Verification Workspace

Review surface for checking extracted outputs against source context and surrounding evidence. When you extract code, proofs, summaries, or structured data, the Verification Workspace lets you confirm that the extraction is accurate, complete, and faithful to the original.

---

Core Capabilities

Verify	Trace	Inspect
Check extracted outputs against original source material for accuracy.	Follow any claim back through the extraction chain to its origin.	Side-by-side context view of outputs alongside their source evidence.

---

Features

Evidence Panel

Side-by-side view showing the extracted output alongside its original source context. Read the extraction and the source simultaneously — see exactly what was captured and what was left out.

Learn more about Evidence Panel →

Source Trace

Follow the provenance chain for any extracted output — from the published collection, through the extraction process, to the original document or conversation message. Every link in the chain is navigable.

Validation Notes

Annotate extracted outputs with verification status, correctness notes, and reviewer comments. Mark outputs as verified, disputed, or needs-review to maintain quality across your knowledge base.

Review State

Track the verification status of your corpus — how many outputs have been verified, what percentage is still pending review, and which sources have the highest dispute rates. Quality dashboard for your knowledge.

---

The Problem We Solve

"I don't trust my AI-extracted outputs"

AI extraction is powerful but imperfect. The Verification Workspace provides the human review layer — evidence panels, source traces, and validation workflows — to ensure extracted outputs are accurate.

"Knowledge drift is invisible"

When extracted outputs are modified, summarised, or restructured over time, they can drift from the original meaning. Source Trace and Evidence Panel make drift visible and correctable.

"I can't cite my own knowledge base"

Published claims need verifiable sources. The Verification Workspace ensures that every claim in your published collections can be traced back to an inspectable, contextualised original source.

---

Verification Workflow

1. Queue — Outputs from the Extraction Studio enter the verification queue
2. Compare — Evidence Panel shows extracted output alongside original source
3. Validate — Reviewer confirms accuracy, completeness, and faithfulness
4. Annotate — Validation Notes record reviewer assessment and any corrections
5. Resolve — Disputes are resolved by returning to source material
6. Publish — Verified outputs are cleared for publication through the Publishing Layer

---

Quality Metrics

Metric	What It Measures
Verification Rate	Percentage of extracted outputs that have been human-reviewed
Accuracy Score	Proportion of verified outputs marked as accurate
Drift Index	How much extracted outputs have changed from original sources
Dispute Rate	Frequency of verification disagreements across reviewers

---

Integration with Other Products

• Corpus Workspace — Source material for verification comes from the corpus
• Conversation Parser — Parsed conversations provide original context for AI-extracted outputs
• Knowledge Graph — Graph provenance supports verification of relationships and connections
• Extraction Studio — Extracted outputs flow into the verification queue
• Publishing Layer — Only verified outputs are cleared for publication

---

Related Solutions

• Trace Provenance — Follow any claim back to its original source
• Extract Knowledge — Verify the accuracy of extracted outputs
• Organize Corpus — Maintain corpus quality through systematic verification

Related Use Cases

• For Researchers — Verify research findings against original sources
• For Developers — Confirm extracted code accuracy before reuse
• For Writers — Ensure published claims are supported by verifiable evidence

---

→ Explore All Products