task charges (each task completed) or per-hour charges (agent time consumed). It aligns costs with actual usage so clients pay proportionally.
Typical Structures
| Component | Low Volume | Mid Volume | High Volume |
|---|---|---|---|
| Monthly retainer | £1,000 | £3,000 | £5,000 |
| Per-task charge | £30-50 | £20-40 | £10-30 |
| Included tasks | 20 | 60 | 150 |
| Overage per task | £40 | £30 | £20 |
Some firms include a block of tasks within the retainer and charge for overages. This gives clients a predictable base cost with flexibility for variable demand.
Pros
Hybrid billing provides predictable base revenue that covers fixed costs regardless of usage. The variable component aligns charges with actual consumption. It is flexible — both firm and client share the risk of demand variability. It works well for ongoing relationships where AI usage fluctuates month to month.
Cons
The model is more complex to explain than pure hourly or per-deliverable pricing. Clients may question the retainer — “Why am I paying £3,000/month when I only used 10 tasks?” The retainer needs clear justification (availability, capacity reservation, infrastructure, support). It also requires solid usage tracking and transparent reporting to maintain trust.
When to Use
Hybrid billing suits ongoing client relationships with variable AI workload. Most professional services firms find this the lowest-risk entry point for AI agent billing because it balances predictability with flexibility. It works particularly well when billing for AI work alongside traditional human services.
Worked Example
A management consulting firm charges a client £3,000/month retainer for AI agent access (covering research, analysis, and drafting agents) plus £30 per completed task. In a busy month, the client uses 80 tasks: £3,000 + (80 x £30) = £5,400. In a quiet month, 15 tasks: £3,000 + (15 x £30) = £3,450. The firm covers fixed costs in both scenarios. The client gets budget predictability with usage flexibility.
How Does Value-Based AI Agent Billing Work?
Value-based billing charges based on the measurable value delivered to the client — cost savings, revenue generated, risk reduced, or time saved.
Setting the Price
Define measurable value metrics with the client before work begins. Common metrics include: cost reduction (agent vs previous manual process), revenue increase (from insights or actions the agent enabled), risk reduction (compliance gaps identified, errors prevented), and time saved (hours freed for the client’s team).
Agree on a pricing formula — typically 10-30% of measured value. The percentage depends on the scale of value and the firm’s negotiating position.
Pros
Value-based billing offers the highest margin potential of any model. If an agent identifies £500,000 in procurement savings, a 15% fee generates £75,000 — far more than any time-based or deliverable-based charge for the same work. It aligns firm and client incentives — both benefit from better outcomes. It justifies premium pricing because the price is anchored to results, not effort.
Cons
Measuring value objectively is hard. Client and firm may disagree on attribution — “Was the saving due to the agent’s analysis or our team’s negotiation?” Payment is delayed until value is realised, which may take months. Disputes on measurement methodology are common. Not all work has easily measurable value outcomes.
When to Use
Value-based billing suits high-value deliverables where the outcome is clearly measurable. Strategic engagements — M&A due diligence, procurement optimisation, tax planning — are natural fits. The client must be open to performance-based pricing and willing to agree on measurement methodology upfront.
Worked Example
An accounting firm’s AI agent analyses a client’s expense data and identifies £200,000 in tax-deductible expenses the client had been missing. The firm charges 15% of the identified savings: £30,000. The agent cost was £45 in compute and tools. Human review cost was £500 (2 hours of partner time). Total cost: £545. Margin: 98.2%.
How Do You Choose the Right Billing Model?
Use this decision framework to match billing models to specific situations.
Decision Framework
Is the deliverable clearly defined?
- Yes → Outcome-based billing is a strong candidate
- No → Move to the next question
Is the value measurable?
- Yes → Consider value-based billing for high-value work
- No → Move to the next question
Is the client relationship ongoing?
- Yes → Hybrid billing provides the best balance
- No → Time-based billing is the simplest option for one-off work
Is the task duration predictable?
- Yes → Time-based billing works well
- No → Outcome-based or hybrid billing reduces risk
Multi-Model Strategy
Most firms will not choose a single model. The realistic approach is:
- Hybrid billing for ongoing client relationships and general AI agent access
- Outcome-based billing for clearly defined, repeatable deliverables
- Value-based billing for strategic, high-impact engagements
- Time-based billing for ad-hoc tasks, internal chargebacks, and clients who require hourly rates
The common foundation across all models is accurate cost tracking. Whether you bill per hour, per deliverable, per retainer, or per value outcome, you need to know what the agent work actually costs. Without that data, margin is a guess. For tracking cost per task, real-time cost monitoring gives firms the foundation for any billing model.
Migration Path
Start where risk is lowest and evolve toward higher-margin models:
- Start with time-based billing — familiar to clients, easy to implement, low risk
- Evolve to hybrid billing — add a retainer component for ongoing clients, improving revenue predictability
- Introduce outcome-based billing — for well-defined deliverables where you have enough cost data to price confidently
- Move to value-based billing — for strategic engagements where outcomes are measurable and the margin opportunity justifies the measurement effort
This migration typically takes 6-12 months as firms build cost data, client trust, and billing infrastructure.
Keito tracks AI agent costs in real time, giving you the data foundation for any billing model — whether you bill by the hour, by the deliverable, or by the value delivered. Build rate cards backed by actual cost data and generate client-ready invoices automatically.
Frequently Asked Questions
What are the main billing models for AI agent work?
The four main models are time-based (charge per agent-hour), outcome-based (charge per deliverable), hybrid (monthly retainer plus variable usage fees), and value-based (charge based on measurable value delivered). Most professional services firms use multiple models across different client relationships and service types.
How does time-based AI agent billing work?
Time-based billing charges clients per hour of agent active time at a published rate. The rate is calculated by taking the agent’s cost per hour (model inference, tools, infrastructure, operational overhead) and applying a 3-5x margin multiplier. Agent hourly rates typically sit at 10-25% of equivalent human rates.
What is outcome-based billing for AI agents?
Outcome-based billing charges a fixed price per deliverable — per report, per analysis, per document reviewed. The price is set by calculating the average cost per deliverable, adding margin, and rounding for client simplicity. Agent efficiency improvements increase the firm’s margin without changing the client price.
How does hybrid AI billing work?
Hybrid billing combines a fixed monthly retainer (covering infrastructure, base capacity, and support) with variable per-task or per-hour charges for actual usage. Some firms include a block of tasks within the retainer and charge for overages. It balances revenue predictability with usage-based charges.
When should you use value-based billing for AI agents?
Use value-based billing when the agent’s output delivers clearly measurable value — cost savings, revenue increases, risk reduction. The firm charges a percentage (typically 10-30%) of the measured value. It suits strategic engagements like procurement optimisation, tax planning, and M&A due diligence where outcomes are quantifiable.
Which AI agent billing model has the highest margin?
Value-based billing has the highest margin potential — often exceeding 90% — because the price is anchored to client value, not agent cost. However, it carries the highest risk (measurement disputes, delayed payment). Outcome-based billing offers the second-highest margin potential with lower risk, particularly as agent efficiency improves over time.