For nearly a century, investors and executives have relied on a consistent set of accounting and valuation frameworks developed for the Industrial and Information Economies. Revenue. EBITDA. Cash flow. Assets under management. These metrics emerged because they correctly measured the productive resources that mattered in each era: physical capital, recurring software subscriptions, transaction volume.
Generally Accepted Accounting Principles — GAAP — was not designed to measure intelligence. It was designed to measure industrial and commercial activity, and it does this with considerable precision. The problem is not that GAAP is wrong. The problem is that it is blind to the most valuable thing an agentic network creates.
The Four Blind Spots
Blind Spot 1: Institutional Memory Has No Balance Sheet Line
When an enterprise deploys AI agents that execute thousands of decisions per day, every execution enriches the network's memory, refines its routing model, and improves the quality of future outputs. This accumulated intelligence — the institutional memory of the network — is not an intangible under any current accounting standard. It does not appear as an asset. It generates no depreciation schedule. It cannot be used as collateral. It is, from a GAAP perspective, invisible.
Yet this invisible asset is precisely what distinguishes a mature agentic network from a newly launched one. The ten million annotated executions that an established platform holds in its memory represent a competitive moat that no competitor can replicate by writing a cheque. GAAP accounting records neither its existence nor its value.
Blind Spot 2: Trust Capital Is Not Measured
The trust score accumulated by an intelligence asset — the verified track record of accurate, well-governed outputs built over thousands of executions — determines its settlement rate, its discovery weight, and its collateral value. Trust capital is an economically significant asset. It is the difference between an intelligence asset that commands premium pricing and one that executes at the floor rate.
No current accounting standard has a category for trust capital. No auditor measures it. No financial statement reports it. An agentic enterprise's most strategically important assets are therefore systematically absent from every financial document it produces.
Blind Spot 3: The Compounding Dynamic Is Invisible to DCF
Standard discounted cash flow analysis assumes that productive assets depreciate over time — that last year's investment is worth less this year, and less still next year. This assumption is correct for most asset classes. It is incorrect for well-maintained intelligence assets, which appreciate as they accumulate execution history, trust scores, and contextual memory.
A conventional DCF model applied to an agentic platform will systematically undervalue it in early periods (when trust scores are still developing and appreciation has not yet compounded) and correctly value it only as the network matures. For investment decisions made in the critical early scaling phase, this produces material mispricing.
Blind Spot 4: Network Effects Are Not In the Multiple
SaaS companies are valued on revenue multiples calibrated to their growth rates. The multiple captures, implicitly, the value of recurring revenue, high margins, and modest network effects. It does not capture the value of a network in which every additional execution makes every future execution more valuable — where growth is multiplicative rather than additive. Applying a SaaS multiple to an agentic network produces the same error as applying a SaaS multiple to a financial exchange. The business model is categorically different.
The SaaS valuation model — enterprise value as a multiple of ARR — became the dominant framework for technology company valuation over the last two decades for a specific reason: it worked. Software companies with recurring revenue, high gross margins, and predictable expansion rates traded at multiples of their ARR that accurately reflected their economic properties. The multiple was not arbitrary; it was calibrated to the specific dynamics of software businesses.
Those dynamics do not describe agentic networks. Three structural differences make SaaS multiples systematically inappropriate.
Structural Difference 1: Revenue is a Lagging Indicator
In a SaaS business, revenue is the primary indicator of economic health. Customers pay for subscriptions; subscriptions grow with the customer base; the customer base grows through sales effort. The relationship between the business's current activities and its future revenue is relatively direct.
In an agentic network, the primary indicator of economic health is not current revenue — it is current intelligence creation, execution quality, and trust accumulation. A platform that is rapidly building its Network Intelligence Capital may have modest current revenue while building the execution infrastructure that will generate far larger revenue in future periods. Conversely, a platform with strong current ARR that is failing to maintain its trust scores and memory quality is deteriorating in economic terms even as its reported revenue grows. Revenue, in this context, is a lagging and potentially misleading indicator.
Structural Difference 2: The Unit Economics Are Different
SaaS economics are characterised by high upfront customer acquisition costs, predictable recurring revenue, high gross margins, and churn as the primary risk factor. The key question in SaaS is: does the lifetime value of the customer exceed the cost of acquiring them?
Agentic economics are characterised by different unit economics entirely. The marginal cost of an additional execution approaches zero as the network scales. Revenue per execution can increase over time as trust scores improve. The primary risk is not churn but quality degradation — the erosion of trust capital that causes execution rates and settlement prices to fall. The right unit economic question is not LTV/CAC. It is: does the intelligence capital created by each execution compound faster than it decays?
Structural Difference 3: The Moat Is Different
SaaS moats derive primarily from switching costs (data migration friction, workflow integration, trained users) and product differentiation. These moats are real but vulnerable to a sufficiently well-resourced competitor who can absorb the switching cost on behalf of customers.
Agentic moats derive from accumulated execution history, trust capital, and network memory — assets that take years to build and cannot be acquired through spending. A competitor cannot buy ten million annotated executions. They cannot purchase a verified trust score. The moat is time and quality, not capital.
The question for an agentic network is not what its revenue multiple should be. It is how to measure and value the Intelligence Capital that no current accounting system can see.
Every major economic transition in history created a new productive asset class that required new measurement frameworks. Land required cadastral surveys and property law. Industrial machinery required depreciation schedules and capital accounting. Financial instruments required mark-to-market accounting and fair value standards. Software required new treatment of intangible assets and subscription revenue recognition.
The Agentic Economy creates its fifth productive asset class: Persistent Executable Intelligence.
What Makes It Different From Every Previous Asset Class
Physical assets depreciate through use. A machine that has produced a million units is worth less than a new machine. Software depreciates through obsolescence — last year's operating system is worth less than this year's. Data depreciates through staleness — last year's customer preferences are less valuable than this year's.
Persistent Executable Intelligence, governed correctly, appreciates through use. An intelligence asset that has executed one million decisions is more valuable than one that has executed one thousand, because it has accumulated the trust scores, memory context, and routing optimisation that make its outputs more reliable and more efficiently delivered. The asset becomes more valuable as it is consumed — a property no previous asset class has had.
The Five Properties
Persistent Executable Intelligence has five properties that distinguish it from every previous productive asset and that must be reflected in any accounting and valuation framework designed to measure it accurately.
Persistence: it does not disappear when its creator retires or when the engagement ends. It continues executing indefinitely under appropriate maintenance.
Executability: unlike a document or a database, it reasons. It takes inputs and produces governed outputs through a reasoning process that can be audited and verified.
Appreciation: well-maintained assets improve in value through execution, as trust scores accumulate and routing models refine.
Attribution: every output is traceable to its contributing knowledge sources, enabling multi-party royalty distribution and creating a legal and economic identity for the asset.
Composability: intelligence assets can be combined, chained, and orchestrated to produce outputs that no single asset could produce alone. This composability creates network effects unique to the asset class.
The Agentic Economy Valuation Standard is designed to make these five properties measurable, auditable, and reflected in financial statements and valuation models. It does not replace GAAP. It extends GAAP to accommodate the productive assets that GAAP was not designed to see.
Every agentic network, regardless of domain, technology stack, or business model, operates through five primitive layers. These layers are not a product architecture or a software design pattern. They are the economic anatomy of any system that creates, distributes, and compounds autonomous intelligence. Understanding them is the prerequisite for understanding how agentic networks create value and how that value should be measured.
Creation — Intelligence assets are produced
↓
Discovery — Intelligence is matched to objectives
↓
Execution — Intelligence produces economic output
↓
Settlement — Value is distributed to contributors
↓
Compounding — The network learns and improves
Each layer generates its own measurable outputs that feed into the AEVS valuation model. The layers are not independent — each layer's health directly affects the layers above and below it. A platform with excellent creation metrics but poor discovery will accumulate unused intelligence that generates no execution value. A platform with excellent execution metrics but poor compounding will fail to improve over time and lose ground to competitors whose learning rate is higher.
The critical insight — the one that makes the AEVS necessary — is that Compounding is not merely the fifth layer. It is the reason the first four layers are worth building. A platform that executes without compounding is a service. A platform that executes and compounds is an infrastructure business.
The Creation layer measures the production of intelligence assets — the agents, workflows, knowledge graphs, digital experts, and reasoning utilities that the network will execute. Creation is analogous to manufacturing in the industrial economy: it is where the productive resource is first brought into existence.
Primary Creation Metrics
Intelligence Asset Count (IAC)
The total number of verified intelligence assets available on the platform at a given point in time. This is the headline supply metric — the equivalent of total SKUs in a marketplace or total listed properties in a real estate platform.
Measurement: counted from the verified asset registry. An asset is counted only after it has passed verification testing and received an initial trust score. Unverified drafts are not counted.
Asset Creation Velocity (ACV)
The rate at which new verified assets are being added to the network. ACV measures the health of the creator pipeline — whether the supply of intelligence is growing quickly enough to meet execution demand.
Measurement: weekly or monthly. Healthy networks show ACV growing faster than execution demand in early stages, then tracking demand growth in mature stages. A declining ACV against growing demand is an early warning signal.
Quality-Adjusted Intelligence Asset Count (QIAC)
A weighted count of intelligence assets that adjusts for quality. An asset with a trust score of 0.95 contributes more to QIAC than an asset with a trust score of 0.60. QIAC is the supply metric that most accurately reflects the economically useful intelligence available on the platform.
Measurement: computed from the verified trust scores and rolling 90-day execution frequencies of all active assets. Trust scores are verified by the domain validation system; they cannot be self-reported.
Creation Economics
The economic health of the Creation layer depends on two relationships: whether creators are being sufficiently compensated to continue creating, and whether the quality of new assets is maintaining or improving the platform's average trust score. A platform where creator economics are attractive but where new assets are low quality faces a specific risk: execution volume may grow while average output quality declines, eroding enterprise trust and eventually reducing settlement rates across the entire network.
The Creator Retention Rate — the fraction of active creators who publish at least one new asset in a given twelve-month period — is the leading indicator of creation layer health. Platforms targeting above 70% creator retention can be considered healthy in their creation economics. Below 50% indicates that creator incentives need structural adjustment.
The Discovery layer matches execution objectives to intelligence assets. It is the platform's allocation mechanism — the system that determines which intelligence executes against which problem. Discovery is not search. Search retrieves documents ranked by relevance. Discovery allocates capability ranked by expected outcome quality.
Primary Discovery Metrics
Intelligence Discovery Volume (IDV)
The total number of discovery events in a given period across all channels: direct queries, API calls, agent-routing requests, marketplace recommendations, and automated workflow triggers.
Measurement: server-side event logging. Each distinct discovery event is counted once. Repeated identical queries within a session are deduplicated.
Discovery Efficiency Rate (DER)
The fraction of discovery events that result in a successful execution. A high DER indicates that the platform's matching algorithm is accurately routing objectives to capable assets. A low DER indicates either insufficient supply in the requested domain or matching algorithm weakness.
Measurement: the numerator is executions that complete with a verification score above the domain-specific quality threshold. The denominator is all discovery events including those where no suitable asset was found. Target DER: above 0.75 for a mature platform.
Discovery Liquidity (DL)
The fraction of published assets that receive at least one execution request per month. DL measures whether the supply of intelligence is being actively used or accumulating as dead inventory.
Measurement: monthly. DL below 0.30 indicates a supply-demand mismatch — the platform is accumulating intelligence that is not being discovered. Above 0.60 indicates healthy utilisation.
Discovery Yield (DY)
The average economic value generated per discovery event. DY captures the quality of the matching — whether the platform is routing executions to high-value assets in high-value domains.
The Execution layer is where intelligence produces economic output. Every execution is an economic event: it consumes resources (reasoning, memory, governance, verification) and produces a governed output with a measurable economic value to the enterprise that requested it. The Execution layer is the engine of the platform — it is where the Intelligence Liquidity Gap is closed, one decision at a time.
Primary Execution Metrics
Gross Intelligence Value (GIV)
The total economic value created by all intelligence executions on the platform in a given period. GIV is the AEVS's primary headline metric — the equivalent of GMV for a marketplace, TPV for a payments network, or compute consumption for a cloud provider. It measures the scale of productive activity flowing through the platform, not the platform's revenue capture from that activity.
The Average Economic Value of an execution is measured against the cost of the best available alternative — typically the cost of human expert labor required to produce an equivalent output. A compliance investigation that would require four hours of senior compliance counsel time at $500 per hour has an execution economic value of $2,000. The sum of these values across all executions in the period is GIV.
Measurement: requires domain-specific benchmarking of human-equivalent costs. The platform maintains a benchmarking library for each active domain, updated annually against market rates for comparable professional services. GIV is an independently auditable figure.
Execution Volume (EV)
The total number of executions in a given period, without weighting for economic value. EV is the volume metric; GIV is the value metric. Both are necessary. Strong EV with declining GIV indicates execution is shifting toward lower-value domains. Declining EV with stable GIV indicates the platform is concentrating on higher-value executions.
Average Execution Value (AEV)
GIV divided by EV. The average economic value created per execution. AEV trending upward over time indicates the platform is successfully routing executions toward higher-value domains and higher-trust assets. AEV trending downward is an early warning of quality degradation or domain mix shift.
Platform Intelligence Capture (PIC)
The platform's revenue from intelligence execution. PIC is the fraction of GIV that flows to the platform as revenue. It is the platform's 'take rate' — analogous to the fee that Visa charges on payment volume or the commission that Amazon charges on marketplace GMV.
Capture Rate varies by revenue channel. Enterprise contracts typically carry a 7–15% capture rate. Marketplace transactions carry 10–20%. API consumption carries 8–12%. The blended Capture Rate is reported as the weighted average across all channels.
The Settlement layer distributes the economic value created by execution to the contributors who made it possible. Settlement is what distinguishes a governed intelligence platform from an AI service. In an AI service, the vendor collects all revenue from the output it produces. In a governed intelligence platform, the revenue is attributed and distributed to each contributor according to their verifiable contribution.
Settlement has two economic functions. The direct function: compensating creators for the value their intelligence produces, creating the incentive for continued high-quality contribution. The indirect function: creating the attribution record that makes intelligence assets auditable, legally defensible, and usable as collateral — the foundation of the Intelligence Capital Market.
Settlement Metrics
Settlement Volume
The total amount distributed to contributors in a given period. Settlement Volume is always equal to PIC minus the platform's operating costs and margin. For platforms operating on a 93/7 creator-platform split, Settlement Volume equals 93% of PIC.
Attribution Depth
The average number of distinct contributors receiving settlement per execution. Attribution Depth measures how richly the platform's royalty system captures the multi-party nature of intelligence production. Low Attribution Depth (approaching 1.0) indicates that settlement is concentrating in primary asset owners and not flowing through to underlying knowledge graph owners, data providers, and methodology contributors. High Attribution Depth (above 4.0) indicates a well-functioning multi-party royalty system.
Settlement Accuracy Rate (SAR)
The fraction of settlement transactions that complete correctly on the first attempt, without dispute or revision. SAR is the quality metric for the settlement infrastructure. Below 99.5% warrants technical review. Below 98% is a serious operational concern.
Compounding is the most important layer in the AEVS framework. It is the mechanism through which an agentic network transcends being a service and becomes an infrastructure business. Every layer above it generates current period value. Compounding determines whether the platform's future periods are more valuable than its current period — and by how much.
In a conventional software business, yesterday's revenue does not make today's product better. A SaaS company that processed a million customer transactions last month has the same underlying software this month as it did last month. In an agentic network, yesterday's executions directly improve today's routing quality, trust scores, and memory depth. The platform becomes more valuable as it operates.
Compounding Mechanics
Compounding operates through four simultaneous feedback mechanisms that reinforce each other continuously.
Memory Compounding
Every execution adds to the platform's operational memory — the historical record of which assets produced which outputs in which contexts with which quality scores. A platform with deep memory routes new executions more accurately and can predict quality outcomes more reliably than a platform with shallow memory. Memory compounds continuously with no upper bound.
Trust Compounding
Every verified execution updates the trust score of every participating asset. High-quality outputs increase trust scores; low-quality outputs decrease them. As trust scores rise, assets command higher settlement rates and receive greater discovery weight, generating more executions, which produce more trust data, which further refines the scores. Trust capital compounds recursively.
Routing Compounding
The discovery layer's matching algorithm improves with every execution it routes. Each completed execution — whether the output was excellent, adequate, or poor — provides training signal for the routing model. A platform with ten million routing decisions has a fundamentally more accurate allocation engine than one with one hundred thousand. Routing quality compounds with execution volume.
Network Compounding
Each new creator that joins the network expands the supply of executable intelligence available to all enterprises. Each new enterprise that joins creates execution demand that benefits all creators. The network effects of an agentic platform are bidirectional and compound with every participant added to either side.
The Network Intelligence Capital (NIC) Metric
Network Intelligence Capital is the AEVS's primary measure of the Compounding layer — the accumulated productive intelligence capital that the network has built through its operating history. NIC is the balance sheet asset that GAAP cannot see and that distinguishes the AEVS from conventional accounting.
Each variable has a defined measurement methodology.
Quality_i: the asset's current trust score (0.0–1.0), verified by the domain validation system.
Trust_i: the asset's verified execution accuracy rate over the trailing twelve months.
Execution_Freq_i: the asset's average monthly execution count over the trailing three months, normalised to a 0.0–1.0 scale against the network maximum.
Reuse_Rate_i: the fraction of the asset's executions that draw on its outputs for downstream executions (a measure of how foundational the asset is to other assets).
Connectivity_i: the number of other assets this asset contributes to or receives contributions from, normalised to a 0.0–1.0 scale.
NIC is computed at the asset level and aggregated to the platform level. A platform's NIC grows every month through three mechanisms: new high-quality assets are published, existing assets accumulate execution history and improve their trust scores, and the network's connectivity deepens as assets are combined into new workflows and orchestration graphs.
NIC Decay: assets that are not updated or executed for twelve months receive a decay factor of 0.05 per quarter applied to their NIC contribution. Assets that are deprecated are removed from NIC. This prevents NIC from inflating through accumulation of obsolete intelligence.
The AEVS defines eleven primary operating metrics that together provide a complete picture of an agentic network's economic health. These metrics are designed to be independently measurable, auditable, and comparable across platforms. Each metric has a defined calculation methodology, a defined data source, and a defined benchmark range.
| Metric | Definition | Healthy Range |
|---|---|---|
| GIV | Total economic value created by executions in period | Growing >20% YoY |
| NIC | Accumulated network intelligence capital (balance sheet) | Growing >15% YoY |
| PIC | Platform revenue from intelligence execution | Capture rate 7–20% |
| IAC | Verified intelligence assets published and active | >500 in primary domains |
| QIAC | Quality-adjusted asset count | QIAC/IAC ratio >0.70 |
| ACV | New verified assets published per month | ACV/IAC ratio >5% |
| IDV | Total discovery events per period | DER >0.75 |
| DER | Fraction of discovery events completing execution | Above 0.75 |
| DL | Fraction of assets receiving execution requests monthly | Above 0.60 |
| AEV | Average economic value per execution | Rising or stable YoY |
| TCR | Verified high-quality executions / total executions | Above 0.90 |
Two of these metrics deserve special attention because they serve as leading indicators of platform health: TCR (Trust Coverage Ratio) and AEV (Average Execution Value). TCR measures quality — whether the executions being produced are meeting the standard required for enterprise reliance. AEV measures value intensity — whether the platform is successfully routing executions toward higher-value use cases. Both metrics can decline while headline GIV and EV continue growing, which is why they must be tracked independently and not inferred from volume metrics alone.
The AEVS extends the conventional balance sheet with five new asset categories that capture the productive intelligence of an agentic network. These categories do not replace conventional balance sheet items — they supplement them. The AEVS balance sheet is additive: a company's AEVS balance sheet includes all conventional assets plus the five new intelligence asset categories.
The Five Intelligence Asset Categories
Memory Capital (MC)
The economic value embedded in the platform's historical execution record — the accumulated memory of which assets produced which outputs under which conditions with which quality scores. Memory Capital is the foundation of the routing model's quality advantage.
Knowledge Retention Rate: the fraction of execution memory that remains operationally relevant given domain evolution. Measured annually through domain council review. For stable domains (constitutional law, mathematics), retention rates approach 0.95. For rapidly evolving domains (cybersecurity, market regulation), retention rates may be 0.60–0.75.
Trust Capital (TC)
The economic value embedded in the platform's verified trust scores — the track record of accurate, well-governed outputs that allows enterprises to rely on the platform's intelligence without extensive independent verification.
Trust Capital is measured in expected annual settlement revenue attributable to trust premiums — the additional settlement that high-trust assets command above the floor rate for their domain. An asset with a trust score of 0.92 generating $2M in annual settlement, where the floor rate for the domain would generate $800K, contributes $1.2M to Trust Capital.
Reputation Capital (RC)
The economic value embedded in the platform's contributor reputation system — the verified track records of individual creators, domain councils, and institutional contributors that make the platform's governance credible.
Agent Capital (AC)
The fair value of the platform's active intelligence asset portfolio, computed as the present value of expected future settlement revenue from each asset.
The discount rate r_i is asset-specific, reflecting the domain's rate of obsolescence (faster-evolving domains carry higher discount rates), the asset's creator maintenance track record, and the competitive density of the domain.
Discovery Capital (DC)
The economic value embedded in the platform's discovery and routing infrastructure — the trained matching model, the execution history that has calibrated it, and the network relationships that enable composable execution plan construction. Discovery Capital is the hardest intelligence asset category to replicate because it is built from execution history that no competitor can acquire through investment.
Routing Model Quality Score: the measured improvement in DER attributable to the routing model versus a naive baseline (random asset selection within domain). A platform with a Routing Model Quality Score of 0.30 has a matching model that improves DER by 30 percentage points above random selection.
The AEVS Balance Sheet Summary
| Intelligence Asset Category | Calculation Basis |
|---|---|
| Memory Capital (MC) | Historical executions × retention × value × reuse |
| Trust Capital (TC) | Trust premium × annual settlement, by asset |
| Reputation Capital (RC) | Contributors × performance × frequency |
| Agent Capital (AC) | PV of future settlement per asset, trust-adjusted |
| Discovery Capital (DC) | Routing quality × addressable GIV × DER |
| Total Intelligence Capital | MC + TC + RC + AC + DC |
The AEVS income statement augments conventional P&L with intelligence-specific line items that capture how the platform's intelligence is creating and distributing value. The intelligence income statement has four sections: intelligence revenue, intelligence cost of goods, intelligence gross profit, and intelligence reinvestment.
Intelligence Revenue
Intelligence revenue captures all economic value flowing through the platform, not merely the fraction the platform retains as conventional revenue. This includes platform revenue (PIC), creator settlements distributed to contributors, and governance fees.
| Revenue Line | Definition |
|---|---|
| Gross Intelligence Value (GIV) | Total economic value created by all executions |
| Platform Intelligence Capture (PIC) | Platform's share: GIV × Capture Rate |
| Creator Settlement Distributed | Contributor share: GIV × (1 − Capture Rate) |
| Governance Revenue | Fees from verification, dispute resolution, council review |
Intelligence Cost of Goods
The costs directly associated with producing intelligence executions: compute costs for AI inference, memory access costs, verification costs, governance overhead, and settlement processing costs.
Intelligence Gross Profit and NIC Investment
Intelligence Gross Profit is PIC minus Intelligence COGS. The critical additional line item is NIC Investment: the portion of gross profit reinvested into activities that increase Network Intelligence Capital — genesis subsidies for new creator onboarding, routing model training, domain council funding, and memory infrastructure. NIC Investment is the AEVS equivalent of R&D capitalisation: it is investment in the platform's most important long-term productive asset.
NIC Investment Rate: NIC Investment / Intelligence Gross Profit (target: >25%)
Beyond the primary operating metrics and the balance sheet categories, the AEVS defines four network health metrics that measure the structural integrity of the platform's intelligence ecosystem. These metrics are leading rather than lagging — they predict future performance rather than measuring current performance.
Intelligence Velocity (IV)
The ratio of GIV to circulating intelligence supply (QIAC). Intelligence Velocity measures how productively the platform's intelligence is being used. A high IV indicates that available intelligence is being actively executed and generating economic value. A low IV indicates supply-demand mismatch — too much supply relative to execution demand, or discovery failures preventing available intelligence from reaching relevant execution requests.
Intelligence Compounding Rate (ICR)
The rate at which NIC grows per unit of GIV. ICR is the core measure of whether the Compounding layer is working — whether executions are translating into accumulated intelligence capital at a healthy rate.
Target ICR: above 0.15 (every $1 of GIV should produce at least $0.15 of NIC growth). ICR below 0.10 indicates the platform is extracting value from its intelligence stock faster than it is building it — a position that is sustainable only briefly.
Intelligence Moat Index (IMI)
A composite measure of the platform's competitive defensibility, combining the four dimensions of the intelligence moat: Memory (execution history depth), Trust (average trust score across active assets), Reuse (average reuse rate of assets), and Connectivity (average asset connectivity in the execution graph).
Each component is normalised to a 0.0–1.0 scale. IMI ranges from 0.0 (no moat) to 1.0 (maximum moat on all dimensions). An IMI above 0.50 indicates a defensible competitive position. Above 0.70 indicates a dominant position in the served domains.
Intelligence Inflation Rate (IIR)
The fraction of newly published assets whose first verification score falls below the domain quality threshold. IIR measures whether the quality bar for new asset creation is being maintained. Rising IIR indicates that the creation incentives are attracting low-quality submissions. This is the earliest detectable signal of trust score degradation before it manifests in execution metrics.
Target IIR: below 0.15. Above 0.25 warrants review of creator onboarding standards. Above 0.35 requires immediate governance intervention.
Discounted cash flow analysis is the foundation of rigorous asset valuation. The logic is impeccable: an asset is worth the present value of all future cash flows it will generate, discounted at a rate that reflects the risk that those cash flows will not materialise. DCF is not wrong for agentic networks. It is incomplete.
The incompleteness has two sources. First, standard DCF models project cash flows assuming that the asset's productive capacity is either constant or declining over time. For an agentic network, productive capacity typically increases over time as NIC accumulates — the platform generates more value per execution in year five than in year one. A DCF model that projects constant productivity from year one understates terminal value systematically.
Second, standard DCF models use a single discount rate to reflect risk. For agentic networks, the risk profile is multi-dimensional: there is execution volume risk, trust score risk, creator retention risk, and platform governance risk, each with different time horizons and different sensitivities to management quality. A single discount rate cannot capture this structure accurately.
The AEVS DCF Modification
The AEVS modifies standard DCF in two ways. First, it introduces a Trust Multiplier that adjusts projected execution value by the platform's current and projected trust score trajectory — reflecting the fact that high-trust platforms command higher settlement rates that compound over time. Second, it introduces a Learning Rate that adjusts the discount rate downward for platforms with demonstrated NIC compounding, reflecting the reduction in risk that comes from a platform whose quality improves with use.
Where Trust(t) = the platform's projected trust score in period t, rising from the current level toward the domain ceiling as execution history accumulates. The Learning Rate is platform-specific and is estimated from the platform's historical ICR — a platform with a demonstrated ICR of 0.20 earns a Learning Rate discount of 1–2 percentage points, reflecting the risk reduction that comes from a demonstrably improving intelligence stock.
The AEVS Enterprise Value formula has five components, each measuring a distinct source of value that conventional models either miss entirely or capture inadequately. The components are additive — not multiplicative. Each represents a separable claim on the platform's future economic output.
This formula replaces the conventional EV = Revenue × Multiple. It is not a simplification — it is more complex to apply, which is appropriate because the underlying economic structure of an agentic network is more complex than a SaaS business. But it is not more complex than necessary: each component has a defined calculation methodology that two independent analysts can apply to the same company and arrive at the same answer.
Infrastructure Value measures the economic worth of the platform as a neutral intermediary — the value it would have even if its current Intelligence Capital stock were stripped away and replaced with generic alternatives. Infrastructure Value is the floor: it reflects the platform's scale, its capture rate, and its operational efficiency.
The Infrastructure Multiple is calibrated from comparable infrastructure businesses in adjacent markets. Payment networks trade at 20–35× annual capture revenue. Cloud infrastructure businesses trade at 15–25×. Content delivery networks trade at 10–20×. Agentic infrastructure businesses should trade at multiples in the upper range of these comparables, given their lower marginal cost structure and higher growth rates, subject to adjustment for stage of development.
| Development Stage | Indicative Infrastructure Multiple |
|---|---|
| Pre-market (< $10M GIV) | 8–12× PIC |
| Early market ($10M–$100M GIV) | 12–18× PIC |
| Growth ($100M–$1B GIV) | 18–28× PIC |
| Scale (>$1B GIV) | 25–35× PIC |
NIC Value measures the economic worth of the platform's accumulated intelligence stock — the Memory Capital, Trust Capital, Reputation Capital, Agent Capital, and Discovery Capital built through the platform's operating history. This is the component that GAAP cannot see and that the AEVS was specifically designed to capture.
The Knowledge Multiple reflects the market's assessment of how much the current NIC stock will grow in future periods — the compounding premium. Platforms with high ICR (strong NIC compounding) and high IMI (deep competitive moat) earn higher Knowledge Multiples than platforms with weak compounding dynamics.
| ICR / IMI Profile | Indicative Knowledge Multiple |
|---|---|
| ICR < 0.10, IMI < 0.40 (weak compounding, weak moat) | 1.0–2.0× NIC |
| ICR 0.10–0.20, IMI 0.40–0.60 (moderate compounding) | 2.0–4.0× NIC |
| ICR > 0.20, IMI 0.60–0.75 (strong compounding) | 4.0–7.0× NIC |
| ICR > 0.25, IMI > 0.75 (exceptional compounding, dominant moat) | 7.0–12.0× NIC |
The GIV Value component captures the present value of the platform's intelligence execution revenue stream, adjusted for the trust trajectory and the platform's capture rate. This is conceptually closest to the conventional DCF approach but is modified to incorporate the Trust Multiplier described in Chapter 14.
In practice, for companies at early to mid-stage, a simplified approximation is acceptable for initial valuation purposes: apply the current-year GIV, adjusted by the Trust Multiplier for the next twelve-month expected trust trajectory, and multiply by an Execution Multiple calibrated to the platform's GIV growth rate.
| GIV Growth Rate (TTM) | Indicative Execution Multiple |
|---|---|
| < 50% YoY | 6–10× Annual GIV |
| 50–100% YoY | 10–16× Annual GIV |
| 100–200% YoY | 16–25× Annual GIV |
| > 200% YoY | 25–40× Annual GIV (subject to quality adjustment) |
Exchange Value captures the worth of the platform's marketplace function — the mechanism through which intelligence assets are listed, discovered, licensed, and traded. As the platform matures, the exchange function becomes increasingly valuable independent of any specific intelligence assets: it is the infrastructure through which the Intelligence Capital Market operates.
The Exchange Multiple is calibrated from financial exchange comparables. Major stock exchanges trade at 20–40× annual revenue. Cryptocurrency exchanges trade at 10–25×. For agentic exchanges in early stages, apply a discount to these multiples pending demonstrated liquidity and volume stability. The Exchange Value component becomes material only after the platform has established consistent daily settlement volume above $1 million.
Platform Optionality captures the option value embedded in the platform's infrastructure position — the future markets and business lines that the platform's current infrastructure makes possible but that are not yet generating revenue. Every major infrastructure business has developed optionality: AWS grew from cloud compute to AI services to quantum; Ethereum grew from smart contracts to DeFi to NFTs. Agentic infrastructure has comparable optionality.
Platform Optionality Index (POI): Addressable Future Markets × Probability × Infrastructure Readiness
Optionality should be valued conservatively in early stages. Applying aggressive POI multiples to speculative future markets is one of the most common errors in technology company valuation and is specifically warned against in the AEVS framework. POI contributes to EV when: the future market is clearly defined and addressable, the platform's current infrastructure directly enables entry, and comparable market development timelines support a probability above 30%.
The terminal value of an agentic network is fundamentally different from the terminal value of a software business, because the asset that drives the terminal value — Network Intelligence Capital — appreciates rather than depreciates.
In conventional DCF, the terminal value assumes that the business reaches a stable growth state where cash flows grow at a constant rate indefinitely. In an agentic network, the terminal value assumption must incorporate the compounding architecture: the platform continues accumulating NIC, the routing model continues improving, and trust capital continues growing, driving secular improvement in execution quality and settlement rates that is not fully captured by a simple constant-growth terminal value.
Where Trust_ceiling is the asymptotic trust score achievable in the domain (typically 0.92–0.96 for well-governed domains), g is the long-run GIV growth rate (anchored to the growth of intelligent decision-making as a fraction of economic activity), and the Learning Rate reduces the effective discount rate to reflect the continuing improvement in execution quality.
The Law of Intelligence Compounding states that in a governed agentic network, the value of intelligence assets appreciates through use rather than depreciating. This is the foundational economic law of the Agentic Economy and the most important distinction from every previous asset class.
Every physical asset depreciates: a machine that has been operated for five years is worth less than a new machine. Every software product depreciates through obsolescence. Every dataset depreciates through staleness. The universality of depreciation is so deeply embedded in accounting and economic thinking that it is rarely made explicit — it is simply assumed.
The assumption is wrong for governed intelligence assets.
An intelligence asset that has executed one million verified decisions in a specific domain has accumulated: ten million annotated training signals that improve its future accuracy; a verified trust score built over thousands of enterprise-grade executions that competitors cannot replicate; contextual memory encoding the specific characteristics of the problems it has solved; a routing priority in the discovery layer that gives it first-mover advantage on new executions in its domain; and a settlement premium that reflects the market's willingness to pay more for verified quality than for unverified alternatives. The sum of these advantages is worth more than a freshly created asset with zero execution history, not less.
This equation holds as long as Learning + Verification + Reuse > Obsolescence + Entropy. Maintaining this condition is the primary governance obligation of the platform and the creator. Assets that are not maintained — not updated as domain knowledge evolves, not submitted for re-verification as standards are updated — will eventually depreciate as obsolescence accumulates. The appreciation dynamic requires active stewardship.
The Law of Recursive Learning states that in an agentic network, the output of each period's executions becomes the input to the next period's intelligence production. The network's productive capacity is not a function of the resources deployed this period — it is a function of the accumulated intelligence of all previous periods.
This is the law that makes agentic networks infrastructure businesses rather than service businesses. A consulting firm that delivers excellent advice this year does not thereby deliver better advice next year unless it consciously invests in capturing and preserving the knowledge produced. An agentic network that executes excellent outputs this year does deliver better outputs next year as a mechanical consequence of the Compounding layer's operation.
The Learning Rate is platform-specific and is one of the most important metrics in the AEVS framework. A platform with a Learning Rate of 0.15 is improving its effective intelligence stock by 15% per year from execution activity alone, before new asset creation is counted. This recursive improvement is the source of the terminal value discount discussed in Chapter 21.
The Law of Intelligence Gravity states that networks with higher Network Intelligence Capital exert disproportionate attraction on new intelligence creators and new enterprise users — creating a concentration dynamic analogous to gravitational attraction in physical systems.
The mechanism is straightforward. A creator choosing where to publish a new intelligence asset will choose the platform that offers the largest addressable execution market for that asset — the platform with the most enterprise users executing in the relevant domain. Enterprise users choosing where to execute will choose the platform with the most verified intelligence supply — the highest QIAC in their domain. Both creator and enterprise behaviour directs them toward the platform with the highest NIC.
The squared connectivity term captures the super-linear nature of the attraction: each new creator adds not just their own assets but the connections between their assets and every previously existing asset. Each new enterprise adds not just its own execution demand but its contribution to the routing model calibration that improves discovery quality for all other enterprises. The gravitational effect compounds non-linearly.
The competitive moat of an agentic platform is structurally different from the moats of software, marketplace, or financial services businesses. It cannot be acquired through capital expenditure. It cannot be replicated through engineering talent. It can only be built through time, execution, and consistently high-quality governance.
The Four Moat Dimensions
Memory Moat
The depth of the platform's historical execution record. A platform with ten million annotated executions has a routing model calibrated to ten million data points. A new entrant starts at zero. The memory moat is non-transferable: no amount of investment can replicate ten million verified executions without performing them.
Trust Moat
The aggregate trust capital accumulated in the platform's asset portfolio. High trust scores take years to build — they require thousands of verified executions per asset, each adding fractionally to the score. An asset with a trust score of 0.93 has years of verified performance embedded in that number. A competitor launching with new assets starts with trust scores in the 0.40–0.60 range and cannot close the gap without performing the executions that raise them.
Reuse Moat
The density of execution-graph connections between assets on the platform. As workflow assets are built that chain multiple knowledge assets, knowledge assets that are heavily referenced by workflow assets become structurally embedded in the platform's execution architecture. Switching the platform would require rebuilding not just the asset but all the workflow integrations that reference it.
Connectivity Moat
The number of creators and enterprises whose economic activity is interdependent through the platform's royalty and settlement architecture. A creator whose assets contribute to dozens of workflow assets on the platform has an economic stake in the platform's success that goes beyond their own direct settlement income. Switching would forfeit not just their own royalties but the contribution income from every workflow that references their assets.
LexAgent Inc. is a hypothetical agentic platform specialising in legal and compliance intelligence. It was founded four years ago and has spent the first two years building its creator network and publishing its initial asset library. In years three and four it has been in commercial production, executing against enterprise clients in financial services, healthcare, and corporate legal departments.
The following financial and operational data is provided for the trailing twelve months (TTM). This worked example demonstrates the complete AEVS valuation methodology applied step by step to a concrete set of inputs.
LexAgent Operating Data (TTM)
| Metric | Value |
|---|---|
| Total verified assets (IAC) | 2,847 |
| Quality-adjusted assets (QIAC) | 2,214 (ratio: 0.78 — above 0.70 target) |
| New assets published in TTM (ACV) | 312 (11% of IAC — above 5% target) |
| Total executions (EV) | 4,200,000 |
| Average economic value per execution (AEV) | $38.00 |
| Gross Intelligence Value (GIV) | $159,600,000 |
| Platform capture rate | 9.5% |
| Platform Intelligence Capture (PIC) | $15,162,000 |
| Creator settlement distributed | $144,438,000 (90.5% of GIV) |
| Total discovery events (IDV) | 5,600,000 |
| Discovery efficiency rate (DER) | 0.75 (at target threshold) |
| Discovery liquidity (DL) | 0.68 (above 0.60 target) |
| Trust Coverage Ratio (TCR) | 0.91 (above 0.90 target) |
| Intelligence Inflation Rate (IIR) | 0.12 (below 0.15 target) |
LexAgent Intelligence Balance Sheet (TTM)
| Intelligence Asset Category | Calculated Value |
|---|---|
| Memory Capital (MC) | $28,400,000 |
| Trust Capital (TC) | $41,200,000 |
| Reputation Capital (RC) | $8,600,000 |
| Agent Capital (AC) | $112,500,000 |
| Discovery Capital (DC) | $19,300,000 |
| Total Intelligence Capital (TIC) | $210,000,000 |
LexAgent Network Health Metrics
| Health Metric | Value · Assessment |
|---|---|
| Intelligence Velocity (IV) | GIV / QIAC = $159.6M / $2,214 = $72,100 per QIAC unit · Healthy |
| Intelligence Compounding Rate (ICR) | ΔNIC / GIV = $31.5M / $159.6M = 0.197 · Above 0.15 target |
| Intelligence Moat Index (IMI) | Memory(0.71) × Trust(0.91) × Reuse(0.62) × Connect(0.58) = 0.234 · Developing |
| Intelligence Inflation Rate (IIR) | 0.12 · Within healthy range, monitoring required |
With LexAgent's operating metrics established, the AEVS valuation proceeds through the five components of the Enterprise Value formula in sequence. Each step is computed from the metrics above using the methodologies defined in Part IV.
Step 1: Infrastructure Value
LexAgent has TTM PIC of $15.16 million and is in the Growth stage ($100M–$1B GIV). The applicable Infrastructure Multiple range is 18–28× PIC. LexAgent's DER of 0.75 (at target threshold, not above it), combined with its strong Trust Coverage Ratio and healthy IIR, supports a multiple toward the midpoint of the range.
Infrastructure Multiple selected: 22×
Step 2: Network Intelligence Capital Value
LexAgent's Total Intelligence Capital is $210 million. Its ICR of 0.197 and IMI of 0.234 place it in the 'Strong compounding, developing moat' band, supporting a Knowledge Multiple of 3.5–4.5× NIC. The IMI below 0.40 on the moat dimension caps the multiple below the upper band.
Knowledge Multiple selected: 3.8×
Step 3: Gross Intelligence Value Contribution
LexAgent's TTM GIV is $159.6 million. GIV growth over the prior year was 87% (from $85.3 million). This places LexAgent in the 50–100% GIV growth band, with an applicable Execution Multiple of 10–16×.
Trust adjustment: LexAgent's current average trust score across active assets is 0.79. The domain ceiling for legal and compliance intelligence is 0.93. Projected trust score trajectory over the next three years is 0.79 → 0.85 → 0.90 → 0.93. The Trust Multiplier applied to forward GIV projections is 1.08 for year one (reflecting 0.85/0.79), 1.06 for year two, and 1.03 for year three.
Applying a 13× multiple to trust-adjusted annualised GIV:
Step 4: Exchange Value
LexAgent's daily settlement volume is approximately $437,000 (annual settlement of $159.6M ÷ 365). This is below the $1 million daily volume threshold for Exchange Value to become material. Exchange Value is therefore set to zero for this valuation. It will be revisited when the platform crosses the threshold.
Exchange Value: $0 (below materiality threshold)
Step 5: Platform Optionality
LexAgent's current infrastructure supports three identifiable near-term optionality paths: expansion into regulatory compliance for pharmaceutical approvals (adjacent domain, infrastructure-ready, estimated 40% probability of successful market entry within three years); litigation intelligence for international arbitration (addressable market $2.1B, infrastructure partially ready, 35% probability); and legal intelligence licensing to large language model providers (emerging market, 25% probability).
Platform Optionality Index calculation:
AEVS Enterprise Value — Total
| EV Component | Value |
|---|---|
| Infrastructure Value | $333,564,000 |
| NIC Value | $798,000,000 |
| GIV Value (trust-adjusted) | $2,240,784,000 |
| Exchange Value | $0 |
| Platform Optionality | $615,500,000 |
| Total AEVS Enterprise Value | $3,987,848,000 (~$4.0 billion) |
Applying conventional technology valuation methodologies to LexAgent's metrics produces materially different results, illustrating precisely why the AEVS framework is necessary.
Conventional ARR Multiple
LexAgent's platform revenue (PIC) of $15.16 million is the closest proxy for ARR in a conventional model. At an 87% growth rate, a generous software multiple of 25× ARR produces an enterprise value of $379 million — less than one tenth of the AEVS valuation.
The conventional model is not wrong about what it measures. It is measuring the wrong thing. PIC is the platform's operating revenue; it is not the economic value flowing through the platform. A toll road valued at 25× its toll revenue would be systematically undervalued if the toll is set at 1% of the goods transported. LexAgent's capture rate of 9.5% means that for every dollar of PIC, approximately $9.50 of economic value is being created and distributed. The conventional model prices only the toll; the AEVS prices the traffic.
Conventional Revenue Multiple on GIV
If GIV of $159.6 million is treated as revenue and a 15× multiple is applied (reflecting high growth, high margins), the enterprise value is $2.4 billion. This is closer to the AEVS valuation but still misses the NIC value ($798M) and the optionality value ($615M), and fails to incorporate the trust trajectory adjustment. The GIV-based model captures the execution layer but remains blind to the accumulated intelligence capital that makes future GIV growth achievable.
Summary Comparison
| Valuation Method | Enterprise Value |
|---|---|
| Conventional ARR × 25× | $379,000,000 |
| GIV × 15× (no intelligence capital) | $2,394,000,000 |
| AEVS Full Model | $3,987,848,000 |
| AEVS vs Conventional ARR | 10.5× higher |
| AEVS vs GIV multiple | 1.67× higher (NIC + optionality delta) |
The gap between the conventional ARR model and the AEVS model ($3.6 billion) is not speculative premium. It is the measured value of LexAgent's intelligence capital assets ($210 million NIC × 3.8× multiple = $798M), the trust-adjusted forward execution value above the base GIV multiple (approximately $550M), and the optionality embedded in the platform's infrastructure position ($615M). Each of these components is grounded in observable, auditable metrics. None of them appear in GAAP accounting.
The AEVS valuation is sensitive to three inputs in particular: the Trust Multiplier, the Knowledge Multiple, and the Infrastructure Multiple. The following sensitivity table holds all other inputs constant and varies each of these parameters across their applicable ranges.
Sensitivity to Knowledge Multiple (NIC Multiple)
| Knowledge Multiple | NIC Value Contribution → Total EV |
|---|---|
| 2.0× (weak compounding) | $420M → $2.6B total EV |
| 3.0× | $630M → $2.8B total EV |
| 3.8× (base case) | $798M → $4.0B total EV |
| 5.0× (strong compounding) | $1,050M → $4.3B total EV |
| 7.0× (exceptional compounding) | $1,470M → $4.7B total EV |
Sensitivity to GIV Growth Rate
| GIV Growth Rate | Execution Multiple → GIV Value → Total EV |
|---|---|
| 40% YoY (deceleration) | 8× → $1.38B → $2.7B total EV |
| 87% YoY (base case) | 13× → $2.24B → $4.0B total EV |
| 120% YoY (acceleration) | 18× → $3.10B → $4.9B total EV |
| 200%+ YoY (breakout) | 28× → $4.83B → $6.6B total EV |
The sensitivity analysis reveals that the AEVS valuation is most sensitive to the GIV growth rate — the pace of execution volume expansion — and least sensitive to the Infrastructure Multiple. This is the correct ordering of sensitivities for an early-stage infrastructure business: the dominant driver of value is the rate at which the execution economy scales, not the current-period capture economics.
Implementing the AEVS balance sheet requires establishing measurement systems for five intelligence asset categories that do not exist in conventional accounting systems. This chapter provides the practical guidance for companies preparing their first AEVS-compliant balance sheet.
What to Measure and How
Memory Capital — Data Requirements
To compute Memory Capital, a company requires: a log of all verified executions with timestamps, domain classifications, and verification scores; an annual domain knowledge retention assessment conducted by the relevant domain council; and a mapping of execution history to current asset portfolio to identify which historical executions remain relevant to active assets. Companies should begin logging execution data at production launch — historical data cannot be reconstructed retrospectively.
Trust Capital — Data Requirements
Trust Capital requires: current verified trust scores for all active assets (from the verification system); a benchmark settlement rate for each asset class in each domain (the floor rate that would be paid for an unverified asset of similar capability); and a mapping of each asset's actual settlement rate to the benchmark, from which the trust premium is derived. Trust Capital computation requires trust scores to have been independently verified — self-reported quality scores are not accepted under AEVS.
Agent Capital — Data Requirements
Agent Capital requires: a verified settlement history for each active asset (trailing twelve months); a domain-specific discount rate (obtainable from the domain council's published rate schedule); and a projected maintenance cost for each asset class. Agent Capital is the most complex component to compute accurately because it requires projections. Companies should use conservative assumptions for trust score trajectories and apply the domain discount rate published by the relevant council rather than self-determined rates.
The AEVS defines five reporting requirements for companies seeking AEVS compliance. These requirements are designed to make AEVS reporting comparable across companies — a necessary condition for the AEVS to function as a standard rather than a self-reported narrative.
Requirement 1: Third-Party Verification of Trust Scores
All trust scores used in AEVS calculations must be verified by the relevant domain validation council or an AEVS-accredited third-party verifier. Self-reported trust scores are not accepted. The verification report must be made available to auditors.
Requirement 2: Execution Log Integrity
The execution log underlying Memory Capital and Trust Capital calculations must be maintained in a tamper-evident system that preserves the full execution history without modification. The integrity of the log must be confirmed by the auditor as part of the annual audit process.
Requirement 3: Benchmark Rate Disclosure
The domain-specific benchmark settlement rates used to compute Trust Capital premiums must be disclosed and sourced from the relevant domain council's published rate schedule. Companies cannot set their own benchmark rates.
Requirement 4: AEVS Reconciliation Statement
AEVS-reporting companies must publish an annual AEVS Reconciliation Statement that bridges from conventional GAAP balance sheet to the AEVS balance sheet, showing the value of each intelligence asset category and the methodology used to calculate it.
Requirement 5: NIC Movement Schedule
An annual schedule showing the opening NIC balance, additions from new asset creation, additions from learning and verification, deductions from decay and depreciation, and the closing NIC balance. The movement schedule must be independently audited.
The AEVS can only function as a trustworthy standard if its calculations are independently auditable. The following audit methodology is recommended for first-year adopters.
The audit of an AEVS balance sheet has four stages. First, execution log integrity: the auditor confirms that the execution log is complete, tamper-evident, and accurately reflects the platform's operational history. Second, trust score verification: the auditor confirms that all trust scores have been independently verified by accredited verifiers and that the scores used in calculations match the verified values. Third, benchmark rate confirmation: the auditor confirms that domain benchmark rates are sourced from published council rate schedules. Fourth, computation verification: the auditor independently recomputes the five intelligence asset categories from the underlying data and confirms agreement with the company's reported values within an acceptable tolerance.
The acceptable tolerance for AEVS audit agreement is ±5% for each intelligence asset category. Discrepancies above 5% in any category require restatement. The audit opinion should specify whether it covers all five intelligence asset categories or only those for which the auditor has obtained sufficient appropriate evidence — partial coverage is acceptable in early adoption phases.
Investors applying the AEVS framework need comparable companies against which to calibrate the multiples used in the valuation model. The following framework provides the relevant comparables for each component of the AEVS EV formula.
Infrastructure Value Comparables
For the Infrastructure Multiple, the relevant comparables are neutral intermediary infrastructure businesses: payment networks (Visa, Mastercard, Stripe), cloud infrastructure providers (AWS, Azure, GCP as standalone businesses), and data routing infrastructure (Cloudflare, Akamai). These businesses are characterised by high volumes, low marginal costs, network effects, and revenue that grows with the economic activity flowing through the infrastructure rather than with the number of subscriptions sold.
NIC Value Comparables
For the Knowledge Multiple, the relevant comparables are businesses whose primary asset is accumulated institutional intelligence: financial data providers (Bloomberg, FactSet, MSCI), legal data platforms (Westlaw, LexisNexis as standalone businesses), and specialised knowledge infrastructure (IHS Markit). These businesses trade at significant premiums to their revenue because the market recognises that their accumulated data and analytical infrastructure is worth far more than their current earnings.
GIV Value Comparables
For the Execution Multiple, the relevant comparables are high-growth marketplace and infrastructure businesses at comparable stages: marketplace platforms (Airbnb, DoorDash) for the exchange structure, financial exchanges (Nasdaq, NYSE Euronext) for the settlement architecture, and professional services automation platforms for the domain intelligence component. GIV multiples should be calibrated specifically to growth rate, as shown in the multiple table in Chapter 18.
The AEVS framework was developed primarily for the valuation of private and public companies operating agentic networks. Its implications, however, extend beyond corporate finance. At sufficient scale, agentic infrastructure becomes national infrastructure — a productive resource whose quality and accessibility affects the competitiveness of the entire economy, not merely the companies operating it.
Roads are valued not by the toll revenue they generate but by the economic activity they enable. Internet infrastructure is valued by the GDP contribution of the companies and citizens it connects. At scale, intelligence infrastructure should be evaluated by the same standard: not only by the settlement revenue flowing through the network, but by the economic activity that becomes possible because the network exists.
This implies that national governments have a legitimate interest in the development of domestic intelligence infrastructure — not as protectionist policy but for the same reasons they have invested in roads, electricity grids, and internet infrastructure. A nation whose healthcare, legal, and regulatory intelligence infrastructure is predominantly foreign-owned is in a position analogous to a nation whose telecommunications infrastructure is foreign-owned: the economic activity it enables benefits the national economy, but the strategic control sits elsewhere.
Sovereign Intelligence Capital
The AEVS framework introduces the concept of Sovereign Intelligence Capital: the aggregate NIC of the intelligence infrastructure operating within a national jurisdiction, weighted by the degree of domestic ownership and governance. Sovereign Intelligence Capital is to the Intelligence Economy what manufacturing capacity was to the Industrial Economy — a strategic asset that governments will increasingly track, develop, and protect.
Strategic Coverage measures the fraction of economically critical domains (healthcare, legal, financial, government administration, critical infrastructure) that have verified domestic intelligence asset supply meeting quality thresholds. A country with excellent domestic legal intelligence but no domestic healthcare intelligence has a partial strategic position that is vulnerable to disruption in its uncovered domains.
The AEVS provides the accounting foundation for an Intelligence Capital Market — a set of financial markets in which intelligence assets, and claims on the yield of intelligence assets, can be bought, sold, and invested in by institutional and retail investors.
The preconditions for a functioning capital market are: standardised measurement of the underlying asset (addressed by the AEVS), independent verification of asset quality (addressed by the domain validation council system), price discovery mechanisms (addressed by the exchange infrastructure), and settlement of financial claims (addressed by the DAAC settlement layer). As these preconditions are met, the Intelligence Capital Market will develop the full range of financial instruments available in mature capital markets.
The Emerging Instrument Stack
Intelligence Bonds: fixed-income instruments backed by the settlement streams of verified intelligence asset portfolios, rated by intelligence rating agencies applying AEVS-compliant standards. Intelligence ETFs: index funds providing diversified exposure to intelligence execution yield across domains. Intelligence venture funds: early-stage investment vehicles that fund genesis of new intelligence assets in under-served domains in exchange for royalty participation rights. Intelligence REITs: structured vehicles that hold portfolios of mature intelligence assets and distribute execution yield to shareholders.
Each of these instruments requires the AEVS as its accounting foundation. Without a standardised, auditable way to measure the value of intelligence assets, none of them can be priced, rated, or traded with the confidence that institutional investors require.
The Agentic Economy Valuation Standard is proposed as a living framework — not a fixed document but a set of principles and methodologies that will evolve as the Intelligence Economy matures and as the empirical evidence on agentic network economics accumulates. The following ten constitutional principles govern that evolution.
Every civilisation has been ultimately governed by its accounting system. Agricultural societies developed land registries. Industrial societies developed corporate accounting. Financial societies developed mark-to-market standards and derivative pricing. Each accounting innovation did not merely describe the economy that produced it — it shaped the economy that followed, by determining what could be owned, traded, and invested in.
The Intelligence Economy will be no different. The most consequential financial innovation of the coming decade may not be a new investment product, a new payment mechanism, or a new financial instrument. It may be the development of accounting standards that make Intelligence Capital visible, measurable, and investable for the first time.
That is what the AEVS proposes. Not a product. Not a company. Not a protocol. A language — precise, auditable, reproducible — for measuring the most important new productive asset class in economic history.
The next century may be remembered not for the invention of artificial intelligence — but for the moment humanity learned how to account for it.