Skip to content

What is Quick Report?

Quick Report (Autopilot) is a one-click pipeline that runs the entire systematic review workflow automatically using sensible default settings. Instead of manually stepping through screening → extraction → analysis → report, click the 🚀 button and let AI4Meta handle it.

Pipeline Steps

The autopilot runs these steps sequentially:

| Step | What it does | Default settings | |------|-------------|-----------------| | 1. Screening | Multi-LLM deliberation screening of all pending papers | Uses project protocol (PICO, inclusion/exclusion criteria) | | 2. Codebook | Generates data extraction variables from your protocol | AI-generated from PICO + research questions | | 3. Extraction | Extracts data from all included papers using the codebook | LLM extraction, fills missing cells only | | 4. Dataset | Compiles extractions into an analysis-ready dataset | Auto-confirms for analysis | | 5. Analysis | Runs random-effects meta-analysis (REML) | REML estimator, 95% CI, no moderators | | 6. Report | Generates a full report with all available sections | All applicable sections included |

Smart Skipping

Each step checks if it's already been completed:

  • Papers already screened → screening is skipped
  • Codebook exists → generation is skipped
  • All cells already extracted → extraction is skipped
  • Dataset already confirmed → compilation is skipped
  • Analysis results exist → analysis is skipped

This means you can safely re-run the pipeline after making manual changes.

How to Use

  1. Set up your project with papers and a protocol (PICO, research questions)
  2. Click the 🚀 Quick Report button in the workspace toolbar
  3. Review the pipeline steps in the confirmation dialog
  4. Click Run Autopilot to start
  5. Watch progress in the button and chat panel
  6. When complete, open the Report tab to view results

When to Use Autopilot vs Manual Steps

Use Autopilot when:

  • You want a quick overview of your data
  • Your review is straightforward (clear PICO, standard outcomes)
  • You're exploring a dataset for the first time
  • You want a draft report to iterate on

Use manual steps when:

  • You need custom screening criteria per paper
  • Your codebook requires domain-specific variables
  • You want to review extractions before analysis
  • You need specific analysis settings (moderators, subgroups)
  • Your review has complex methodology requirements

Limitations

  • Uses default settings for all steps (REML random-effects, no moderators)
  • Screening uses AI consensus — may flag papers for human review
  • Extraction quality depends on abstract/full-text availability
  • Report is a starting draft — always review and edit before publishing
  • Cannot customize individual step parameters during autopilot run