Four Monetisation Models
In Lesson 4, you learned how to choose between Cowork and Frontier based on your organisational context. Now the question shifts from "which platform?" to "how does this agent pay for itself?"
Before you deploy a domain agent, you need to understand how it creates value and how that value is captured. Technology without a value model is a cost centre. It gets funded once, questioned twice, and cut in the next budget cycle. The agents that survive are the ones whose value is visible, measurable, and tied to a model that the organisation understands.
The enterprise agentic landscape has converged on four monetisation models. Each fits a different pattern of value delivery. Choosing the right model is not a minor detail -- it determines whether your agent deployment is seen as an investment or an expense.
Model 1: Success Fee
Deploy the agent. It produces a measurable outcome. Capture a percentage.
The success fee model is the most naturally compelling because value is directly visible: the agent did something, and that something produced a result you can measure. But it requires one critical precondition -- a clean attribution methodology agreed before deployment. Without pre-agreed attribution, you cannot determine which outcomes the agent caused versus outcomes that would have happened regardless.
Natural Domains
| Domain | Typical Fee Structure | Why It Fits |
|---|---|---|
| Sales | $3-8 per qualified lead, 0.5-1.5% of attributed closed revenue | Leads and revenue are directly measurable |
| Finance | 1.5-2.5% of attributed savings identified | Cost reduction is quantifiable against baseline |
| Supply Chain | 0.5-1% of attributed procurement savings | Spend reduction is tracked against purchase history |
The Attribution Requirement
The word "attributed" is doing heavy lifting. Before deploying a success-fee agent, you must agree on:
- What counts as agent-attributed? A lead the agent identified, qualified, and handed to sales? Or any lead the agent touched?
- What is the baseline? What would have happened without the agent? You need a comparison period or control group.
- Who measures? An independent measurement prevents disputes.
Get attribution wrong, and the model collapses into argument. Get it right, and it is the most powerful justification for continued investment.
Model 2: Subscription
Recurring fee regardless of value in a given period. Per-seat, per-team, or enterprise-wide.
The subscription model works when value is continuous but difficult to attribute to specific outcomes. The agent helps every day, but you cannot point to one moment and say "that generated $X."
Natural Domains
| Domain | Why Subscription Fits | Typical Range |
|---|---|---|
| HR | Continuous value across recruiting, onboarding, policy questions -- hard to tie to specific revenue | Team-level: $800-$2,500/month |
| Product Management | Diffuse, ongoing value across discovery, planning, and stakeholder communication -- hard to tie to specific revenue | Team-level: $2,000-$8,000/month |
| Operations | Ongoing process documentation and compliance value that prevents failures but cannot be attributed to specific savings | Department: $1,500-$5,000/month |
The Self-Justification Problem
Subscription's weakness is that it does not self-justify. A success-fee agent proves its value every time it generates a fee. A subscription agent requires active measurement to demonstrate that the recurring cost is worth paying.
Without deliberate value tracking, subscriptions become line items that finance questions during budget reviews. The fix: build measurement into the deployment from day one. Track time saved, errors prevented, queries handled -- whatever makes the value visible even when it cannot be attributed to specific revenue.
Model 3: License
High-stakes, regulated, or proprietary domains where security and compliance reviews are expected and the pricing reflects the risk profile.
License agreements are annual contracts with significant upfront negotiation. They fit domains where the consequences of agent failure are severe enough that both parties need contractual protections.
Natural Domains
| Domain | Typical Annual Range | Why License Fits |
|---|---|---|
| Legal | $40,000-$150,000/year | Regulatory compliance, attorney-client privilege, malpractice risk |
| Banking / Finance | $60,000-$200,000/year | IFRS 9, Basel III/IV compliance, AML/KYC regulatory exposure |
| CA/CPA Practice | $40,000-$120,000/year | Audit standards, tax compliance, professional liability |
Requirements
Deploying under a license model means passing through:
- Security review: How is data stored, transmitted, and accessed?
- Legal review: Who is liable if the agent produces incorrect output?
- Compliance assessment: Does the agent meet regulatory requirements specific to the domain?
This procurement process takes months, not weeks. It is appropriate for Level 3+ maturity organisations (you will learn about maturity levels in Lesson 6) that have the governance infrastructure to manage these reviews.
Model 4: Marketplace
Publish your SKILL.md as a reusable plugin. Other teams or organisations subscribe.
The marketplace model turns your domain expertise into a product. You write a SKILL.md that encodes general best practices in your domain -- not your organisation's proprietary knowledge, but the knowledge that any practitioner in your field would benefit from.
Economics
- Revenue per subscriber: $200-$900/month
- Marginal cost of each additional subscriber: effectively zero
- IP distinction: Organisation-specific knowledge = not publishable. General domain best practice = publishable.
The IP Distinction
This is the critical boundary. Your company's internal compliance procedures, client lists, and proprietary methods are not marketplace material. But your knowledge of how to structure a regulatory review, how to approach building code analysis, or how to qualify a sales lead in your industry -- that general domain expertise is publishable and valuable.
Comparison Table
| Model | Value Pattern | When to Use | Key Risk |
|---|---|---|---|
| Success Fee | Measurable, attributable outcomes | Sales, finance, cost reduction | Attribution disputes |
| Subscription | Continuous, diffuse value | HR, product management, operations | Fails to self-justify |
| License | High-stakes, regulated domains | Legal, banking, CA/CPA practice | Lengthy procurement |
| Marketplace | Reusable domain expertise | General best practices | IP boundary confusion |
Try With AI
Use these prompts in Anthropic Cowork or your preferred AI assistant to explore these concepts further.
Prompt 1: Personal Application
I work in [YOUR DOMAIN -- e.g., "financial compliance at a regional
bank"]. If I deployed a domain agent to help with [describe a specific
task -- e.g., "reviewing loan applications against our risk criteria"],
which of the four monetisation models would best capture the value?
Walk me through each model and explain why it does or does not fit my
situation. Then recommend the best model and explain what I would need
to set up before deployment (e.g., attribution methodology, measurement
framework, compliance review).
What you're learning: You are practising model selection against your own domain. The AI forces you to evaluate each model against your specific value delivery pattern, not just pick the first one that sounds reasonable.
Prompt 2: Framework Analysis
Analyse this scenario: A 40-person architecture firm wants to deploy
an AI agent that reviews building plans against local building codes
and flags potential violations before submission. The firm charges
clients per project, and a code violation caught early saves an
average of $15,000 in rework costs.
Which monetisation model fits best? Could a hybrid model work (e.g.,
subscription base + success fee per violation caught)? What are the
trade-offs of each approach?
What you're learning: You are evaluating whether a single model fits or whether a hybrid approach is needed. Real deployments often require blending models, and this prompt teaches you to think about trade-offs rather than defaulting to one answer.
Prompt 3: Domain Research
Research how AI agents are currently monetised in [YOUR INDUSTRY --
e.g., "legal technology," "healthcare IT," "sales enablement"]. What
pricing models are the leading vendors using? Are they charging per
seat, per outcome, per license, or through marketplaces? How do the
prices compare to the benchmarks I learned (e.g., $3-8 per qualified
lead for sales, $40,000-$150,000/year for legal licenses)?
What you're learning: You are grounding the abstract models in current market reality. Knowing what competitors charge and how they structure pricing gives you a reference point for your own deployment decisions.