SciSpace Academy
TutorialLiterature review › End-to-end research workflow with AI agents

Agentic Research Workflow: From Literature Search to Submission

Dr. Faheem UllahEnglishLiterature reviewEnd-to-end research workflow with AI agents

TL;DR

1-min read

This tutorial presents an end-to-end agentic research workflow in SciSpace, following a single project from the first literature search through to journal submission. It is designed for researchers, PhD students and faculty who want to run a full project in one AI-supported workflow rather than spreading the work across many separate tools and documents.

Key takeaways

  • In an agentic setup, AI links steps such as searching for papers, screening them, extracting key information, synthesising evidence and drafting text, rather than answering single questions.
  • Keeping literature search, screening, synthesis and drafting in one SciSpace workflow cuts the time lost moving material between disconnected tools.
  • Working in one environment also makes it easier to trace how a manuscript was built, from the sources found to the claims in the draft.
  • The researcher sets the goals and criteria, and human checkpoints remain essential for inclusion decisions, source verification and conclusions.
  • An agent can help organise findings into a draft and check formatting, but authors must verify every citation and confirm that conclusions follow from the evidence.
  • Before submitting, check the target journal's AI policy and disclose AI use where it is required.
Apply it

To apply this, set up your next project in SciSpace with clear inclusion criteria, let the agent handle the linked search and drafting steps, and review each stage before the work moves forward.

How to do it

  1. Set clear goals and inclusion criteria

    Before starting, define your project goals and inclusion criteria in SciSpace. Clear criteria guide every later step of the agentic research workflow.

  2. Link search, screening and extraction

    Let the agent run linked steps: searching for papers, screening them and extracting key information. Keeping these in one workflow cuts time lost moving material between tools.

  3. Make inclusion decisions yourself

    Review the agent's screening results and make the final inclusion decisions. Human checkpoints here protect the quality of your evidence base.

  4. Synthesise evidence into a draft

    Ask the agent to organise findings into a draft. Working in one environment makes it easier to trace how each claim was built from its sources.

  5. Verify citations and conclusions

    Check every citation and confirm that your conclusions follow from the evidence. As the author, you remain responsible for the manuscript.

  6. Check the journal's AI policy

    Before submitting, read your target journal's AI policy and disclose AI use wherever it is required.

Watch the tutorial38 min

Agentic Research Workflow in SciSpace: From Search to SubmissionPlay video: Agentic Research Workflow in SciSpace: From Search to Submission

Chapters

  1. Welcome and tutorial outline
  2. What is a literature review, and which type do you need?
  3. What types of literature review are there?
  4. How do you write a good research question for a review?
  5. Which databases and inclusion/exclusion criteria should you use?
  6. How does paper screening work: title, abstract, full text?
  7. What is snowballing in a literature review?
  8. How do you build a data extraction form?
  9. How do you analyse extracted data with thematic analysis?
  10. Can AI agents find and rank papers for a literature review?
  11. How do you extract data from 73 papers at once with AI?
  12. How do you chat with a research paper or turn it into a podcast?
  13. Can SciSpace generate a systematic literature review?
  14. Why can't you submit an AI-generated literature review?
  15. Diagrams, AI detection, journal finder, citation audit and posters
  16. Why is submitting AI-written work unethical? Disclosing AI use
  17. What skills do future researchers need? Prompt engineering

Try this on SciSpace

Frequently asked questions

What is an agentic research workflow?

It is a way of working in which an AI agent carries out linked research tasks, such as searching, screening, summarising and drafting, as part of one continuous process. The researcher defines the question and criteria, reviews outputs at each stage and decides what goes into the final manuscript, so responsibility stays with the people doing the research.

Can an AI agent prepare a manuscript for submission?

An agent can help organise findings into a draft, suggest structure and check formatting against journal guidelines. It cannot take responsibility for the science. Authors must verify every citation, confirm that conclusions follow from the evidence, check the target journal's AI policy and disclose AI use where it is required.

Related tutorials