Research progresses through 12 defined stages, from initial intake to final publication. At each stage, SciAI captures checkpoints, logs agent actions, and maintains full provenance.
Each stage represents a distinct phase of research. Transitions are tracked, and you can always see where your project stands.
Project setup and initial requirements
Define your research question, scope, and constraints. Create your project and configure initial settings.
Mission planning and task decomposition
Break your research into missions. The Planner agent helps decompose complex goals into actionable tasks.
Paper search and claim extraction
Search PubMed, upload PDFs, and extract claims. The Literature agent helps identify relevant papers and key findings.
Knowledge consolidation and patterns
Connect claims across papers. Identify patterns, contradictions, and gaps in the existing research.
Method design and pre-registration
Define your statistical methods. Pre-register hypotheses. Lock confirmatory analyses to separate from exploratory work.
Data collection and ingestion
Upload datasets, define schemas, and link to sources. All data ingestion is logged with provenance.
Quality control and validation
The QC agent validates data quality. Flag issues, run checks, and approve data before analysis.
Statistical analysis execution
Execute analysis scripts. Run meta-analyses, calculate effect sizes, test hypotheses. All runs are logged.
Visualization generation
Generate forest plots, tables, and diagrams. Visualizations link back to the analyses that produced them.
Writing and citations
The Manuscript agent helps draft sections. Citations are automatically linked to your library with proper formatting.
Internal review and revisions
The Reviewer agent provides critical feedback. Collect comments, address issues, and finalize your manuscript.
Final output and archival
Generate final outputs, export manuscripts, and archive your project with complete provenance records.
Separate your pre-registered, hypothesis-testing work from exploratory analysis. This distinction is captured in the provenance trail.
Pre-registered analyses that test specific hypotheses. Once locked, changes require formal deviation workflows with rationale and approval.
Open-ended exploration and hypothesis generation. Actions are still logged, but with more flexibility for iteration.
Every action in SciAI contributes to a complete audit trail. From data ingestion to final publication, the provenance is preserved.
Papers, datasets, and missions flow through agent runs → producing artifacts → requiring approvals → graduating to stage checkpoints.
Every run, decision, and change is logged
Locked artifacts can't be silently changed
See how SciAI can bring clarity to your workflow.
Book a Demo