Your prices live in a document
The page links to a PDF or a rate card. An agent that cannot parse it reports that pricing is unavailable and moves on.
You get compared on the assumption that you are expensive.Agent readiness
Build an agent-ready business
AI agents now do the finding, the comparing and the shortlisting that buyers used to do themselves. A business an agent cannot read is left out of the answer, and nothing in your analytics records the loss. Agent readiness is the work of being findable, understandable and callable by the software that shops on your buyer's behalf.
The invisible loss
Six things that break when an agent tries to buy from you
An agent evaluated four suppliers, one of them returned nothing it could use, and that one was never mentioned to the buyer. There is no bounce rate for this. No failed enquiry, no abandoned form, no signal at all.
The page and the structured data disagree
A price changed on the page and the machine-readable copy behind it kept the old number, or the other way round.
An agent finding both picks one, and you do not know which.Every way in is a form
Contact, quote and booking are all built for a person with a mouse. Nothing returns an answer to a caller that is not a browser.
An agent can read about you and can do nothing with you.Your crawler policy is whatever came with the site
The file was written for search engines and says nothing about AI use, training or grounding.
You are permitting everything or refusing everything, and you chose neither.Someone already connected an assistant to your systems
It saved them an hour a week and it was never registered, scoped or reviewed.
You cannot name which agents reach your data, or switch one off.Assistants describe you from whatever they found
Old pages, a stale directory listing, a competitor comparison, a job ad from four years ago.
The first pitch a buyer hears about you is one you did not write.The five domains
Discover, Understand, Access, Act, Govern
Five things an agent tries to do, in the order it tries them. Most businesses are strong on the first, thin on the second, and have never been asked the last three. Each domain is scored 1 to 5, for 25 in total.
- Discover
Discovery
Whether an AI system can find your business and identify it as one company rather than four half-matching records.
- A crawler policy that states your position on AI use
- An index of the site written to be read by a model
- One canonical identity across every property you own
- Understand
Data
Whether an agent can work out what you sell, at what price, on what terms, and be right about it.
- Services, prices and eligibility published and current
- Structured data that agrees with the page it sits on
- One source of truth per fact, with a named owner
- Access
Interfaces
Whether an agent can call something and get a useful answer, instead of reading prose and guessing.
- At least one endpoint that returns business data
- A published specification that matches actual behaviour
- Authentication, rate limits, logging and error handling
- Act
Transactions
Whether an agent can get work quoted, checked, booked or started without a person in the middle of every step.
- Quote requests and eligibility checks it can call
- Booking or order initiation that returns a confirmation
- A defined handoff to a person, and the conditions that trigger it
- Govern
Governance
Whether the access you have opened is identified, bounded, recorded and reversible. This is the domain that turns the rest from a marketing exercise into a board conversation.
- An identity for every agent that reaches you
- Approval thresholds and spend limits that are enforced
- An audit log, and a revocation path that has been tested
How each domain is scored
One ladder, five rungs, used on every domain
The same five levels apply everywhere, so a score means the same thing in Governance as it does in Discovery, and a re-score next year measures what changed rather than who assessed it.
-
Absent
Nothing exists in this domain.
-
Incidental
Something exists as a side effect of another decision. No owner, no date, no intent behind it.
-
Declared
Published deliberately. Current, dated, and one named person answers for it.
-
Structured
Machine-readable and specified. A third party can parse and verify it without asking you anything.
-
Operated
Monitored and versioned. Drift, staleness and breakage raise an alert to the person who owns it.
Every score cites evidence that someone outside your organization can check. Four of the five domains are scoreable from the public internet with no access and no permission, which means we can tell you where you stand before you have engaged us, and it means a competitor can do the same to you.
What this is
Three things agent readiness gets confused with
Each of these overlaps with a piece of the work and stops well short of it. The distinction matters, because buying the wrong one leaves the expensive domains untouched.
"This is AI SEO"
Being cited in an AI answer is Discovery, which is one domain of five. The work that changes a buying decision sits in Data, Interfaces and Governance, and a search agency sells none of the three.
"This is a chatbot on our website"
An assistant on your site serves people who already arrived. Agent readiness serves the agents that never load the page, on behalf of buyers you never hear from.
"This is an MCP project"
MCP is the dominant interface standard today and it will be replaced by something. The work is scoped around the interface requirement, on the same reasoning that keeps the transaction work independent of any one payment protocol.
Our own work
We did this to ourselves first
Begine Fusion publishes a crawler policy that states our position on search, AI grounding and AI training as three separate answers rather than one blanket yes or no. We publish a hand-written index of this site for models to read, a generated file containing every indexable page as plain text, and a markdown version of every page at its own address. Structured service and pricing data sits behind all eleven offer pages.
We also ran the scoring on ourselves before running it on anybody else, and published the result. Fourteen out of twenty-five on the first run, fifteen once the catalogue endpoint shipped, with the evidence for every point linked and the domains we score poorly on named. Read our own score, or check the evidence yourself in a single click: our crawler policy, our index for models, or the catalogue an agent can call.
Where this is delivered
Enter at the point that matches where you are
Agent readiness is delivered through the engagements we already run, at the prices those engagements already publish. Find the line that describes your organization. Any of them can be first.
Fit
Whether this is the conversation to have now
This page is for organizations whose buyers research before they make contact. If the question you are holding is a different one, the right column says where it is answered.
Start here if
- You sell something a buyer researches before they contact anyone
- You have wondered what AI assistants say about your company, and never checked
- Your staff have connected assistants to systems that hold client data
- A customer or partner has asked how to reach your data programmatically
- You are rebuilding the website anyway, so doing it right costs very little more
Start somewhere else if
- The question is how your own people use AI internally. Go to AI-Enabled Workforce.
- You need the rules for AI use before anything else. Go to FusionGuard.
- You want a free read on your internal AI readiness first. Take the AI readiness assessment.
- You have no public digital surface and no plan to build one. This can wait.
Find out what an agent sees when it looks for you
Bring your website and whatever agent access your team has already granted. We will score the five domains against evidence, show you the point where a machine stops, and tell you which engagement closes the gap.