Data Readiness

FactVerse AI Agent workflows depend on trusted operational context. Before a workflow is used for regular review, confirm that the relevant assets, signals, documents, scenes, and feedback records are available and traceable.

Readiness dimensions

Dimension	Check
Asset identity	Asset IDs, names, locations, owners, and hierarchy are consistent across FactVerse Platform and source systems.
Source freshness	Signals, work records, documents, and scene metadata show clear timestamps.
Signal quality	Units, sampling intervals, missing periods, outliers, and known sensor issues are visible.
Work history	Inspections, work orders, alarms, replaced parts, and field notes are connected to the right assets.
Knowledge sources	Manuals, standard procedures, troubleshooting notes, and site constraints can be searched and cited.
Spatial context	Designer scenes, Inspector views, or simulation packages identify the same assets used by the workflow.
Feedback loop	Reviewed outputs, accepted actions, rejected suggestions, and field corrections are recorded for later improvement.

Minimum data by workflow

Workflow	Minimum useful data	Stronger operating data
Facility operations	Asset identity, current status, latest alarms, inspection records, and open work orders.	Meter readings, documents, location context, nearby equipment, and operator feedback.
Predictive maintenance	Equipment identity, time-series signals, health or anomaly output, and maintenance records.	Component hierarchy, operating mode, parts history, false-positive labels, and engineer review notes.
Physical AI	Designer scene version, asset version, task goal, and simulation assumptions.	SimReady metadata, process constraints, field observations, robot or equipment task traces, and validation records.

Data readiness workflow

1. Define the boundary: tenant, site, asset group, equipment set, time window, or scene.
2. Resolve identity: map user-facing names to stable asset IDs and source-system identifiers.
3. Check freshness: record the newest timestamp for each source used by the workflow.
4. Check quality: flag missing intervals, unit mismatches, outliers, or unmapped records.
5. Bind evidence: connect signals, work records, documents, and scene context to the same asset boundary.
6. Run a read-only sample: confirm the Agent can return evidence and missing-data notes without creating actions.
7. Record corrections: update mappings, source rules, or operating notes before expanding the workflow.

Readiness states

State	Meaning	Next step
Ready	Required sources exist, timestamps are clear, and identity mapping is stable.	Run the workflow with evidence and review capture.
Partial	Some evidence exists, but freshness, mapping, or quality gaps remain.	Use the output as a data-readiness report before making recommendations.
Blocked	Required identity, source data, or permissions are missing.	Fix the source or access issue before workflow execution.

Output expectations

A data-readiness response should include:

• workflow boundary and time window;
• source systems and latest timestamps;
• asset IDs and related source identifiers;
• data quality issues and missing fields;
• documents, work records, scenes, or signals used as evidence;
• whether the workflow should run, run with limits, or wait for correction.

Common gaps

Gap	Impact	Response
Asset name differs across systems	Agent may summarize the wrong asset or merge unrelated records.	Require stable asset ID mapping before workflow use.
Timestamps are hidden	Reviewers cannot tell whether an answer reflects current conditions.	Show source freshness in every accepted output.
Sensor units are inconsistent	Trend or health output may be misleading.	Normalize units and record known conversion rules.
Work history is incomplete	Recommendations miss field context.	Connect Inspector records, maintenance notes, and operator feedback.
Scene version is missing	Physical AI results cannot be reused safely.	Require scene and asset versions in each scenario package.