ActiveSLR
Let's talk

ActiveSLR Plus

AI acrossyour whole review.

An AI reviewer at every stage. It screens, reads and extracts the way a reviewer does, shows its reasoning and its sources, and a person always makes the final call.

The Studies page of a project with the AI runs strip above the studies table: one card each for title/abstract, full text and data extraction, with their progress and Start AI
AI-enabled screening

A reviewer,
not a suggestion box.

At title/abstract and full-text screening the AI reviewer takes a free review slot and writes its decision into it exactly as a person would, with its reasoning. A person always has the final say.

  1. 01Title/abstract screening

    Screens like a reviewer.

    Reads each title and abstract, weighs it against your PICOS and applies your exclusion criteria, then records a decision with its reasoning in a free review slot.

  2. 02Full-text screening

    Reads the whole paper.

    Reads each study's PDF once it has been processed to text, and writes its decision and reasoning into the full-text reviewer slot for QC or reconciliation.

AIon every decision it makes, with its reasoning labelled AI:

A project's AI tab with the AI reviewer switched on: title/abstract screening, full-text screening and data extraction each set to AI on one side, with the model for its slot and its cost and quality tier
The project's AI tab: each stage is set on its own, and each AI slot gets its own model, with its cost and quality tier beside it.

Set up per project, decided by a person.

  1. 01

    Choose how much AI does

    In the project's AI tab, set each stage to No AI, AI on one side or AI on both sides, and pick the model for each slot. Both sides need a double-screening project.

  2. 02

    Start AI on pending work

    Nothing runs until someone presses Start AI on the AI runs strip of the Studies page, where every run is watched and stopped too.

  3. 03

    AI fills a free slot

    It takes whichever review slot is still free and records its decision, with its reasoning in the slot's comment. It never overwrites a decision a person has made.

  4. 04

    A person decides

    On single screening, quality check samples AI's decisions like anyone's. On double screening, every disagreement goes to a reconciler.

AI data extraction

Every value, with its source.

The AI reads the full text against your frozen grid, identifies the treatment arms and answers every field, each value with the source quote, the page and a confidence percentage.

A trial's data extraction beside its PDF: the eye button next to an AI answer has moved the PDF to the cited page and highlighted the quoted sentence

View source on page N

One click to the sentence.

The eye button beside a cell moves the PDF to the cited page and highlights the quoted sentence where it sits in the paper, so you read what surrounds it before you accept the value.

Every quote is tagged by how closely it backs the value:

  • direct quote
  • same section
  • inferred
  • Page 6
A completed extraction quality check beside the trial's PDF: an extractor's eye button has opened its source quote, highlighted in the paper, next to the extracted answers and the final values

Quality check and reconciliation

Checked against the paper.

An AI-completed extraction goes to quality check or reconciliation like anyone's. Either extractor can be the AI: the second person opens each extractor's quote in the PDF with the same eye button and chooses the final value.

  • Only empty or failed cells

    The AI only ever answers cells that are still empty or failed, so nothing a person has entered is overwritten.

  • Extract with AI, one trial

    Draft the empty cells of your own extraction slot with a model the other side is not using. You review every answer and mark it complete yourself.

  • Marked and restorable

    Every AI-filled cell is marked Answered by AI. Type over it and the AI answer stays beside yours, ready to restore.

AI tools

More AI, across the review.

Plus also brings AI to the search, the protocol and the extraction grid, and shows every run and what it used.

  1. Runs & usage

    Start it, watch it.

    Every AI run is started, watched and stopped from the AI runs strip on the Studies page, with one card per stage. A gauge in the sidebar shows how much of your monthly AI allowance you have used.

    • One card per stage
    • Start AI on pending work
    • Monthly usage gauge
    AI Usage in Settings: your share of this month's AI allowance and each team member's usage

  2. PubMed search agent

    Search PubMed from your PICOS.

    The PubMed search agent builds and checks a PubMed search from the project's PICOS, then imports the results straight into the project as studies.

    • Built from your PICOS
    • Search checked for you
    • Imported as studies
    The PubMed search agent tab on a project's Upload Files page, with Generate strategy from PICOS above the PubMed search string

  3. PICOS from your protocol

    Your protocol, read for you.

    Generate from document reads your review protocol, a PDF (scanned documents are fine), and proposes the review question, the PICO and your inclusion and exclusion criteria for you to apply.

    • Reads scanned PDFs
    • All eight fields drafted
    • You apply what you keep
    A project's PICOS tab with the Read the PICO from your document dialog open, ready for a review document as a PDF

  4. Grid & extraction

    Draft the grid, fill the sheet.

    Draft the extraction grid from a short brief or edit it through an AI chat. Extract with AI drafts a trial's empty cells into your own slot, each value with a source quote, a page and a confidence percentage, for you to review and sign off.

    • Draft the grid with AI
    • Edit with AI chat
    • Quote, page and confidence
    The data extraction grid, drafted by AI, with the Edit the grid with AI chat panel open beside it
Guardrails

A person always makes
the final call.

  1. A person always decides.

    AI's decisions go through the same quality check, reconciliation and override screens as anyone else's. AI never acts as the quality checker or the reconciler.

  2. Never overwrites a person.

    A decision someone has already made is never replaced, and a trial someone has started extracting is never taken. Your team keeps working as normal.

  3. Every AI decision is badged AI.

    Decision badges read AI with a sparkle icon instead of Reviewer 1 or Reviewer 2, and its reasoning is the note labelled AI: in the comments strip.

  4. Every value is quoted, with its page.

    Each AI value keeps its source quote, tagged direct quote, same section or inferred, and the page it is on, one click from the highlighted sentence.

  5. Usage is always in view.

    A gauge in the sidebar shows how much of the monthly AI allowance you have used. Administrators see team usage and can switch AI off for a member.

  6. On per organization, per project.

    ActiveSLR Plus is switched on for the whole organization, and AI is set up project by project, stage by stage.

Add-ons

Two add-ons for
faster reviews.

Rapid Review and Auto Project are optional add-ons for fast, AI-led reviews of the PubMed literature. Each is switched on for your team separately from ActiveSLR Plus, and its AI work counts towards the same AI allowance.

Rapid Review add-on

A rapid review, step by step.

One project, one question. AI does the heavy work at every stage, from the search to the PRISMA diagram, and you check each stage and override anything you disagree with.

  1. 01 / 08Question

    Frame the question

    Start a project in a folder of your Rapid Review library, choose a framework (PICOS, PICOC, CoCoPop, PEO, PFO or PIRD) and write the question as one plain sentence. AI drafts one element per framework part for you to review and refine.

    The Rapid Review dashboard with recent folders and recent projects, each project card showing its framework

  2. 02 / 08PubMed search

    An AI-built PubMed search

    An AI agent builds the search string from your framework elements, testing candidate queries against PubMed as it goes, and shows the live hit count. Refine it by chat, then import from PubMed or upload a file.

    Strategy and Imports in a Rapid Review project: the AI-built PubMed search string with its live hit count, beside the Refine the search string with AI chat

  3. 03 / 08Ti-Ab screening

    AI title/abstract screening

    AI screens every study on its title and abstract in one background run and files it as include, exclude or uncertain, with its reasoning. Filter by verdict and override any decision you disagree with.

    Ti-Ab Screening in a Rapid Review project after a finished AI run: verdict filters and each study's decision with its AI reasoning

  4. 04 / 08Full-text screening

    AI full-text screening

    Find Full Text looks for each study's PDF, then the AI reads the whole paper and gives every study a verdict with its reasoning. You decide which verdicts to accept.

    Full Text Screening in a Rapid Review project after a finished run: PDF status filters and each study's full-text decision with its AI reasoning

  5. 05 / 08Summary table

    A summary table, filled for you

    Describe the columns you want. The AI reads the title and abstract of every included study and fills them in, one row per finding, each row keeping the sentence it came from. Correct any cell and download CSV or Excel.

    The Summary Table of a Rapid Review project after extraction: one row per finding with study design, population, sample size, intervention and comparator, each with its evidence

  6. 06 / 08Narrative synthesis

    A narrative synthesis, drafted

    AI writes a narrative across the included studies and ties each point to the study it came from. Refine it by chat, ask for a change to one selected passage or edit it yourself, then download it as a Word file.

    The Narrative Synthesis of a Rapid Review project, with numbered citations to the included studies, beside the Refine with AI chat

  7. 07 / 08Data extraction

    Data extraction with source quotes

    Draft a grid from your PICO, freeze it and let the AI read each included study's PDF against it. Every value keeps the sentence and page it came from, one click away in the PDF.

    Data extraction results in a Rapid Review project: the study's PDF with the quoted sentence highlighted beside the extracted values

  8. 08 / 08PRISMA

    PRISMA, drawn from the counts

    The PRISMA 2020 flow diagram is drawn from the live screening tally, with nothing to fill in, and downloads as a PNG for your report or manuscript.

    The PRISMA 2020 flow diagram of a Rapid Review project, with Download PNG
Auto Project add-on

One question in,
a landscape out.

Give it a name, a framework and one clinical question. It runs the early review on its own and ends in a Treatment Landscape: the drugs and regimens the included studies describe, with the evidence behind each value.

  1. Stage 1 of 5

    Framework

    Drafts the framework elements from your clinical question.

  2. Stage 2 of 5

    Search

    Searches PubMed live to assemble and refine the query.

  3. Stage 3 of 5

    Import

    Fetches the matching records and removes duplicates.

  4. Stage 4 of 5

    Screen

    AI reads every title and abstract against the eligibility criteria.

  5. Stage 5 of 5

    Summary

    Extracts the treatments and looks up each drug's class, brand, FDA and EMA status.

The New auto-search dialog with a folder, a project name, a framework and a clinical question filled in, ready to start
01Ask one questionNew auto-search takes a project name, a framework, the clinical question and a folder. The run starts at once.
A finished auto search reading Your treatment landscape is ready, with the imported, included and table row counts and the review question and framework
02Leave it runningFive stages run unattended on the server, with running counts of imported, included and table rows. Close the tab and come back.
The Treatment Landscape of an auto project on the Drugs tab, with drug classes, lines of therapy, brands, FDA and EMA status and the number of supporting studies; the manufacturer column is blurred
03Read the landscapeDrugs and Regimens tables, with a Data confidence bar that sorts every looked-up value from Verified to Not assessed. Export CSV or Excel.
The Treatment Landscape with the Supporting studies drawer open for one drug, each study card showing the verbatim sentence the model grounded on; the manufacturer column is blurred
04Check the evidenceOpen a row's supporting studies: each card shows the verbatim sentence behind the value. The Studies screen shows every screening decision with its AI reasoning.

Not a full systematic review. An auto project stops at the Treatment Landscape and its screening decisions are read-only. When you need full-text screening, a synthesis, data extraction or a PRISMA diagram, run the question as a Rapid Review project.

Next steps

Bring AI to your next review.

ActiveSLR Plus is switched on for your whole organization. Talk to us about your team.

Ready when you are

Let's accelerate
your evidence.

Start your ActiveSLR® journey today. Experience structured, AI-assisted screening, and bring in ActiveSLR Plus when you want AI at every stage.

Book a demoactiveslr@pharmaquant.org