Agent Readiness Levels Explained: From Dark to Interoperable
Not every business needs to be fully autonomous on day one. The Agent Readiness Level (ARL) framework gives you a clear, cumulative progression from completely invisible to AI agents (ARL-0) to running your own agent that negotiates with customer agents (ARL-6).
The 7 Levels at a Glance
Why ARL-3 Is the Revenue Inflection Point
The jump from ARL-2 to ARL-3 is where everything changes. At ARL-2, agents can read about your business. At ARL-3, agents can transact with your business. This is the difference between being a listing and being a revenue channel.
Our data shows that businesses at ARL-3 and above receive 4.7x more agent interactionsthan businesses at ARL-2, because agents prioritize businesses they can actually do something with. An agent asked “book me a dentist” will skip the ARL-2 dentist (can only read about) and book with the ARL-3 dentist (can actually schedule).
The business case: Getting from ARL-0 to ARL-3 with AgentHermes takes 60 seconds and costs $29/month. The ROI is the first agent-booked appointment your competitor does not get.
Detailed Level Breakdown
ARL-0: Dark
Invisible to AI agents
The business exists but has zero machine-readable presence. No structured data, no API, no agent card, no llms.txt. When an AI agent searches for this type of business, it will never find this one. The business is a ghost in the agent economy — it might as well not exist.
Real Examples
Phone number on a static HTML page. Menu is a JPG. Hours are on Facebook.
No website, no Google Business Profile, walk-in only. Completely invisible to any digital system.
LinkedIn profile and a PDF portfolio. No structured data, no way for an agent to understand services or pricing.
To reach ARL-1: Add Schema.org markup, create a Google Business Profile, or publish any machine-readable description of what the business does.
ARL-1: Discoverable
Agents can FIND the business
An AI agent can discover that the business exists and understand what it does at a high level. The business has some form of structured data — Schema.org markup, a Google Business Profile with structured hours and categories, or an agent-card.json file. An agent asked "find a dentist near me" could include this business in results.
Real Examples
Hours, address, phone, category, and photos are structured. An agent can recommend it but cannot book or order.
Docs describe capabilities. An agent can understand what the product does, but cannot sign up or use it.
Requirements
- Discovery score (D1) at 40+
- Total weighted score at 20+
- At least one structured data source (Schema.org, agent card, llms.txt)
To reach ARL-2: Publish structured service/product catalogs. Make pricing, availability, and detailed descriptions machine-readable instead of embedded in HTML paragraphs.
ARL-2: Readable
Agents can UNDERSTAND offerings
An agent can read the business's offerings, pricing, and availability in structured formats without scraping HTML. Product catalogs, service lists, and pricing are available as JSON, XML feeds, or well-structured Schema.org markup. An agent can now comparison shop — it knows what this business offers, at what price, and whether items are available.
Real Examples
Products have Schema.org Product markup with price, availability, and reviews. Agents can compare across stores.
Services, pricing ranges, and accepted insurance are in structured data. An agent can match patients to providers.
Requirements
- All ARL-1 requirements met
- Data Quality (D6) or API Quality (D2) at 50+
- Total weighted score at 35+
- Structured product/service catalog accessible to machines
To reach ARL-3: Add booking, ordering, or signup capabilities. Let agents DO something, not just READ about you.
ARL-3: Bookable
Agents can START a transaction
This is the revenue inflection point. An agent can now initiate a transaction — book an appointment, place an order, request a quote, or create an account. The business has some form of API or booking system that accepts programmatic input. The jump from "agents can read about you" to "agents can book you" is where real value begins to flow.
Real Examples
An agent can place a takeout order with customizations, paying through the existing checkout flow.
An agent can create a trial account, get API keys, and start testing the product — all programmatically.
Scheduling API accepts service type, preferred date, and address. Agent can book an appointment.
Requirements
- All ARL-2 requirements met
- Onboarding (D3) at 40+ — programmatic booking or signup exists
- API Quality (D2) at 50+ — callable API endpoints
- Total weighted score at 50+
To reach ARL-4: Add payment processing, order tracking, and modification capabilities. Let agents COMPLETE the transaction, not just start it.
ARL-4: Transactable
Agents can COMPLETE a transaction
The full transaction cycle is agent-accessible: create, pay, track, modify, and cancel. An agent can handle the entire customer journey without human intervention at any step. Payment is programmatic (Stripe, Square, or similar). Order status is queryable. Modifications and cancellations have API endpoints.
Real Examples
Agents can create payment intents, manage subscriptions, issue refunds, and track payment status through a comprehensive API.
Cart, checkout, payment, order tracking, and returns are all API-accessible. Agents handle the full purchase flow.
Requirements
- All ARL-3 requirements met
- Payment (D5) at 50+ — programmatic payment capability
- Security (D7) at 50+ — proper auth, TLS, rate limiting
- Reliability (D8) at 50+ — consistent uptime, fast responses
- Total weighted score at 60+
To reach ARL-5: Enable autonomous relationship management: reordering, plan optimization, proactive notifications, and agent-driven account management.
ARL-5: Autonomous
Agents can MANAGE the relationship
An agent can manage the ongoing business relationship without human intervention. This goes beyond single transactions to lifecycle management: reordering when supplies run low, optimizing subscription plans based on usage, escalating issues through support channels, and negotiating renewals. The business has essentially delegated customer relationship management to the agent layer.
Real Examples
Agents can provision resources, optimize costs, set alerts, auto-scale, and manage billing across the entire lifecycle.
Agents handle reorders, subscription modifications, loyalty points, and proactive recommendations based on purchase history.
Requirements
- All ARL-4 requirements met
- Onboarding (D3) at 60+ — full self-serve lifecycle management
- Payment (D5) at 60+ — subscription billing, refunds, modifications
- Agent Experience (D9) at 60+ — SDKs, error handling, tracing
- Total weighted score at 70+
To reach ARL-6: Publish A2A agent card and MCP server. Let your business operate its own agent that communicates with customer agents directly.
ARL-6: Interoperable
Agent-to-agent communication
The business operates its own AI agent that communicates with customer agents via standardized protocols (A2A, MCP). This is the fully realized agent economy: a customer's agent contacts the business's agent, they negotiate terms, and the transaction happens without any human on either side. The business's agent can handle complex negotiations, multi-party coordination, and dynamic pricing.
Real Examples
MCP server with database tools, A2A capabilities, comprehensive API, and agent-native documentation. Close to full interoperability.
A restaurant whose agent negotiates group booking rates with a corporate travel agent, coordinates with delivery agents, and adjusts pricing in real-time.
Requirements
- All ARL-5 requirements met
- Published A2A agent card at /.well-known/agent-card.json
- MCP server exposing business capabilities as tools
- Total weighted score at 75+
Maximum level reached. You have reached the highest level. Continue improving individual dimensions and expanding agent capabilities.
How to Check Your ARL Level
Enter your URL
Go to agenthermes.ai/audit and paste your business website.
Run the scan
We probe 9 dimensions across the full 6-step agent journey in seconds.
See your ARL
Your score page shows your ARL level, overall score, and dimension breakdown.
Level up
Follow dimension-specific recommendations to advance to the next ARL.
The scan is free and takes about 10 seconds. You get a permanent score page at agenthermes.ai/score/yourdomain.com that shows your ARL level alongside your full 9-dimension breakdown. Share it with your team to align on agent readiness priorities.
The ARL level is cumulative — you cannot be ARL-4 without meeting all requirements for ARL-0 through ARL-3. This prevents gaming: a business with great payment capabilities but no discovery is not truly agent-ready. Every level must be earned in order.
Where Do Businesses Fall Today?
After scanning 132+ businesses across 15 verticals, here is the distribution of Agent Readiness Levels in our database:
Key finding:63% of businesses we scan are at ARL-0 or ARL-1 — meaning agents can barely find them, let alone transact with them. Only 9% have reached ARL-3 or above, where agent-driven revenue starts to flow. The gap between “agents know you exist” and “agents can book you” is where most businesses stall.
What is your Agent Readiness Level?
Scan your business for free and see your ARL level, dimension scores, and exactly what you need to reach the next level.