Sovereign AI use case · Universities and Research

AI for literature and evidence review screening in research

Reviews take months of screening, delaying grants, papers and policy advice. Fatigue leads to inconsistent decisions between screeners. Extraction errors carry into findings. Teams struggle to document decisions in the detail that reporting guidelines such as PRISMA expect.

Inaaya use-case library · Last reviewed 19 November 2026

In short

Literature and evidence review screening is the process of searching, screening and extracting data from studies for systematic, scoping or rapid reviews and background research. An AI assistant can apply the team's criteria and produce extraction tables, while researchers make and verify screening decisions according to their review protocol.

Who does this today.

Graduate students, research associates, librarians and investigators do this work. It is hard because searches return hundreds or thousands of records, screening must follow the protocol consistently, and extraction is repetitive yet must be accurate.

What Inaaya can do, step by step.

  1. Reads the review protocol, research question and inclusion and exclusion criteria.
  2. Screens titles and abstracts against the criteria and records a suggested decision with reasons for each record.
  3. Marks uncertain records for human screening.
  4. Extracts agreed data items from included full texts into a review table with citations and page references.
  5. Lists conflicts and discrepancies between suggested and human decisions for resolution.
  6. Keeps a record of every screening decision and its reason, for the review's flow diagram and methods.
  7. Drafts a narrative summary of findings by theme for researchers to revise.
You provide
  • Review protocol and criteria
  • Search results export
  • Full-text articles the team has access to
  • Data extraction template
You get
  • Screening decisions with reasons
  • Review table with citations
  • Decision record for reporting
  • Draft thematic summary
Your people decide
  • Researchers make and verify all inclusion decisions as the protocol requires
  • Principal investigator approves the extraction template and findings
Data and controls
  • Can run with Canadian data residency, or in the environment a research agreement requires, on a model and host the institution approves
  • Research data handled under applicable privacy and sector rules and the terms of each data agreement
  • Access limited to the project team and research office staff the institution designates
  • Every action, draft and approval recorded
  • Licensed full texts used only under the institution's access terms
How to measure it
  • Records screened per week
  • Agreement between suggested and human decisions
  • Extraction errors found at verification
  • Time from search to completed review table
The aim

Designed to reduce manual screening time, with researchers verifying every decision.

Questions people ask.

Can AI screen papers for a systematic review?

An AI assistant can screen records against the protocol's criteria and give reasons for each suggested decision. Researchers verify decisions as their protocol and reporting standards require.

How should AI use be reported in a literature review?

Describe how the assistant was used in the methods, how human screeners verified its decisions and how conflicts were resolved, following the journal's and reporting guideline's expectations.

Can AI extract data from research papers?

It can extract agreed data items into a table with page references, so researchers can check every value against the source.

Related use cases.

Apply to be a founding partner