What is an AI component
What is an AI component
An AI component is the specific part of a project where real AI work happens — analysis, generation, prediction, ranking, or decision-making. It is not a marketing label or a brand mention. In DecAIHub project cards, the AI component field answers one question: what exactly does the AI do here?
In short: an AI component is the concrete mechanism where data goes in and an AI-driven result comes out — not a logo, a tagline, or a roadmap promise.
Analogy for beginners
Think of a project like a restaurant. The use case is the menu — what the customer sees and orders. The AI component is the kitchen — the actual equipment and technique that turns raw ingredients into a dish. A restaurant can have a beautiful menu (use case) but no working kitchen (AI component). When DecAIHub reviews a project, it walks into the kitchen to check whether the stove is real, whether anyone is cooking, and whether there are actual recipes — not just photos of food printed on a sign outside.
What it is
An AI component is a functional module or subsystem that performs work typically associated with artificial intelligence or machine learning. This includes:
- Inference engines — running a trained model to produce outputs (predictions, classifications, generated text or images)
- Training pipelines — updating model weights on new data
- Analytics models — extracting patterns, scores, or signals from data (market analytics, risk scoring, sentiment analysis)
- Agent frameworks — orchestrating autonomous decision-making, tool use, and multi-step reasoning
- Data labeling systems — managing human or automated annotation to build datasets
- Compute orchestration — distributing AI workloads across nodes or GPUs
The key distinction: a project has an AI component if it contains a subsystem that processes data using models, algorithms, or learned representations. A project lacks an AI component if "AI" appears only in its name, marketing copy, or future plans without any verifiable technical implementation.
An AI component does not need to be custom-built from scratch. A project that fine-tunes an open-source model, wraps an inference API with meaningful pre- and post-processing, or trains its own classifier on proprietary data still has an AI component — as long as the AI layer adds a verifiable function beyond simple pass-through.
The spectrum: from absent to fully verifiable
Not all AI components are equal. They exist on a spectrum of maturity and verifiability:
| Level | Description | What to expect |
|---|---|---|
| Absent | No AI found in the project | The ai_component field reads none or unknown |
| Positioning only | AI claimed in marketing but no technical evidence | The field reads positioning only; ai_score is 0 |
| Declared | AI is mentioned in official docs or roadmap, but no working artifacts | AI_0 = 1, remaining AI flags are 0 |
| Demonstrated | A working demo, prototype, or documented pipeline exists | AI_1 and AI_2 are 1; partial evidence |
| Verified | Benchmarks, technical documentation, model versions, and reproducible results are available | AI_3 and AI_4 are 1; strong Tier-1 evidence |
This spectrum matters because "having AI" is not binary in practice. A project may have a trained model but no benchmarks, or a working demo but no published architecture. The AI component field captures what exists, not what could exist in the future.
Common AI component values in DecAIHub
Here are real values from project cards in the registry:
AI agent platform (workflows/tools/KB)— agent infrastructure with tools and knowledge baseAI data labeling + model evaluation (human experts)— labeling with a human-in-the-loop pipelineAI/ML crypto market analytics (app)— machine learning models applied to market datacompute orchestration— routing and scheduling AI compute tasksAI trading signals— models generating trading-related signals from dataagentic framework (memory/actions/plugins)— an agent runtime with memory and tool integrationsAI summaries + crypto intelligence terminal— NLP-based summarization and data analysisinference engine— a system dedicated to running model inference
Each value describes what the AI actually does — not what the project promises to become. These values are drawn from the normalized registry of repeating fields to ensure consistency across project cards. When a reviewer fills in the AI component field, they choose from observed values or add a new one that precisely captures the technical mechanism.
Where it appears in DecAIHub project cards
The AI component (ai_component) field is part of the Project Passport table. It sits alongside the Use case field. Together, they separate the technical mechanism from the product description.
If the field reads positioning only, none, or unknown, it means evidence of a working AI component has not been confirmed as of the verification date. These are not technologies — they are evidence-reading modes.
The AI component field is one of the first things an experienced reader checks. It gives an immediate signal: does this project do something with AI, or does it only say it does? Combined with the Evidence table and the verification date, it paints a fast picture of the project's technical substance.
How AI components are verified
Verification follows the Tier-1/2/3 evidence hierarchy. For the AI component to move beyond positioning only, reviewers look for:
- Tier-1 artifacts — source code in public repositories, model weights or architecture descriptions, benchmark reports with methodology, API documentation showing inference endpoints, published technical papers by the team.
- Tier-2 context — third-party reviews mentioning the AI pipeline, ecosystem integrations that depend on the AI output, aggregator descriptions that reference specific AI functionality.
- Tier-3 signals — marketing claims, influencer mentions, press releases without technical detail. These are noted but do not raise the AI score on their own.
The review also checks for common anti-patterns: an empty GitHub repository listed as "AI code," benchmarks copied from another project, or an API wrapper presented as a proprietary model. These red flags keep the ai_score at 0 until verifiable evidence appears.
What an AI component means for DecAI Fit
The AI component directly feeds the AI reality axis of the DecAI Fit score. Without Tier-1 evidence of a working AI component, a project's ai_score is 0 — and its overall DecAI Fit cannot rise above 0.
This is the strictest rule in the scoring system: no AI reality, no DecAI Fit above 0. Even if a project has strong on-chain infrastructure and a well-designed token, it cannot score above zero without a verifiable AI component. This reflects DecAIHub's core thesis — the registry evaluates alignment with the "Decentralized AI" thesis, and AI is the non-negotiable starting point.
This does not mean the project is bad or worthless. It means the evidence was not available at the time of review. When verifiable artifacts (code, benchmarks, demos, technical documentation) appear, the card is updated.
What this term alone does not prove
- Having an AI component does not prove the product is useful, the token is necessary, or the on-chain integration is meaningful.
- The AI component field describes only the mechanism. Quality, maturity, and scale are assessed separately through the Evidence table and the AI reality flags.
- A project can have a real AI component that is trivial — for example, a simple sentiment classifier calling an external API. Real does not mean significant.
- Conversely, a project with a sophisticated AI component may still score low on DecAI Fit if its on-chain integration is superficial or its token serves no function.
Do not confuse with
- Use case describes what the user gets from the product. AI component describes what the AI does inside. See AI component vs use case.
- Category (e.g.
AI/Agents) classifies the project at a higher level. The AI component is more specific: the actual technology. - Automation is not always AI. Simple rule-based scripts or if/then workflows are not AI components. See Automation vs real AI component.
- AI model refers to the trained artifact (weights, architecture). The AI component is broader: it includes the model plus the pipeline around it (data ingestion, pre-processing, inference serving, post-processing).
- API wrapper — a project that only forwards requests to a third-party AI API (such as a hosted LLM) without adding meaningful processing. This may or may not qualify as an AI component depending on the level of integration and added value.
Frequently asked questions
What is the difference between an AI component and a use case?
The AI component is the technical mechanism — what the AI does (e.g., runs inference, trains models, scores data). The use case is the product outcome — what the user gets (e.g., DeFi copilot, portfolio tracker, risk alerts). A project can have the same AI component but different use cases, or the same use case powered by entirely different AI components. In DecAIHub project cards, these are deliberately separate fields so readers can evaluate the technology and the product independently. See AI component vs use case for a detailed comparison.
Can a project have AI without an AI component?
No — in DecAIHub terminology, if a project has AI, it has an AI component. The question is whether that component is verifiable. A project might genuinely use AI internally but provide no public evidence: no code, no documentation, no benchmarks, no demos. In that case the AI component field will read unknown and the ai_score remains 0. This does not mean the AI is absent — it means evidence is absent. The card is updated when evidence appears.
What does "positioning only" mean?
It means the project uses AI language in its marketing, branding, or positioning — but no Tier-1 or Tier-2 evidence of a working AI component has been found. The word "AI" may appear in the project name, website tagline, or pitch deck, yet there is no code repository, no model documentation, no demo, and no technical description of what the AI actually does. "Positioning only" is a factual observation, not a judgment. Some projects in this state may be pre-launch with AI components in development; others may never build one.
Does a higher ai_score mean a better project?
No. The ai_score measures verifiability of the AI component, not the quality of the product or the usefulness of the technology. A project with ai_score = 3 has strong evidence that its AI is real and documented. A project with ai_score = 1 has limited evidence. Neither score says anything about whether the product solves a real problem, whether users need it, or whether the token model makes sense. DecAI Fit combines four dimensions — AI reality, Onchain necessity, Token necessity, and Evidence — to give a more complete picture.
External references
- Machine Learning Glossary — Google Developers — reference definitions for ML and AI terminology used across the industry.
- The promises and challenges of crypto + AI applications — Vitalik Buterin — a framework for thinking about where AI and blockchain genuinely complement each other.
- Stanford HAI — AI Index Report — annual data on the state of artificial intelligence, including adoption, research, and technical progress.
No comments to display
No comments to display