AI Essentials: From Tools to Strategy
Week 4: AI in Scholarly Workflows & Infrastructure
● Identify how AI is structurally reshaping specific steps in scholarly
workflows
● Identify variation and accuracy between some GenAI tools
● Discuss where these tools fit within library or publishing workflows
Four Ways AI Changes Workflows
● Manual Labor to Scalable Assistance
● Step-by-Step Process to Responsive Systems
● One Standard Output to Multiform Interpretation
● Local Control to Shared Systems
From Manual Labor to Scalable Assistance
● Crossref and OpenAlex now use machine learning to automatically
disambiguate author names and affiliations.
● OpenAIRE applies AI models to enrich metadata across repositories.
● Elsevier’s JournalFinder scans abstract language to recommend likely
publication venues.
● Clarivate’s Web of Science uses AI to assign topical classifications.
These innovations help us keep up with growing research volume, but they also raise
important questions: What role should human oversight still play? How do we
balance speed with local context and interpretation?
From Step-by-Step Process to Responsive Systems
● Springer Nature’s Geppetto uses AI to screen for indicators of fraud or AI
generation in submitted manuscripts.
● Prophy’s reviewer recommender is now embedded in Editorial Manager
● NEJM has experimented with AI-generated reviews to support editorial triage
From One Standard Output to Multiform Interpretation
● Dimensions and Scopus produce different article summaries from similar
content.
● EBSCO’s Scholarly Graph and other knowledge graphs reflect different
underlying taxonomies.
● Ex Libris Primo’s Research Assistant produces contextualized summaries in
user-facing interfaces.
From Local Control to Shared Systems
● figshare and Dryad leverage structured metadata workflows that can integrate
with enrichment services.
● Elsevier’s Article Transfer Service uses AI to re-route submissions across
journals.
Try AI for Metadata
Each group will test two AI tools and compare how each performs the same core tasks:
● Extract subject keywords
● Generate a plain-language summary
● Create a metadata block (e.g., in JATS or Dublin Core)
Suggested Tools:
● https://chat.openai.com
● https://elicit.org
● Other tools participants know (e.g., Scite.ai, Consensus.app)
Prompt: Paste a research abstract or brief article paragraph (not sensitive or proprietary) into the tool’s
chat. Use the same prompt for each tool (feel free to refine the prompts if your results are not good).
Note: ·
● Differences in tone, terminology, accuracy
● How “visible” the tool is about its methods or sources
● What labor this replaces, and what still requires human judgment
● How accurate or biased was the result?
● Would you trust this in your catalog or repository?
● Who’s responsible for errors or omissions?
Group Report out & Reflection
● Did any tool feel more trustworthy or transparent?
● Would you deploy either of these in your workflow?
● What would need to change (in people, policy, training) to support its use?
Take-Home Worksheet: Drafting Principles for AI Adoption
Reflect on how AI is emerging in your workflows, and begin articulating guiding principles for adopting or
integrating AI tools in alignment with your institution’s mission, values, or community expectations.
Instructions:
1. Think of 1–2 places where AI is (or could be) active in your organization.
2. For each, consider what it enables, what questions or uncertainties it introduces, and what kind of
principle might guide responsible or effective use.
3. Bring your reflections to Week 6, where we’ll begin building alignment strategies.

Hudson Vitale "AI Essentials: From Tools to Strategies: A 2025 NISO Training Series, Session Four - AI in Scholarly Workflows and Infrastructure"

  • 1.
    AI Essentials: FromTools to Strategy
  • 2.
    Week 4: AIin Scholarly Workflows & Infrastructure ● Identify how AI is structurally reshaping specific steps in scholarly workflows ● Identify variation and accuracy between some GenAI tools ● Discuss where these tools fit within library or publishing workflows
  • 4.
    Four Ways AIChanges Workflows ● Manual Labor to Scalable Assistance ● Step-by-Step Process to Responsive Systems ● One Standard Output to Multiform Interpretation ● Local Control to Shared Systems
  • 5.
    From Manual Laborto Scalable Assistance ● Crossref and OpenAlex now use machine learning to automatically disambiguate author names and affiliations. ● OpenAIRE applies AI models to enrich metadata across repositories. ● Elsevier’s JournalFinder scans abstract language to recommend likely publication venues. ● Clarivate’s Web of Science uses AI to assign topical classifications. These innovations help us keep up with growing research volume, but they also raise important questions: What role should human oversight still play? How do we balance speed with local context and interpretation?
  • 6.
    From Step-by-Step Processto Responsive Systems ● Springer Nature’s Geppetto uses AI to screen for indicators of fraud or AI generation in submitted manuscripts. ● Prophy’s reviewer recommender is now embedded in Editorial Manager ● NEJM has experimented with AI-generated reviews to support editorial triage
  • 7.
    From One StandardOutput to Multiform Interpretation ● Dimensions and Scopus produce different article summaries from similar content. ● EBSCO’s Scholarly Graph and other knowledge graphs reflect different underlying taxonomies. ● Ex Libris Primo’s Research Assistant produces contextualized summaries in user-facing interfaces.
  • 8.
    From Local Controlto Shared Systems ● figshare and Dryad leverage structured metadata workflows that can integrate with enrichment services. ● Elsevier’s Article Transfer Service uses AI to re-route submissions across journals.
  • 9.
    Try AI forMetadata Each group will test two AI tools and compare how each performs the same core tasks: ● Extract subject keywords ● Generate a plain-language summary ● Create a metadata block (e.g., in JATS or Dublin Core) Suggested Tools: ● https://chat.openai.com ● https://elicit.org ● Other tools participants know (e.g., Scite.ai, Consensus.app) Prompt: Paste a research abstract or brief article paragraph (not sensitive or proprietary) into the tool’s chat. Use the same prompt for each tool (feel free to refine the prompts if your results are not good).
  • 10.
    Note: · ● Differencesin tone, terminology, accuracy ● How “visible” the tool is about its methods or sources ● What labor this replaces, and what still requires human judgment ● How accurate or biased was the result? ● Would you trust this in your catalog or repository? ● Who’s responsible for errors or omissions?
  • 11.
    Group Report out& Reflection ● Did any tool feel more trustworthy or transparent? ● Would you deploy either of these in your workflow? ● What would need to change (in people, policy, training) to support its use?
  • 12.
    Take-Home Worksheet: DraftingPrinciples for AI Adoption Reflect on how AI is emerging in your workflows, and begin articulating guiding principles for adopting or integrating AI tools in alignment with your institution’s mission, values, or community expectations. Instructions: 1. Think of 1–2 places where AI is (or could be) active in your organization. 2. For each, consider what it enables, what questions or uncertainties it introduces, and what kind of principle might guide responsible or effective use. 3. Bring your reflections to Week 6, where we’ll begin building alignment strategies.