Pricing in the agent economy
See also How to price AI agents (published April 28, 2025)
The move towards the agent economy is gathering momentum. Look at virtually any major B2B software website, and there is an Agent or Agentic AI story emerging.
In recent conversations, we have been privy to venture capital board members who have been pushing companies to adopt an agent-first strategy and to conceal other functionality behind agents. See Bessemer Venture Partners on Business Model Invention in the AI Era and Dan Bartus and Mishca Vaughn The Agent Economy.
This slide from Thoma Bravo summarizes the emerging consensus.
Agents are becoming a major design pattern in their own right, reframing human-computer interactions and, in some cases, replacing them with agent-to-agent interactions. See Jakob Nielsen’s Hello AI Agents: Goodbye UI Design, RIP Accessibility.
So how might agents and the agent economy get priced, and how might that impact the pricing of other B2B SaaS applications?
An agent is a piece of software that takes an action on behalf of a user.
An AI agent is an agent that leverages AI in order to decide what action to take and/or to execute the action.
The actions can be combined in various ways, and agents are often developed as families of agents or, if they interact and exchange data, as systems of agents.
One of the best descriptions of agents comes from the AI agent platform vendor Vellum in The Six Levels of Agent Behavior.
Caller (Basic Responder)
Actor (Uses Tools)
Operator (Observe, Plan, Act)
Explorer (Fully Autonomous)
Inventor (Fully Creative)
Today, most agents are Callers or Actors, but over time, more and more agents will climb this ladder.
Here are some examples. (List generated using Perplexity; see the full thread with examples and sources.)
Sales and Marketing AI Agents
1. AI Sales Development Representatives (SDRs)
2. Dynamic Pricing AI Agents
3. Product Recommendation Engines
Customer Experience AI Agents
4. Autonomous Customer Support Agents
5. Visual Product Search Agents
Financial Services AI Agents
6. Loan Processing and Business Development Agents
7. Economic Value Modeling Agents
8. Contract Negotiation Agents
Healthcare AI Agents
9. Medical Diagnostic Agents
10. Medical Imaging Analysis Agents
Operational AI Agents
11. Autonomous Vehicle Navigation Agents
12. E-commerce Operation Agents
13. Sales Forecasting Agents
Emerging AI Agent Applications
14. Competitive Intelligence Agents
15. Regulatory Compliance Agents
Each of these agents performs different jobs, creates value in different ways, and needs its own approach to pricing.
If only it were that simple. Most important business processes will be handled by teams of agents. A simple example from the world Ibbaka works in is sketched below.
How will this buzzing confusion of agents be priced?
One of the best approaches we have seen is Gary Bailey’s Jobs to be Done approach. Most people know Clayton Christensen’s Jobs to be Done framework. You can follow his ideas in these three LinkedIn posts.
Gary Bailey Part 1: AI Agents Jobs-Led-Pricing
Gary Bailey Part 2: Ai Agents: From Blob Pricing to Job Pricing
Gary Bailey Part 3: AI Agents Repackage Features into Jobs
Core Idea: The Jobs to Be Done (JTBD) framework, developed by Clayton Christensen, focuses on understanding why customers "hire" products or services to accomplish specific functional, emotional, or social tasks (the "job").
Circumstances Over Demographics: It prioritizes the context and motivations driving customer decisions rather than relying on traditional demographic-based segmentation.
Solution-Agnostic: JTBD emphasizes solving the customer's core problem or need, regardless of the specific product features or tools used.
Customer-Centric Innovation: By aligning offerings with the "job" customers want done, businesses can create more effective, differentiated, and enduring solutions.
Practical Example: For instance, customers might buy a milkshake not for its taste but to make a long commute more enjoyable—a deeper insight that guides product design and marketing.
Application to Pricing: Define pricing metric in terms of the value of the job to be done and the value of completing that job (see the Ibbaka Agent AI Pricing Layer Cake below)
How does one apply this to pricing?
An agent is designed to do a job for you. There are many different jobs that agents can do, and many more to come. Agents are already being used for a vast array of business activities.
Here are some examples of what a Jobs-to-be-Done approach to pricing agents could look like.
The increasing complexity of AI agents in today's market requires strategic and well-structured pricing approaches. The Ibbaka Agentic AI Pricing Layer Cake framework provides a comprehensive method for organizing pricing metrics into four distinct layers, enabling businesses to develop pricing models that align with value creation, maintain predictability, and scale effectively with costs.
Role (the Job to be Done or the type of agent)
Access (a retainer to assure access or manage expense)
Usage (how often the agent is used)
Outcomes or Performance (the value the agent is meant to deliver)
The Agentic AI Pricing Layer Cake is a structured framework that organizes pricing metrics for AI agents into four strategic layers. This layered approach simplifies the pricing design process by categorizing metrics according to their relationship to value delivery and customer engagement. By understanding these layers, organizations can create pricing models that balance complexity with effectiveness.
The Role layer focuses on the fundamental purpose of the AI agent—what job it performs or what type of agent it is. This layer addresses the core functionality that solves specific customer problems.
Role-based pricing metrics might include:
Agent type categories (research assistant, customer service agent, coding assistant)
Specialization level (basic, advanced, expert)
Industry-specific applications (healthcare, finance, education)
Capability tiers (standard features, premium capabilities)
Role metrics establish the baseline value proposition and often form the foundation of tiered pricing structures where different agent types or capability levels command different price points.
The Access layer covers metrics related to ensuring the customer can reliably use the agent when needed. This often takes the form of a retainer or subscription model that guarantees service availability or helps manage expenses.
Access-based pricing metrics typically include:
Subscription periods (monthly, annual)
Service level agreements (standard, premium, enterprise)
Number of authorized agents or agent actions
Priority access guarantees
Reserved capacity or computing resources
Access metrics provide predictable revenue streams for providers while giving customers assurance of service availability within defined parameters.
The Usage layer tracks how frequently and extensively the agent is utilized. These metrics capture the volume and intensity of customer engagement with the AI agent.
Common usage-based pricing metrics include:
Number of queries or interactions
Processing time or compute resources consumed
Data volume processed
Number of tasks completed
Feature utilization rates
Session duration or frequency
Number of tokens consumed (input, output, reasoning)
Usage metrics allow pricing to scale with actual consumption, aligning costs with benefits received while providing flexibility for customers with varying needs.
The Outcomes or Performance layer focuses on the actual value delivered by the agent. This represents the most direct connection between pricing and customer value but can also be the most challenging to measure.
Outcome-based pricing metrics might include:
Revenue generated or influenced
Cost savings achieved
Productivity improvements
Success rates or quality metrics
Time saved
Customer satisfaction scores
Business objectives achieved
Outcome metrics create the strongest alignment between pricing and value, though they require robust tracking mechanisms and a clear definition of success measures.
Effective pricing design begins with a clear understanding of value creation and strategic pricing objectives. The layer cake framework provides a systematic approach to translating value understanding into practical pricing structures.
See Core Concepts: Value Model.
The foundation of agent pricing is developing a comprehensive value model. A value model identifies and quantifies the specific benefits an AI agent delivers to customers. Each value model consists of value drivers expressed as equations that estimate specific types of value.
For example, a customer service AI agent might have value drivers such as:
Call center labor cost reduction = (Average call handling time × Agent labor cost per hour × Number of calls) × Percentage reduction achieved by AI
Customer satisfaction improvement = Increase in satisfaction score × Value per point of satisfaction
These value driver equations contain variables that can be mapped to specific pricing metrics within the layer cake framework.
Once value drivers are defined, the next step is mapping their variables to the four layers of pricing metrics:
Role variables might include agent capabilities or specialization level
Access variables could involve the number of users or service level requirements
Usage variables typically include interaction volume or resource consumption
Outcome variables directly measure value creation, such as cost savings or revenue increases
This mapping process creates a direct connection between the value delivered and how it's reflected in pricing.
The most effective pricing models typically incorporate metrics from one or two layers of the cake. Using metrics from three or four layers creates excessive complexity that can confuse customers and complicate sales processes.
The selection of which layers to use depends on strategic priorities:
Startups might favor Role and Access metrics for simplicity and predictability
Enterprise solutions might combine Access with Outcomes for alignment with business value
Consumer applications might focus on Role and Usage for transparency and scalability
The goal is to balance simplicity with value alignment, choosing metrics that create the right incentives while remaining understandable to customers.
Effective AI agent pricing should satisfy several key design objectives that balance provider and customer needs.
When pricing tracks value, customers perceive fairness because they pay in proportion to benefits received. This alignment encourages adoption and reduces churn. Metrics from the Outcomes layer create the strongest value alignment, though they may introduce variability.
To strengthen the connection between pricing and value:
Regularly validate pricing metrics against actual value delivered
Gather customer feedback on perceived value versus price
Adjust pricing tiers to reflect value differences between customer segments
Predictable pricing helps customers budget appropriately and reduces friction in the buying process. Access and Role metrics typically provide the greatest predictability, while Usage and especially Outcomes metrics may introduce variability.
Techniques to enhance predictability include:
Offering fixed-price tiers with clear feature sets
Providing usage estimates or calculators
Setting usage caps or thresholds
Creating hybrid models that combine predictable base rates with variable components
The ideal pricing model scales revenue faster than underlying costs as adoption increases. This creates expanding margins that support ongoing investment in the agent's capabilities.
To achieve favorable scaling:
Identify which costs scale linearly with usage and which are more fixed
Design pricing that captures a portion of increasing value at scale
Consider volume-based pricing tiers that reward but also monetize increased usage
Focus on metrics that track value creation rather than just resource consumption
The fundamental pricing equation states that Value > Price > Cost must be maintained across all scales of operation. This ensures that customers receive net value, providers generate profit, and both parties benefit from the relationship.
To preserve this critical relationship:
Regularly reassess the value delivered as agents evolve
Monitor cost structures, especially as scale increases
Adjust pricing strategies to maintain appropriate margins
Segment customers based on their value profiles
Consider different pricing models for different customer segments
As AI agent technology evolves rapidly, so too will pricing approaches. The layer cake framework provides adaptability for this changing landscape while maintaining consistent principles.
In the near future, specialized AI agents will assist in pricing other AI agents. These pricing agents will analyze context about agent capabilities, value delivery patterns, and target customer characteristics to generate optimized pricing models.
These pricing agents will:
Process vast amounts of usage and outcome data
Dynamically adjust pricing based on value patterns
Identify optimal pricing metrics across the four layers
Suggest pricing experiments to optimize revenue and adoption
Balance the four design objectives automatically
As measurement capabilities improve, we can expect a gradual shift toward more outcome-based pricing models that directly tie costs to value creation. This evolution will require:
Advanced analytics to attribute outcomes to AI agent actions
Greater transparency in value measurement
New contractual frameworks for sharing value
Industry standards for measuring AI agent effectiveness
The Agentic AI Pricing Layer Cake provides a structured approach to designing pricing for AI agents that balances complexity with effectiveness. By categorizing pricing metrics into Role, Access, Usage, and Outcomes layers, organizations can select the right combination of metrics to achieve their strategic objectives.
Effective pricing design begins with understanding value and mapping it to appropriate metrics. By thoughtfully selecting metrics from the layer cake framework, organizations can create pricing that tracks value, maintains predictability, scales efficiently, and preserves the fundamental Value > Price > Cost relationship.
As AI agent technology continues to evolve, this framework offers both structure and flexibility to adapt pricing approaches while maintaining a focus on value delivery. The future of AI agent pricing will likely see increasing sophistication, with automated pricing optimization becoming a reality sooner than many anticipate.