Machine-Readable Market Access. When Legibility to AI Agents Becomes a Condition of Trade

Synthocracy Institute Working Paper
Martin Novak
July 2026

Abstract

A business may be legally registered, economically productive, technically competent, and commercially competitive, yet remain operationally absent from an AI-mediated market.

This can happen without a formal ban. The company may simply lack the structured catalogue, standard identifiers, product classification, current availability data, digitally verifiable credentials, supported units of measure, payment information, software interfaces, agent profile, or executable checkout path required by an automated purchasing system.

Human buyers may still be able to find the company’s website, read a PDF catalogue, interpret an email, call a salesperson, understand an unusual product, request clarification, and negotiate an exception. An AI agent may not. It must be able to identify, parse, authenticate, compare, qualify, route, and act.

This paper introduces machine-readable market access as a distinct category of economic participation. It also develops the concept of executable visibility: visibility that does not end when a firm or product is found, but continues until an agent can verify the participant, interpret the offer, compare it with alternatives, establish authority, and complete or initiate a transaction.

The transition is already visible in emerging commerce and procurement infrastructures. Google’s Universal Commerce Protocol integration requires merchants to provide structured product and account data, returns and customer-support information, declared capabilities, technical endpoints, and programmatically accessible checkout processes. The Agent2Agent protocol uses Agent Cards to describe an agent’s identity, skills, interfaces, capabilities, and authentication requirements. Peppol catalogue specifications similarly depend on structured identifiers, coded units, product references, and machine-processable business documents. These systems do not yet constitute one universal agentic market, but they reveal a broader direction: commercial participation is becoming progressively dependent on whether a business can be interpreted and executed by machines. (Google for Developers)

The central question is therefore:

Does a business have meaningful market access if it can be found by humans but cannot be interpreted or executed by agents?

Keywords

Machine-readable market access; executable visibility; market synthocracy; AI agents; agentic commerce; agentic procurement; structured product data; commercial admissibility; Agent Cards; UCP; A2A; Peppol; SME access; digital identity.

Claim Status and Scope

This paper distinguishes three kinds of claims.

(A) Empirical claims describe observable standards, systems, documentation, or market developments.

(B) Normative and argumentative claims interpret those developments and propose categories for understanding them.

(C) Foresight claims describe plausible future developments rather than established outcomes.

The paper does not claim that every business must immediately deploy its own autonomous agent, that human commerce is about to disappear, or that machine-readable standards are inherently exclusionary. Structured data, interoperability, digital credentials, and automated transactions can reduce costs and expand access.

The argument is narrower and more consequential:

As AI agents become intermediaries between market participants, machine legibility can shift from a source of efficiency to a practical condition of participation.

The Synthocracy Institute’s wider research programme studies precisely this movement of power into upstream layers: the systems that determine who becomes visible, prioritised, trusted, routed, and capable of acting before a final human decision appears.

1. The Business That Exists but Does Not Appear

Imagine a medium-sized manufacturer with twenty years of experience.

It produces a reliable industrial component. Its prices are competitive. Its quality record is good. It can customise dimensions, offer short production runs, and ship across Europe. Its employees understand the product deeply. Long-term customers trust the company.

The business has a website. It also has a PDF catalogue, a spreadsheet with technical specifications, several declarations of conformity, test reports stored in email attachments, and a salesperson who can answer questions.

To a human buyer, the company exists.

A purchasing manager can search for the product, open the PDF, notice that one dimension is described differently from the internal specification, call the supplier, ask whether a similar version is available, clarify whether a price refers to one item or one package, and request an updated certificate.

A capable human can interpret ambiguity.

Now imagine that the first purchasing decision is made by an agent.

The agent is instructed to identify suppliers that can deliver 20,000 units within fourteen days, meet a specific technical standard, provide a defined certificate, accept sixty-day payment, quote in euros per thousand units, and connect to the buyer’s automated RFQ workflow.

The manufacturer’s website mentions the correct product, but not in the expected classification. The PDF uses millimetres in one place and centimetres in another. Availability is not published. The certificate is a scanned image rather than a digitally verifiable record. The catalogue price refers to a box, but the box quantity is not represented in a structured field. The company has no API, no supported agent interface, and no standard mechanism for declaring commercial capabilities.

The agent may not conclude that the supplier is bad.

It may never reach a conclusion at all.

The company can disappear at several points:

during discovery because its product is not classified correctly;
during interpretation because the system cannot map its terminology;
during qualification because its certificate cannot be verified;
during comparison because units are inconsistent;
during availability checking because inventory data is missing;
during commercial validation because payment conditions are not machine-readable;
during execution because the supplier cannot receive or respond to an automated RFQ;
during checkout because no supported transactional interface exists.

No formal rejection is required.

No platform must issue a ban.

No buyer needs to decide against the supplier.

The business may simply fail to become an actionable option.

This is the first principle of machine-readable market access:

A firm can be economically real and commercially absent at the same time.

2. From Visibility to Executable Visibility

The digital economy has already created several forms of visibility.

A company may be visible in a search engine. Its content may appear in a generated answer. Its products may be listed on a marketplace. Its brand may be recognised by a language model.

None of these guarantees that an agent can transact with it.

The development can be represented as a sequence.

2.1 Search visibility

The business can be found through a search query.

This is the traditional domain of search-engine optimisation. The objective is to ensure that a human searching for a product, service, or supplier can reach the relevant page.

Search visibility asks:

Can the market participant be found?

2.2 Answer visibility

The company or product appears in an AI-generated answer, comparison, summary, or recommendation.

This requires not only being indexed but being sufficiently understandable, relevant, and credible for a model to use the information.

Answer visibility asks:

Can the participant be represented in a machine-generated account of the market?

2.3 Agent legibility

An agent can identify the firm, interpret the offer, and understand the relevant product or service attributes.

It can determine, for example:

what is being sold;
in which configuration;
under which identifier;
in what unit;
at what price;
with which certifications;
in which territory;
under which delivery constraints.

Agent legibility asks:

Can the participant be translated into a structured commercial object?

2.4 Agent comparability

The system can compare the firm’s offer with alternatives without relying on extensive human interpretation.

This requires semantic compatibility. Two offers must be recognised as comparable even when their internal naming conventions differ.

Comparability asks:

Can the agent place this offer inside a valid decision set?

2.5 Agent qualification

The agent can verify that the company and offer satisfy the buyer’s legal, technical, financial, security, and operational requirements.

Qualification asks:

Can the agent establish that the participant is admissible?

2.6 Transaction readiness

The agent can initiate a commercially meaningful process: request a quotation, reserve stock, create a basket, propose terms, or prepare an order.

Transaction readiness asks:

Can the participant move from information into workflow?

2.7 Executable visibility

The participant is not only findable, understandable, and eligible. The agent can take or initiate an authorised action involving that participant.

This may include:

creating an RFQ;
obtaining a current quotation;
negotiating within a mandate;
confirming availability;
creating a checkout session;
placing an order;
initiating payment;
booking logistics;
recording the transaction.

Executable visibility can therefore be defined as:

The condition in which a market participant is sufficiently identifiable, interpretable, verifiable, comparable, authorised, and technically connected for an AI agent to include it in an executable commercial process.

The distinction matters because the future market may contain many businesses that are visible but not executable.

A supplier can appear in search results yet remain absent from procurement.

A merchant can be mentioned in an answer yet remain unavailable for agentic checkout.

A manufacturer can be listed in a directory yet fail automated qualification.

A company can therefore possess informational visibility without operational presence.

The project’s earlier framework identified this emerging hierarchy: visible and executable firms, visible but non-transactional firms, gateway-dependent firms, manual-exception suppliers, machine-invisible firms, and formally excluded participants.

3. Why Machine Legibility Is Becoming More Important

Machine-readable commerce is not entirely new.

Electronic data interchange, product identifiers, procurement standards, structured catalogues, APIs, and automated ordering have existed for decades. Large enterprises already exchange standardised orders, invoices, dispatch notices, and catalogue data.

What changes with AI agents is the breadth of interpretation and action being brought into the same environment.

Traditional electronic commerce often required two organisations to agree in advance on:

document formats;
system mappings;
product identifiers;
communication channels;
contractual rules;
workflow states.

Agentic systems promise more flexible discovery and interaction. Agents may identify capabilities, interpret natural-language requests, coordinate across services, and dynamically select tools or counterparties.

But flexibility does not remove the requirement for structure.

It moves the boundary.

A language model may understand a poorly written description better than conventional software, but a consequential transaction still requires precision. The system must know whether “500” means pieces, kilograms, metres, packs, pallets, or boxes. It must distinguish a manufacturer’s identifier from a distributor’s internal code. It must determine whether stock is real, forecast, reserved, or available subject to confirmation.

The closer a system moves towards execution, the less ambiguity it can safely tolerate.

This creates an apparent paradox:

AI makes commerce more tolerant of natural language at the interface while making reliable structured data more important beneath the interface.

A person may tell an agent:

“Find me an equivalent stretch film for our wrapping machines, available next week, but do not use recycled material if it reduces puncture resistance.”

The instruction is conversational.

The execution layer is not.

The agent needs machine-readable information about:

film width;
thickness;
stretch percentage;
roll length;
core diameter;
material composition;
puncture performance;
recycled content;
machine compatibility;
pallet quantity;
availability;
lead time;
price unit;
minimum order;
certification;
delivery conditions.

The front end becomes easier for the buyer.

The data obligation becomes heavier for the supplier.

4. The Emerging Infrastructure of Machine-Readable Access

Several current infrastructures illustrate this direction.

They should not be treated as one integrated global system. Each addresses a different part of the market. Together, however, they show how identity, capability declaration, structured product information, commercial policies, and execution interfaces are becoming connected.

4.1 Universal Commerce Protocol: from product feed to executable checkout

Google describes the Universal Commerce Protocol as an open standard intended to support commerce actions through AI interfaces, beginning with direct purchasing in AI Mode and Gemini. Merchant integration is not limited to having a website or product page. Google’s implementation documentation requires preparation of Merchant Center data, including shipping information, return policies, customer-support details, and product-feed attributes. A merchant must also publish a UCP profile declaring supported capabilities, protocol versions, service endpoints, fulfilment options, discount handling, order-management functions, and payment configuration. (Google for Developers)

For native checkout, the merchant must provide a RESTful API that Google can call to create and manage checkout sessions. Google also provides an embedded route for merchants whose checkout involves more complex logic. This demonstrates a practical distinction between being visible on Google and becoming executable inside an AI-mediated purchase flow. (Google for Developers)

Google’s Merchant API schema contains structured fields for identifiers, pricing, availability, categories, dimensions, shipping, handling times, technical product details, quantities, and unit-pricing measures. Google also provides a reporting context for determining product eligibility for UCP checkout. (Google for Developers)

This does not mean that all merchants will be forced to use Google’s protocol or that UCP will become the universal standard.

It reveals a more general architecture:

the product must be represented correctly;
the merchant must declare capabilities;
policies must be available in structured form;
endpoints must be reachable;
authentication must work;
checkout must be executable;
the integration must satisfy platform validation.

The company has moved from being a webpage to being a callable commercial service.

4.2 A2A: the agent as a discoverable economic actor

The Agent2Agent protocol standardises communication between agents created by different vendors and frameworks. Its Agent Card is a structured JSON document containing an agent’s identity, description, version, service interfaces, capabilities, skills, supported media types, and authentication requirements. Clients use this information to discover suitable agents and configure interaction with them. (a2a-protocol.org)

A2A documentation allows different discovery strategies. Agent Cards may be retrieved from known locations, private registries, curated catalogues, or other discovery infrastructures. The protocol therefore supports interoperability without determining who will control discovery. (a2a-protocol.org)

That distinction is central to market access.

An open communication protocol does not guarantee that every supplier agent will be discovered.

The protocol may be open while:

the registry is private;
the catalogue is curated;
the identity provider is concentrated;
the buyer searches only pre-approved agents;
reputational data is not portable;
external agents receive fewer capabilities.

A supplier may therefore be technically compatible with A2A yet practically absent from the agent’s accessible market.

The Agent Card makes capabilities legible.

The discovery architecture determines whether those capabilities are ever seen.

4.3 Peppol and UBL: structured commerce before autonomous agents

Peppol shows that machine-processable business exchange already depends on detailed semantic conventions. Its catalogue and ordering specifications support structured product identification, supplier identity, catalogue references, standard item identifiers such as GTINs, manufacturer identifiers, buyer identifiers, coded units, quantities, pricing information, and formal order states. (Peppol Documentation)

These infrastructures were not created primarily for generative AI agents. They were built to support interoperable electronic procurement and invoicing.

Their significance for the agentic market is methodological.

They show why commercial execution requires more than natural-language understanding. The system must share sufficiently precise meanings for:

parties;
products;
quantities;
units;
prices;
orders;
changes;
acceptance;
rejection;
legal status.

An agent can help translate between systems, but it cannot safely eliminate the need for valid commercial semantics.

4.4 GS1 and Digital Product Passports: the product as an information endpoint

GS1 standards provide globally used identifiers and mechanisms for capturing and sharing product and supply-chain information. GS1 Digital Link connects identifiers such as a GTIN to online resources that may be presented in either human-readable or machine-readable form. (GS1)

The European Digital Product Passport develops the same broad direction from a regulatory and sustainability perspective. European Commission research describes DPP data as harmonised, structured, machine-readable information linked to unique product identities. Current work on steel and textiles emphasises transparency, traceability, semantic interoperability, searchability, and transferability. (Susproc)

The DPP is not an agentic-commerce protocol.

But a machine-readable product passport can become part of an agent’s qualification and comparison environment.

An agent may eventually use such information to determine:

whether a product meets regulatory requirements;
what materials it contains;
where it originated;
whether it can be repaired or recycled;
which sustainability criteria it satisfies;
whether a substitution is acceptable.

Machine-readable compliance information can therefore increase both transparency and the power of automated admission criteria.

5. The Ten Layers of Executable Visibility

A company does not become executable through one field or one API. Executable visibility is cumulative.

The following ten layers form a practical model.

5.1 Entity identity

The system must be able to establish who the firm is.

Relevant information may include:

legal name;
registration identifier;
tax number;
domain;
physical location;
ownership;
authorised representatives;
payment identity;
sanctions and compliance status;
digital credentials.

A website name is not necessarily a legal identity. A marketplace profile is not necessarily portable. An agent identifier is not automatically evidence that the represented company authorised the agent.

Identity is the first gate because every later claim depends on knowing who made it.

5.2 Product identity

The system must know which product is being offered.

A product may have:

seller item number;
manufacturer part number;
GTIN;
buyer-specific identifier;
model number;
batch number;
CAS number;
customs classification;
category code.

Without reliable identity, the agent may merge different products or treat equivalent products as unrelated.

5.3 Semantic classification

The product must be placed in a shared conceptual structure.

This may require:

marketplace taxonomy;
industry ontology;
procurement category;
technical standard;
commodity code;
regulatory class;
material classification.

Classification determines whether the product appears in the correct search universe.

A good product placed in the wrong ontology may be indistinguishable from no product at all.

5.4 Technical attributes

The agent must be able to determine whether the product satisfies the functional need.

For a simple packaging material, this can require many attributes. For a chemical, machine component, pharmaceutical ingredient, or construction product, the requirements become much more complex.

A narrative description may help.

A structured attribute permits comparison.

5.5 Units and commercial quantities

Units are not a minor formatting issue.

The agent must understand:

price per item;
price per roll;
price per kilogram;
price per thousand;
package size;
pallet quantity;
minimum order;
conversion rules;
tolerance;
base unit;
delivery unit.

A company may appear expensive because the system compares a box with a piece, a tonne with a kilogram, or a 1,500-metre roll with a 500-metre roll.

Machine-readable markets can scale small semantic errors into systematic exclusion.

5.6 Availability and time

The system must know not only what exists but what can be delivered.

Availability information may include:

current stock;
reserved quantity;
production capacity;
minimum lead time;
maximum lead time;
handling time;
dispatch calendar;
validity period;
geographic availability.

Stale availability can be worse than no availability. An agent that repeatedly encounters failed confirmation may reduce the supplier’s confidence score or remove it from future consideration.

5.7 Certificates and evidence

The company must prove claims in a form the system can verify.

Evidence may include:

quality certifications;
safety documentation;
declarations of conformity;
environmental credentials;
audit reports;
insurance;
licences;
country-of-origin records;
test results;
product passports.

A PDF may contain the relevant fact. An automated system may still need:

a standard issuer identity;
certificate number;
scope;
issue date;
expiry date;
status;
verification endpoint;
cryptographic signature;
machine-readable subject.

The future dividing line may not be certified versus uncertified.

It may be verifiably certified versus merely claiming certification.

5.8 Commercial terms

The offer must express conditions such as:

price;
currency;
tax treatment;
minimum order;
payment term;
discount;
delivery term;
quotation validity;
return policy;
warranty;
cancellation condition.

A human salesperson can combine several emails into one understanding.

An agent requires a state that can be represented, validated, and acted upon.

5.9 Agent and interface capability

The firm or its intermediary must declare what interactions are supported.

This may include:

search;
quotation;
negotiation;
reservation;
order;
status update;
cancellation;
return;
dispute;
payment;
post-purchase service.

A2A Agent Cards and UCP profiles illustrate this movement from general web presence to explicit capability declaration. (Google for Developers)

5.10 Transaction execution

The final layer is the capacity to convert representation into consequence.

Can the agent:

request and receive a current offer;
create a valid basket;
establish shipping cost;
confirm the final price;
obtain authorisation;
initiate an order;
receive a contractual acknowledgement;
follow fulfilment;
reverse or cancel where permitted?

At this level, the company is not simply machine-readable.

It is machine-operable.

6. Commercial Admissibility Comes Before Competition

Traditional descriptions of markets emphasise competition.

Firms compete on:

price;
quality;
innovation;
service;
reliability;
speed;
reputation.

But competition begins only after participants enter the comparison set.

An agent may apply an admission sequence before assessing commercial attractiveness:

Is the entity identifiable?
Is the product interpretable?
Are units compatible?
Is the certification current?
Is availability confirmed?
Is the supplier’s risk score acceptable?
Is the payment method supported?
Can the process be executed?
Is the supplier inside the permitted registry?
Does the buyer’s policy allow this route?

A supplier eliminated at question three never competes on price.

A supplier eliminated at question six never competes on quality.

A supplier eliminated at question eight may remain visible as information but cannot become a transaction.

This is commercial admissibility:

The set of technical, semantic, identity, evidence, risk, protocol, and execution conditions that a participant must satisfy before entering an agent-mediated commercial comparison.

Commercial admissibility can be legitimate.

A hospital should not buy medical supplies from an unverifiable supplier. A manufacturer should not procure hazardous chemicals without correct identification and documentation. A buyer should not allow an unknown agent to issue purchase orders.

The governance problem begins when admission rules are:

hidden;
disproportionate;
technically unnecessary;
controlled by an interested platform;
impossible for smaller firms to satisfy;
based on inaccurate data;
non-portable;
not subject to appeal.

The defining market question becomes:

Are these requirements necessary to execute trade safely, or do they function as a private licence to participate?

7. The SME Risk: Exclusion Without Rejection

Large enterprises often possess:

master-data teams;
ERP systems;
product-information management;
APIs;
integration budgets;
compliance departments;
digital certificates;
structured catalogues;
platform partnerships.

Small and medium-sized firms may possess the product but not the interface.

This creates several forms of disadvantage.

7.1 Translation costs

The firm must translate human commercial knowledge into:

taxonomies;
schemas;
codes;
APIs;
credentials;
product feeds;
protocol profiles.

This work is often underestimated because it does not look like manufacturing or selling. Yet it determines whether manufacturing and selling can reach the market.

7.2 Gateway dependency

An SME may use an intermediary to make itself agent-compatible.

The gateway may:

convert spreadsheets into structured data;
host the supplier’s agent;
verify identity;
translate protocols;
manage certificates;
connect to buyers;
process payments.

This can expand access.

It can also create a new dependency. The intermediary may gain visibility into prices, customers, margins, inventory, and transaction history. It may become impossible to change providers without losing accumulated reputation or integrations.

7.3 Semantic disadvantage

Large firms can influence standards and taxonomies. Smaller suppliers must adapt to categories created elsewhere.

A specialist product may be poorly represented by a general classification. A non-standard but effective solution may be invisible because the system recognises only standard parameters.

Innovation can become harder when admissibility requires fitting an established machine category.

7.4 Stale-data penalties

Small firms may lack real-time inventory integration.

Their product data may be correct monthly but not continuously. Agents optimised for immediate certainty may prefer larger suppliers with live feeds, even when the smaller supplier could fulfil the requirement after a brief confirmation.

Speed becomes a proxy for reliability.

Integration becomes a proxy for competence.

7.5 Unobservable exclusion

The firm may never know that it was considered and rejected.

It receives no RFQ, no explanation, and no error message.

Sales decline, but there is no visible decision to appeal.

This makes agentic exclusion different from losing a tender. The firm cannot learn from an event that it cannot observe.

The Synthocracy framework describes a wider pattern in which people and institutions are not always explicitly denied. They are simply not shown, not prioritised, or routed into a lower-access path.

7.6 Synthetic path dependence

Suppose an agent chooses a larger supplier because it has better structured data.

The supplier receives more orders.

More orders produce better operational data.

Better data improve confidence and reputation.

Higher confidence produces further selection.

The smaller competitor receives fewer transactions and therefore less evidence from which to improve its score.

The system can begin treating the consequence of its earlier preference as proof that the preference was correct.

8. Machine Readability Can Expand Access

The argument should not be reduced to a defence of unstructured commerce.

Machine-readable standards can produce substantial benefits.

A small supplier with a well-structured catalogue may become discoverable to international buyers without maintaining a large sales team. Standard identifiers can reduce confusion. Digital certificates can increase trust. Automated translation can bridge language barriers. Shared procurement documents can reduce bespoke integrations. Real-time availability can help small firms capture urgent demand.

Structured information may also reduce the advantage of personal networks.

A human buyer may repeatedly choose familiar suppliers. An agent with a broad and correctly configured discovery scope may identify qualified alternatives.

Executable visibility can therefore democratise access when:

protocols are interoperable;
discovery is open;
identity is affordable;
standards are publicly documented;
gateways compete;
reputation is portable;
data errors can be corrected;
manual exceptions remain possible;
ranking does not favour the infrastructure owner.

The question is not whether the market should become machine-readable.

It is who defines readability and whether all capable participants can obtain it on fair terms.

9. Principles for Fair Machine-Readable Market Access

9.1 Discovery scope should be disclosed

A buyer using an agent should be able to know whether the system searched:

the open market;
one marketplace;
an approved-vendor list;
a private registry;
a partner ecosystem;
only API-connected suppliers.

A restricted search should not be presented as an objective view of all available suppliers.

9.2 Rejection should leave a reason code

When a supplier is excluded before comparison, the system should preserve a meaningful reason:

identity not verified;
missing field;
unsupported unit;
expired certificate;
unavailable delivery date;
policy restriction;
transaction interface unavailable.

This need not expose confidential buyer strategy. It should make correction possible.

9.3 Commercial data should be correctable

Businesses need a route to inspect and correct the data used to represent them.

This should include:

legal identity;
product parameters;
certificates;
delivery records;
complaints;
risk classifications;
ownership;
sustainability data;
payment status.

The right to correct machine-readable commercial identity may become as important to firms as the right to correct consumer credit information.

9.4 Equivalent evidence should be accepted

A system should not reject valid certification merely because it was issued by a competing but recognised provider.

Security and compliance should not become indirect requirements to use one platform, one cloud, or one identity service unless such restriction is objectively necessary.

9.5 Unit and ontology mappings should be contestable

When an offer is rejected because of classification or unit conversion, the supplier should be able to identify the mapping error.

Semantic interpretation should not become an unchallengeable decision layer.

9.6 Reputation should be portable

Where feasible, firms should be able to transfer authenticated transaction and performance history between platforms.

A company should not lose its commercial existence merely because it changes infrastructure provider.

9.7 Gateway markets should remain competitive

SMEs will often need intermediaries.

The solution is not to eliminate gateways but to prevent one gateway from becoming the compulsory landlord of agentic market access.

9.8 Human escalation must reach someone with authority

A support chatbot that repeats the automated rejection is not an appeal.

A meaningful escalation requires a person who can:

inspect the data;
understand the classification;
suspend the outcome;
admit an exception;
correct the record.

9.9 Standards should be evaluated for distributional effects

A technically elegant standard may impose disproportionate implementation costs on smaller firms.

Governance bodies should ask not only whether a standard is interoperable, but who can afford to comply.

9.10 Market observability should include the unseen

Regulators and large buyers should monitor:

how many suppliers are excluded before comparison;
which requirements cause exclusion;
whether exclusion disproportionately affects SMEs;
how often data errors are corrected;
whether manual exceptions succeed;
how concentrated discovery and identity services become.

A market cannot be assessed only by analysing completed transactions. It must also examine opportunities that never reached execution.

10. The Executable Visibility Diagnostic for SMEs

The following diagnostic is designed as a practical first assessment. It does not certify legal or technical readiness. It helps a business identify where it may disappear from an agent-mediated market.

Score each dimension from 0 to 4:

0 — absent
1 — mostly manual or inconsistent
2 — structured internally but not externally accessible
3 — machine-accessible with limitations
4 — verified, current, interoperable, and transaction-ready

Dimension 1: Legal and digital identity

Is the legal entity clearly identifiable?
Are registration, tax, and contact data consistent across systems?
Can identity be verified digitally?
Is there an authorised representative for agentic transactions?
Can the business prove that an agent acts on its behalf?

Score: 0–4

Dimension 2: Product identity

Does every product have a stable identifier?
Are manufacturer, seller, and standard identifiers distinguished?
Are variants represented separately?
Can batches, lots, or configurations be identified when required?
Are discontinued and replacement products mapped?

Score: 0–4

Dimension 3: Classification and ontology

Is the product assigned to recognised categories?
Are industry-specific codes available?
Can an external system understand what the product is used for?
Are equivalents and substitutes represented?
Are unusual products mapped to standard attributes without losing essential meaning?

Score: 0–4

Dimension 4: Technical attributes

Are specifications stored in structured fields rather than only PDFs?
Are mandatory and optional attributes separated?
Are tolerances stated?
Are machine compatibility and application limits represented?
Can the data support automatic comparison?

Score: 0–4

Dimension 5: Units, packaging, and quantities

Are sales units clearly distinguished from base units?
Is package quantity stated?
Are conversion rules reliable?
Are MOQ, pallet quantity, and delivery units structured?
Can the agent compare the real unit price?

Score: 0–4

Dimension 6: Price and commercial terms

Is the current price machine-readable?
Are currency and tax status explicit?
Are discount rules represented?
Are payment terms available?
Are quotation validity and minimum-order rules defined?
Can prices be requested programmatically when they are not public?

Score: 0–4

Dimension 7: Availability and fulfilment

Is stock or capacity data current?
Are handling and production lead times stated?
Are shipping regions and delivery conditions structured?
Can the firm confirm or reserve availability electronically?
Can the system distinguish current stock from forecast supply?

Score: 0–4

Dimension 8: Certificates and compliance

Are certificates current?
Are issuer, scope, number, issue date, and expiry date structured?
Can authenticity be checked?
Are SDS, declarations, audits, or test reports linked to the correct product?
Can outdated documents be removed automatically?

Score: 0–4

Dimension 9: Machine interface

Is product information available through a feed or API?
Can systems request quotations electronically?
Are standard business documents supported?
Can the company receive structured orders?
Does it support recognised commerce or agent protocols directly or through a gateway?

Score: 0–4

Dimension 10: Agent capability profile

Can an external agent determine what the company’s agent or interface can do?
Are supported actions declared?
Are authentication requirements clear?
Are versions and endpoints documented?
Can capabilities be updated without breaking previous integrations?

Score: 0–4

Dimension 11: Transaction execution

Can an agent create a valid quotation or checkout session?
Can it submit or prepare an order?
Can the company acknowledge, amend, or reject the order electronically?
Are payment and fulfilment states available?
Can an order be cancelled or corrected through the same environment?

Score: 0–4

Dimension 12: Traceability and human fallback

Can the company reconstruct what the agent received and returned?
Are rejected or failed transactions logged?
Can a human intervene?
Is there a real escalation route?
Can incorrect product or company data be corrected rapidly?
Can the firm continue trading when the agentic route fails?

Score: 0–4

11. Interpreting the Score

The maximum score is 48.

0–12: Human-only market participant

The business depends primarily on human interpretation and manual communication. It may be visible online but is unlikely to enter automated comparison or execution without an intermediary.

Primary risk: machine invisibility.

13–24: Digitally visible but non-executable

The company has digital content and some structured information, but critical data remain inconsistent, inaccessible, or unsuitable for automated qualification.

Primary risk: the firm is found but removed before execution.

25–34: Agent-compatible through adaptation

The business can participate in some automated workflows, often through platform-specific feeds, manual updates, or gateways.

Primary risk: dependency on intermediaries and fragmented data.

35–42: Agent-ready

The company has stable identity, structured product data, current commercial information, machine-accessible documentation, and functional electronic workflows.

Primary risk: platform-specific lock-in or incomplete traceability.

43–48: Executable visibility

The participant can be discovered, interpreted, verified, compared, qualified, transacted with, monitored, and reached through meaningful human fallback.

This does not guarantee selection.

It means the company can genuinely enter the decision environment.

12. A Ninety-Day SME Readiness Programme

Days 1–30: Establish the canonical commercial record

The firm should identify one authoritative source for:

company identity;
product identifiers;
technical attributes;
units;
certificates;
price rules;
availability;
delivery conditions.

Conflicting spreadsheets, catalogues, PDFs, and website descriptions should be reconciled.

The first objective is not to build an autonomous agent.

It is to create one coherent machine-readable version of the business.

Days 31–60: Build external legibility

The firm should then:

assign or validate identifiers;
map products to recognised categories;
structure technical attributes;
standardise units;
digitise certificate metadata;
create current availability fields;
establish feeds or APIs;
document supported commercial actions.

At this stage, the company should test whether an external system can interpret the offer without a salesperson explaining it.

Days 61–90: Test execution and failure

The firm should simulate several workflows:

product discovery;
qualification;
RFQ;
price comparison;
certificate validation;
stock confirmation;
order creation;
cancellation;
error correction;
human escalation.

The most important question is not whether the ideal transaction succeeds.

It is whether the company remains visible when something goes wrong.

13. Research Questions for Synthocracy Institute

The transition towards machine-readable market access requires systematic empirical investigation.

13.1 What causes pre-comparison exclusion?

Research should measure how often firms are excluded because of:

missing structured data;
classification errors;
unsupported units;
inaccessible certificates;
absent inventory feeds;
identity failures;
protocol incompatibility;
platform restrictions.

13.2 Who bears the cost of legibility?

The cost may fall on:

suppliers;
buyers;
platforms;
gateways;
public infrastructure;
standardisation bodies.

The distribution of this cost will shape which firms remain competitive.

13.3 Are agents searching markets or ecosystems?

An agent may appear to search widely while operating inside one platform’s commercial boundaries.

Research should distinguish the apparent market from the searched market.

13.4 Can firms observe their own exclusion?

A key indicator of market fairness will be whether a supplier can know:

that it was discovered;
that it was parsed incorrectly;
that it failed qualification;
which condition caused the failure;
how to correct it.

13.5 Does machine legibility favour incumbents?

Large firms may possess better integration while small firms may be more flexible, innovative, or specialised.

The research question is whether agentic markets evaluate commercial value or mainly reward data maturity.

13.6 Do gateways broaden access or create digital tenancy?

Gateways may become essential infrastructure for SMEs.

Their pricing, data rights, portability, conflicts of interest, and switching conditions deserve scrutiny.

13.7 How should commercial identity correction work?

A standard procedure may be needed for disputing:

erroneous classifications;
expired-status errors;
false ownership information;
incorrect delivery records;
misinterpreted certificates;
duplicated identities;
synthetic blacklisting.

13.8 What constitutes a functional manual fallback?

Research should determine when a supposedly available manual channel is too slow, costly, or powerless to represent real access.

Conclusion: The Company the Agent Can Act Upon

The first digital economy asked whether a business had a website.

The platform economy asked whether it could be listed, ranked, and reviewed.

The answer economy asks whether its information can be cited and summarised.

The agentic economy asks a more demanding question:

Can the business be acted upon?

Can it be identified?

Can its products be interpreted?

Can its units be compared?

Can its certificates be verified?

Can its availability be trusted?

Can its commercial conditions be represented?

Can its capabilities be discovered?

Can an authorised system request, negotiate, order, pay, cancel, and preserve a record?

A company that fails these tests may continue to exist for law, taxation, employees, owners, and established customers.

It may still cease to exist for the system that increasingly prepares new commercial decisions.

This is why machine-readable market access cannot be reduced to technical optimisation. It concerns the distribution of economic opportunity.

Structured information can broaden markets, reduce friction, improve trust, and help smaller firms reach buyers. But when one interpretation standard, registry, platform, identity provider, or execution path becomes the practical gate to participation, readability becomes power.

The defining inequality of the agentic market may not be between firms that have websites and firms that do not.

It may be between:

firms that can be read;
firms that can be verified;
firms that can be compared;
firms that can be admitted;
and firms that can be executed.

The concept of executable visibility names the final threshold.

A firm possesses executable visibility when it is not merely present in the market’s information layer but can enter its operational decision layer.

The central question therefore remains:

Does a business have meaningful market access if it can be found by humans but cannot be interpreted or executed by agents?

The answer is increasingly no.

A human may still know that the business exists.

The market may increasingly behave as though it does not.