3  Guide 0 — The Opportunity: AI Research Agents & the “AI Research OS”

3.1 Why research is the best vertical for AI agents

Research is text-heavy, process-heavy and repetitive — exactly what LLM agents are good at:

Research task Today (manual) With an agent
Finding papers & gaps Weeks–months of reading Hours, with citation graph & gap statement
Systematic / literature review 6–12 months 2–4 weeks with PRISMA trail
Experiment design & analysis Needs statistician + bioinformatician Guided pipeline + auto-generated notebook
Writing & submitting a paper 2–6 months, many rejections Draft in days; journal matching; simulated review
PhD supervision 1 meeting per month if lucky Weekly structured feedback

The million-dollar opportunity is not another ChatGPT wrapper. It is an agent that owns a complete workflow, keeps state, uses real research tools (arXiv, PubMed, Semantic Scholar, Crossref, statistical libraries), and produces auditable outputs (tables, PRISMA diagrams, notebooks, manuscripts).

3.2 Market size (order-of-magnitude, for planning)

Segment Global count (approx.) Willingness to pay
Universities & HEIs ~25,000 $20k–500k / year for research tools (they already buy Turnitin, EndNote, Scopus, SciVal)
PhD students ~3–4 million enrolled $10–50 / month
Active researchers / faculty ~9 million $20–50 / month; labs $2k–5k / year
Corporate R&D (pharma, agri, engineering) tens of thousands of labs $100k–1M / year
Emerging-market universities (Pakistan, GCC, South Asia, Africa) ~3,000+ Underserved; regional pricing; strong policy push for research output

Comparable products & prices (to anchor pricing): Elicit, Consensus, Scite, SciSpace, Research Rabbit, Litmaps, Paperpal, Writefull — individual plans typically $10–50/month; institutional deals $10k–200k/year.

3.3 The five agents at a glance

# Agent Core loop Teaches you Guide
1 Research Discovery Search → cluster → gap → idea → proposal Tool use, search APIs, citation networks 01_research_discovery_agent.md
2 Literature Review PRISMA search → screen → extract → write Structured outputs, pipelines, auditability 02_literature_review_agent.md
3 Research Experiment Plan → analyse → model → test → methods Code + data + statistics agents 03_research_experiment_agent.md
4 Writing & Publication Draft → match journal → simulate review → rebut Multi-persona simulation, critique loops 04_academic_writing_publication_agent.md
5 PhD Supervisor Roadmap → weekly check-in → critique → slides Memory, long-running state, coaching 05_phd_supervisor_agent.md

3.4 Ranking (from the council)

Rank Agent Market Revenue potential Why
🥇 Literature Review Agent Massive $$$$$ Everyone needs it; measurable ROI; easiest to demo
🥈 Research Discovery Agent Massive $$$$$ Top-of-funnel; viral; feeds all other agents
🥉 Writing & Publication Agent Very high $$$$ Recurring — every paper; upsell to institutions
4 Experiment Agent High $$$$$ Highest enterprise value (pharma/agri) but domain-specific
5 PhD Supervisor Agent High $$$$ Retention & institutional stickiness

3.5 The AI Research OS (Phase 2 vision)

Research Idea
     │
     ▼
[1] Research Discovery Agent ──► research gaps, proposal
     │
     ▼
[2] Literature Review Agent ───► PRISMA review, evidence table
     │
     ▼
[3] Experiment Agent ──────────► design, analysis notebook, figures
     │
     ▼
[4] Writing & Publication Agent ► manuscript, journal, reviewer sim, rebuttal
     │
     ▼
[5] PhD Supervisor Agent ──────► roadmap, weekly progress, defence prep

Think “Cursor + ChatGPT + EndNote + a research assistant + a bioinformatics analyst — for scientists.” A university could pay $50k–500k / year for a platform that owns this whole pipeline.

3.6 Common architecture (all five agents share it)

            ┌──────────────┐
 user goal ►│   Planner    │ (LLM: decomposes the task into steps)
            └──────┬───────┘
                   ▼
            ┌──────────────┐     ┌──────────────────────┐
            │   Workers    │◄───►│ Tools (arXiv, PubMed, │
            │ (LLM + tools)│     │ Semantic Scholar,     │
            └──────┬───────┘     │ pandas, sklearn ...)  │
                   ▼             └──────────────────────┘
            ┌──────────────┐
            │  Structurer  │ (Pydantic structured outputs → tables/JSON)
            └──────┬───────┘
                   ▼
            ┌──────────────┐
            │   Critic     │ (LLM reviews for gaps, errors, hallucinations)
            └──────┬───────┘
                   ▼
            ┌──────────────┐
            │   Writer     │ (LLM produces Markdown/Docx/Notebook)
            └──────────────┘

3.7 Where the money is (summary)

  • Individuals: $20–30/month × thousands of PhD students = steady base revenue and word-of-mouth.
  • Institutions: $20k–100k/year × even 10 universities = $200k–1M ARR.
  • Enterprise R&D: $100k–500k/year per pharma/agri company for private deployments and custom pipelines.
  • Services: training workshops, custom pipeline development, data-migration — often 20–30% of early revenue.

See 07_go_to_market_and_pricing.md for the full sales playbook.

Codanics — Python ka Chilla: Build AI Agents

  • Course: https://codanics.com/courses/python-ka-chilla-build-ai-agents/
  • Website: https://codanics.com · YouTube: @codanics
  • WhatsApp channel: https://whatsapp.com/channel/0029Va7nRDq3QxRzGqaQvS3r
  • Complete free course: https://www.youtube.com/watch?v=UE3o-7KRNXs&list=PLaQmdklCcX40