> ## Documentation Index > Fetch the complete documentation index at: https://vaquill.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Analysis steps > Run a matrix over a document set, diff two documents, check compliance gaps, score against benchmarks, or reason across a whole set Five steps. This is where a workflow stops collecting material and starts producing judgments. *** ## Build and Run Matrix AI Slow Applies a matrix template to a set of documents and extracts every cell. Rows are documents, columns are the template's questions. This is the same engine behind [Document Matrix](/docs/guides/document-matrix) and behind most of the built-in review workflows. **Config** | Field | Type | Required | Default | Notes | | --------------- | --------------- | -------- | ----------------- | ------------------------------------------------------------- | | Matrix template | Template picker | Yes | | One of yours or a built-in template. | | Documents | Document list | Yes | | 1 to 200. One row per document. | | Title | Text | No | "Workflow Matrix" | How the matrix appears in the Matrix list and the run viewer. | **Outputs** | Field | Contents | | ---------------------------- | ------------------------------------------------------------------------------------------------------------------ | | Matrix Id | Open the Matrix surface to drill into individual cells. | | Rows, Columns | Matrix dimensions. | | Extracted, Not Found, Errors | Per-cell result counts. | | Truncated | True when polling hit the time cap before every cell finished. The matrix keeps filling; the step stopped waiting. | | Answers By Question By Doc | Per-document answers as `{question: {filename: answer}}`. **Wire into Cross-Doc Synthesis.** | | Findings By Filename | Per-document findings in the shape Verify Synthesis expects. **Wire into Verify Synthesis.** | The two dict outputs exist specifically so a matrix can feed the synthesis and verification steps without you reshaping anything. Wiring **Answers By Question By Doc** into Cross-Doc Synthesis, and **Findings By Filename** into Verify Synthesis, is the canonical portfolio-review chain. *** ## Compare Two Documents Deterministic Medium Redline-style diff between two documents. **Config** | Field | Type | Required | Notes | | ----------------- | --------------- | -------- | -------------------------- | | Original document | Document picker | Yes | The baseline. | | Revised document | Document picker | Yes | The proposed version. | | Title | Text | No | Defaults to the filenames. | **Outputs** | Field | Contents | | --------------------- | --------------------------------------------------------------- | | Comparison Id, Status | The created comparison. | | Url Path | Relative path to open the comparison viewer and accept changes. | *** ## Compliance Gap Check AI Slow Scores one document against a regulatory regime and surfaces per-requirement gaps. **Config** | Field | Type | Required | Notes | | ----------------- | --------------- | -------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Document | Document picker | Yes | | | Regulation | Select | Yes | GDPR, CCPA, DPDP, HIPAA, SOX, UK GDPR, LGPD, PCI-DSS, DORA, NIS2, SOC 2. | | Document category | Select | Yes | DPA, privacy policy, terms of service, breach plan, consent form, DSR process, vendor agreement, information security policy, incident response plan, financial report, and others. | | Context | Text | No | Industry, jurisdiction, or data-type context to sharpen the check. Up to 2,000 characters. | **Outputs** | Field | Contents | | -------------- | ---------------------------------------------------------------------- | | Analysis Id | The stored analysis. | | Overall Status | The headline verdict. | | Requirements | Per-requirement results. | | Gap Count | How many requirements came back as gaps. Useful as a branch condition. | *** ## Score Against Benchmarks Deterministic Fast Scores extracted contract facts against market-standard benchmarks. No model call, so it is fast and repeatable. **Config** | Field | Type | Required | Notes | | --------------- | ----------------------------------- | -------- | -------------------------------------------------------------------------------------------------- | | Benchmark group | NDA, MSA, Employment, M\&A, Privacy | Yes | Each suite scores a different contract family. | | Facts | Dict | Yes | Wire from an upstream Extract Contract Facts step. Facts the suite does not recognize are ignored. | **Outputs** | Field | Contents | | --------------------------------- | ---------------------------------------------- | | Overall Tier | The headline risk tier. Good branch condition. | | Overall Score | Numeric score. | | Benchmark Scores | Per-benchmark detail. | | Flagged Benchmarks, Flagged Count | The redline-ready list. | *** ## Cross-Doc Synthesis AI Slow Answers comparison and aggregation questions across multiple documents: "which is worst", "where do they disagree", "which ones are missing this clause". **Config** | Field | Type | Required | Notes | | -------------------- | ------------- | -------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Documents | Document list | Yes | **2 to 50.** Order does not matter. | | Questions | String list | Yes | 1 to 10 cross-document questions. | | Per-document answers | Dict | No | An optional starting point. Wire Answers By Question By Doc from an upstream Build and Run Matrix so the synthesizer starts primed rather than reading everything cold. | **Outputs** | Field | Contents | | ------------------------------ | -------------------------------- | | Answers | One answer per question. | | Document Count, Question Count | What was actually reasoned over. | This is the highest-hallucination-risk step in a portfolio chain, because it reasons over summaries rather than raw documents. Always follow it with [Verify Synthesis](/docs/workflows/builder/steps/verification), which drops findings that reference documents outside the input set. *** ## Related The pass that should follow every synthesis. Turning a matrix or a synthesis into a deliverable.