Evaluation focus
What it does, how it works, and where data lives.
Answer those three questions before the walkthrough.
Proof pack
Review the DecisionTrail evaluation pack before the walkthrough.
Review the workflow, trust model, deployment boundary, and sample output in one evaluation path before a walkthrough.
Automatic by default Background QA for edge cases Privacy-safe deployment model
What this is for
Clearer workflow review, not inflated claims.
Use the pack to understand the workflow, then validate it in the walkthrough.
Synthetic examples
Six synthetic examples for workflow review.
Synthetic examples showing operator surfaces, trust model, and output format.
Synthetic example
Synthetic coverage example
Source tagging shows which Teams surfaces enter the evidence graph and which stay quarantined.
Sample decision timeline
Sample decision timeline
The casebook separates a ready decision path from a restricted cross-wall signal.
Synthetic example
Example audit trail detail
The workbench aligns chronology, supporting evidence, operator actions, and decision outcome.
Synthetic example
Synthetic QA review example
A constrained review lane handles uncertain or restricted signals without making the whole workflow manual.
Non-production walkthrough
Synthetic governance boundary
Governance rules show why restricted content stays contained before export.
Sample narrative report
Sample narrative report
The approved decision ends as a reviewable narrative package with chronology and audit receipt.
Workflow and trust
Trust model and deployment boundary in one glance.
Synthetic example
Background QA review example
The non-production walkthrough shows how constrained signals stay in review while clear evidence continues toward export.
Non-production walkthrough
Governance boundary example
Wall rules and effective scope show how restricted content remains controlled inside the workflow.
Workflow benchmark
Validation metrics to review during the walkthrough.
Use these benchmark categories to review coverage, timing, QA posture, and holdback controls.
How quickly a scan becomes a reviewable chronology.
How quickly a reviewer can move from timeline to report.
How often sampled or held-back events need correction.
Whether the client is running mostly automatic or stricter review.
Synthetic workflow validation
Large-scale workflow benchmarks for walkthrough validation.
Use these results to review coverage, confidence, controls, and export quality before testing your own workflow.
Validation run
Banking workflow benchmark
Meridian Bank NA
Treasury approvals, exception handling, and supervision review.
10,842 Teams-style messages across payments, credit memo, treasury exception, surveillance review workflows.
64 stitched conversations22 channels148 actors8 decision types 89 batched analysis requests
Messages fetched 10,842
Events created 1,276
Teams conversations stitched 64
Teams channels sampled 22 channels, 8 group chats
Unique actors identified 148
Workflow families Payments, Credit memo, Treasury exception, Surveillance review
Decision types modeled Approve, Decline, Escalate, Evidence request, Exception granted, Exception denied, Defer, Override
Decisions detected 312
Exceptions detected 86
Approvals detected 184
QA samples reviewed 292
Duplicate events removed 137
Average confidence score 91.8%
Time to first usable timeline 17m 42s
Time to export 4m 18s
Blocking holdbacks 6 restricted-context items
Validation run
Regulated workflow benchmark
Aurora Regulated Operations
Cross-functional investigations, vendor risk, and policy exceptions.
12,418 Teams-style messages across compliance review, legal hold, vendor risk, operational exception workflows.
81 stitched conversations31 channels213 actors11 decision types 101 batched analysis requests
Messages fetched 12,418
Events created 1,934
Teams conversations stitched 81
Teams channels sampled 31 channels, 14 group chats
Unique actors identified 213
Workflow families Compliance review, Legal hold, Vendor risk, Operational exception
Decision types modeled Approve, Reject, Remediate, Escalate, Hold, Release, Request evidence, Accept risk, Transfer risk, Defer, Close
Decisions detected 428
Exceptions detected 119
Approvals detected 267
QA samples reviewed 386
Duplicate events removed 211
Average confidence score 89.6%
Time to first usable timeline 22m 11s
Time to export 5m 47s
Blocking holdbacks 11 policy-review items
Benchmarks are synthetic workflow simulations for validation. They test scale, coverage, review controls, and export quality before customer-specific validation. Demo feed: /demo/synthetic-workflow-benchmark-feed.json
Next step