Session 2: Building the Human-AI Research Pipeline

Building the pipeline once teaches more than reading about it ten times.

This guide offers a hands-on workshop for designing and executing a systematic review with AI tools. The approach is cognitive-first: instead of just learning to operate tools, we learn to think with them. We engage with AI as a mirror that makes our own research processes visible, helping us deconstruct complex tasks like "finding a gap" or "building a theory" into explicit, repeatable steps.

By the end of this session, you'll move beyond generic AI interaction to purposeful orchestration.

Learning outcomes

By the end of this session, you'll be able to:

• Construct a high-quality, curated literature set using discovery tools (Research Rabbit for the human eye, OpenAlex for the agent) and a management tool (Zotero)
• Deconstruct complex research tasks (e.g., "synthesizing a framework") into explicit, step-by-step cognitive operations
• Design and implement structured, multi-step prompts ("cognitive blueprints" (Structured multi-step prompts that guide AI through complex research tasks)) that guide AI through sophisticated analytical work
• Critically evaluate AI-generated outputs, identifying common failure modes (like "botshit" and paradigm blindness)
• Architect a research workflow that strategically combines your domain expertise with AI capabilities

The complete research pipeline

     seed papers
          |
          v
  citation discovery
          |
          v
    expanded set
          |
          v
 +--------+--------+
 | Human curation  |
 +--------+--------+
          |
          v
  reference library
          |
          v
AI-assisted analysis
          |
 +--------+--------+
 | quality control |
 +--------+--------+
          |
  issues? +----> revise method
          |            |
          +<-----------+
          |
          v
 research synthesis

Human judgment provides quality gates throughout this pipeline: AI handles scale and pattern detection, while researchers provide curation and evaluation.

Core readings: the AI toolkit

Mindset & mental models

Techniques & best practices

Risks & responsibility

Supplementary readings

Quick reference guides

Foundational papers

Advanced context

Session structure

Our hands-on session will follow this structure:

1. Pipeline Overview - Understanding the complete workflow
2. Discovery & Curation - Research Rabbit, OpenAlex, and Zotero
3. Tool setup - Adding only the capabilities needed for the selected working mode
4. Pipeline Practice - Working with a sample literature set

We use the scaling and scalability literature as a shared case study throughout these guides: research on how organizations grow, and on whether their structures, processes, business models, and capabilities can handle that growth. In this course shorthand:

Scaling: growth in size or scope.

Scalability: the capacity to grow without the organization breaking or losing coherence.

For a taught course, start with the papers your instructor assigns and follow the access route in the course brief. If that route is missing, ask the instructor. If you are learning independently, choose 3 to 5 papers relevant to your question, or begin with the Student Vault’s fictional practice records. The starter includes no paper pack. Expand your literature set with the discovery methods below when your question calls for it.

MCP server setup (Live demo)

For the connected-tool demonstration, we'll add the first MCP servers. Follow along only if this route fits your course brief and sharing permissions. Manual copying and saving remain valid alternatives; provided-output, human-peer, and individual-reading modes need no MCP connection.

MCP has had less attention lately than skills and plain command-line tools, and it is still worth the hour. It is the standard interface through which plugins in Claude Code, Codex, and Cherry Studio reach anything outside the machine, and it gives an organization a seam to put controls on: which servers an agent may reach, what each server is allowed to do on its behalf, and where that is logged. The protocol enforces none of that itself; the host's approvals, the server's own authorization, and the operating system's permissions do. An agent that does everything through a shell has no such seam.

For more MCP exploration, see the MCP Explorer Guide.

Record which parts of the workflow you tested and which remain a design. A working connection does not establish that a research claim is sound.

Pre-class setup for Session 2

Understand the task before setting up tools. Bring the notes/benchmark/ files from Session 1: your question, protocol, and search log are the checks on this session's work. Arriving with worksheets and no AI tooling configured is a valid pre-class state. Your instructor's course brief determines the working mode and any tool requirements.

In the connected-tool demonstration, we will:

• Set up knowledge bases together
• Practice Zotero MCP searches
• Create a first agent in Cherry Studio's Agent Workspace and compare its output with the chat view
• Export conversations to Obsidian
• Create literature note templates
• Practice conversation forking for different analyses

In other modes, follow the worksheet with the assigned account or your own evidence checks. Record unperformed comparisons and runs as not performed.

Check the result

Related Resources

Navigation: Previous: Session 1 • Return to Case Study Overview • Next: Session 3