Agent Readiness Index (ARI): Tech 100
Summary
As AI systems evolve from interpreting content to executing actions, agent-readiness becomes the new frontier of visibility.
The Agent Readiness Index (ARI) evaluates ~100 leading technology companies using a weighted, signal-based audit of machine-readable infrastructure, including tool registries, APIs, plugin manifests, and dataset exposure.
Despite leading the AI narrative, many companies remain inaccessible to AI agents, revealing a structural gap between innovation and implementation.
This report extends the Entity Clarity Index (ECI) into the agent execution layer, highlighting which companies are positioned to be used — not just understood — by AI systems.
Methodology
The Agent Readiness Index (ARI) evaluates publicly observable, machine-readable infrastructure.
Scores are derived from signals that can be accessed, validated, and interpreted by external AI agents through open or discoverable endpoints.
This means ARI reflects an organization’s public agent surface, not its full internal AI capability.
In many cases, companies may possess advanced internal systems, proprietary infrastructure, or private agent frameworks that are not exposed externally. These capabilities are not captured in ARI unless they are presented in a form that can be discovered and used by external agents.
For example, NVIDIA demonstrates leadership in AI infrastructure through platforms such as NVIDIA NIM (Inference Microservices). However, if these systems are not exposed through publicly accessible, machine-readable interfaces (such as registries, manifests, or standardized endpoints), they will not materially impact ARI scores.
This distinction is intentional.
ARI is designed to measure whether an organization is:
- discoverable by agents
- interpretable by agents
- usable by agents in open or semi-open environments
—not whether it has built advanced AI systems internally.
As AI agents increasingly mediate discovery, integration, and execution across the web, the public availability of machine-readable infrastructure becomes a critical layer of visibility and participation.
Findings
1. No true public leaders have emerged yet
The highest observed ARI score in the dataset was just 22, earned by Fortinet. Robinhood followed at 20, while IBM, Infineon, and Datadog each scored 19. Even the strongest publicly observable companies remain below true readiness, confirming that the field is still early.
2. Reachability does not equal readiness
Many companies were reachable but still scored poorly. Google, Amazon, Intel, Qualcomm, Snowflake, and Palo Alto Networks each scored 10, while Apple, Meta, CrowdStrike, ADP, Nintendo, and Workday landed at 6 or below. Public presence alone does not translate into usable agent infrastructure.
3. A meaningful mid-tier cluster is beginning to form
A group of firms including Microsoft, Netflix, Palantir, AppLovin, Texas Instruments, Booking.com, Arista Networks, Spotify, Cloudflare, and Monolithic Power Systems clustered around 16, suggesting an emerging class of partially machine-legible companies with some infrastructure in place but clear gaps remaining.
4. Zero-score exposure is more common than expected
A notable set of reachable companies scored 0, including SK Hynix, Applied Materials, Uber, MercadoLibre, DoorDash, Tokyo Electron, Synopsys, Coinbase, Seagate, NXP Semiconductors, and Roper Technologies. This highlights a key structural issue: many firms are publicly accessible on the web but effectively invisible to agents.
5. Some of the most strategically important firms remain blocked or unobserved
NVIDIA, TSMC, Oracle, SAP, AMD, Micron, Adobe, Salesforce, Broadcom, Tesla, and ASML were among the companies that could not be meaningfully observed through the public audit flow. In several cases, badges suggested latent infrastructure signals such as MCP or Docs, but these did not translate into usable public agent surfaces.
Landscape
The Technology sector presents a paradox:
companies leading the AI revolution are often not accessible to AI agents themselves.
While many organizations invest heavily in AI models, far fewer have implemented the machine-readable infrastructure required for agent interaction.
This creates a growing divide between:
- AI Leaders (Narrative Layer)
- Agent-Ready Entities (Execution Layer)
The ARI Tech 100 reveals that agent readiness is not correlated with market leadership, but instead with infrastructure discipline and architectural clarity.
As AI agents become primary interfaces for discovery, decision-making, and execution, this gap will directly influence visibility, integration, and long-term relevance.
Archetypes
Public Frontrunners
These are the strongest publicly observable companies in the dataset, including Fortinet, Robinhood, IBM, Infineon, and Datadog. They are not yet fully ready, but they show the clearest signs of deliberate machine-readable infrastructure.
Partial Infrastructure Builders
This group includes companies like Microsoft, Palantir, Netflix, Cloudflare, Spotify, Booking.com, and Arista Networks that have some visible agent-facing architecture but remain incomplete. They are structurally ahead of the broader field, yet still far from fully agent-ready.
Reachable but Agent-Thin
These companies are publicly reachable but provide little meaningful agent utility. Google, Amazon, Intel, Qualcomm, Snowflake, Apple, Meta, and Palo Alto Networks fall into this category. They are highly visible institutions with shallow public agent surfaces.
Visible but Functionally Invisible
This archetype includes firms that are reachable online yet scored 0, such as Uber, Coinbase, Synopsys, Seagate, NXP, and DoorDash. They are present on the internet but absent in machine-usable form.
Concealed Operators
These are companies that appear blocked or unobserved despite signs of deeper infrastructure intent. TSMC, SAP, Tesla, Sony, Cadence, Marvell, Autodesk, NEC, and Reddit fit this pattern. Some carry badges such as MCP, Docs, or Runtime, suggesting latent capability without dependable public accessibility.
Opaque Giants
These are strategically central institutions whose public agent posture is effectively closed. NVIDIA, Oracle, Broadcom, AMD, Micron, Adobe, Salesforce, and ASML represent some of the most important names in modern technology, yet remain difficult or impossible to observe through public agent-readiness audits.
Index
Strategic Implications
1. The next competitive layer is machine accessibility
The Technology sector is already competing on AI capability, but ARI shows that the next layer of advantage will come from being usable by agents. Visibility alone will not be enough. Institutions will need machine-readable infrastructure that allows agents to discover, interpret, and interact with them directly.
2. Market leadership and agent readiness are diverging
Many of the world’s most powerful technology firms remain weakly scored, blocked, or unobserved. This creates a structural disconnect between who leads the AI narrative and who is actually positioned to participate in the agentic web.
3. Public infrastructure discipline is becoming a form of strategic leverage
The strongest performers in this dataset did not win by brand power alone. They surfaced more clearly because they presented cleaner public machine-readable signals. Over time, this discipline will influence integration opportunities, discovery pathways, and downstream relevance in AI-mediated workflows.
4. Blocked entities may be protecting themselves — but also excluding themselves
For some companies, gating may be intentional. But regardless of motive, the result is the same: reduced public agent legibility. As more software agents mediate research, routing, procurement, and decision support, closed surfaces may carry increasing opportunity cost.
5. The field is still early, which makes the advantage asymmetric
Because the top observed score is only 22, the market has not yet established a mature public standard for agent readiness. That means the leaders of the next phase have not been fully determined. Companies that move early to expose registries, manifests, endpoints, datasets, and documentation can gain disproportionate visibility before the category becomes crowded.
6. ARI expands ECI from perception into execution
Entity Clarity asks whether an institution is understandable to AI systems. Agent Readiness asks whether it is actionable to them. Together, they define a broader institutional question for the AI era: not just whether a company can be seen, but whether it can be used.
Full Report
The Agent Readiness Index (ARI): Tech 100 reveals a structural truth that the market still has not fully internalized: technology leadership does not translate automatically into agent accessibility.
Across the 95 companies provided in this dataset, the dominant pattern is not readiness, but absence. Only 45 companies were reachable and scored, while 50 were blocked or unobserved. That means more than half of this universe did not present a public surface that could be meaningfully evaluated through the ARI audit flow. Even before comparing scores, that alone is the story: the modern technology sector is still largely opaque to agents.
Among the companies that were reachable, readiness remains shallow. The average observed ARI score was just 9.3 out of 100, with a median of 10. The highest score in the dataset was 22, earned by Fortinet, followed by Robinhood at 20, then IBM, Infineon, and Datadog at 19. That is an important result. Even the best-performing companies in this sample remain in the “not ready” band. In other words, the frontier is not crowded. There is no true agent-ready leader yet in this set. There are only firms that are less underbuilt than others.
This is what makes the report strategically significant. We are not observing a mature race with clear winners. We are observing the early phase of a new infrastructure divide.
The second major finding is that blocking behavior is widespread even among the most strategically important firms in the AI and semiconductor stack. The dataset shows NVIDIA, TSMC, Tesla, Oracle, SAP, AMD, Micron, Salesforce, Adobe, Lam Research, Arm, ASML, and Broadcom all as blocked or unobserved. That is not a fringe group. That is the upper spine of modern compute, software, and digital infrastructure. Some of these entities show hints of machine-readable intent through badges such as MCP or Docs, but the public surface still resolves as gated or non-observable in practice. Strategically, that means the sector contains a large layer of latent infrastructure with no dependable public agent interface.
This distinction matters. In several cases, the badges suggest organizations may have pieces of the stack internally or partially exposed. TSMC, Tesla, SAP, Sony, Cadence, Marvell, Autodesk, NEC, and others carry some combination of MCP and documentation signals while still remaining blocked or unobserved. Uber, Equinix, Intuit, and Reddit show runtime-related signals, yet their public-facing posture remains gated or inaccessible. That creates a new archetype within ECI: the concealed operator — an institution that may be building toward machine interoperability, but does not yet present a reliable surface for open agent interaction. In practical terms, these firms may be preparing for an agentic future without yet participating in it publicly.
Among the reachable companies, the score distribution is highly compressed. There is a cluster at 16 points including Microsoft, Netflix, Palantir, AppLovin, Texas Instruments, Booking.com, Arista, Spotify, Cloudflare, and Monolithic Power Systems. There is another cluster at 10 points including Google, Amazon, Intel, Qualcomm, Schneider Electric, Palo Alto Networks, Cambricon, and Snowflake. There are 11 companies scoring 0, including SK Hynix, Applied Materials, Uber, MercadoLibre, DoorDash, Tokyo Electron, Synopsys, Coinbase, Seagate, NXP, and Roper Technologies. That is striking because it shows that being reachable does not guarantee meaningful agent legibility. A company can be online, widely known, and commercially powerful while still presenting effectively no usable agent infrastructure through this audit.
Some of the more interesting signal mismatches come from household AI and internet brands. Microsoft reaches only 16, despite its central role in enterprise AI. Google and Amazon both sit at 10. Meta lands at 6. Apple is also at 6. These are not low-profile firms lacking resources. They are among the most powerful companies in the world. Their weak public ARI posture suggests a broader institutional pattern: the market has invested massively in AI capabilities, but far less consistently in public agent-addressable architecture. That gap between narrative leadership and machine accessibility is exactly where future visibility asymmetry will emerge.
There is also an important geographic and sectoral undertone in the blocked cohort. A large number of semiconductor, hardware, manufacturing, and Asian platform companies appear as unobserved or blocked: TSMC, Samsung, Tencent, Alibaba, AMD, Micron, Foxconn, Advantest, SMIC, MediaTek, Baidu, JD.com, NAURA, Pinduoduo, Meituan, NetEase, and others. This does not mean these organizations lack technical sophistication. It means their publicly inspectable surfaces are not yet aligned with open agent discoverability. Within ECI terms, this is a posture question as much as a capability question. Some firms may be intentionally restrictive. Others may simply not prioritize public machine-readable exposure. But the result is the same: lower agent legibility.
The badge patterns add another layer of nuance. Across the full set, the most common badge state is Unknown, while a substantial minority is explicitly Gated. Only a smaller subset shows affirmative signals such as MCP, Docs, or Runtime. This implies that the market has not yet converged around a standard public signaling language for agent readiness. Said differently, most companies are still accidental participants in the agentic web, not deliberate ones.
From a strategic standpoint, this report points to a coming shift in competitive advantage. Historically, firms optimized for human users, search crawlers, developers, investors, and media. The next layer is optimization for agents that discover, evaluate, route, and act. Once those systems become a meaningful traffic and decision layer, firms without machine-readable public infrastructure will not simply appear behind. They may become structurally excluded from key workflows.
That is why the most important implication of ARI is not “who scored highest.” It is that the entire field is still early. A top score of 22 means this market has not yet produced a clear public standard for agent readiness. This leaves room for fast movers to gain disproportionate advantage, because the bar is still low. The winners in the next phase will likely not be the loudest AI brands, but the companies that quietly build the cleanest public infrastructure: registries, manifests, machine-readable endpoints, documentation, and data surfaces that agents can reliably interpret and use.
Viewed through the broader ECI lens, ARI functions as the execution-side extension of entity clarity. ECI asks whether an institution is clearly legible to AI systems. ARI asks whether that institution is actionable to them. The Tech dataset shows that the answer, for most of the sector, is not yet yes.
The conclusion is simple and sharp: technology companies are racing to build AI, but very few have built themselves to be usable by it. That is the hidden readiness gap. And over time, it will matter far more than most of the market currently realizes.
Read: The AI Legibility Divide: A Cross-Industry Institutional Assessment
Read the related Four Forces of AI Power Framework
Read the Entity Clarity Report - Tech 100 that compliments this ARI report
Read the Lexicon for Agent Readiness Index (ARI)