Synthesizing 12 research studies into novel insights is senior consultant work, not an agent task.

Every research synthesis is structurally unique — different client domains, different research questions, different participant populations, and different analytical frames. There is no repeatable template an agent can reliably apply across fintech and e-learning simultaneously.

Success criteria are deeply subjective: what counts as a 'novel insight the team missed' or a 'cross-client theme' requires human judgment to validate. An agent cannot know when it has found something genuinely novel versus something obvious the team already considered and discarded.

140 hours of video requires transcription, speaker diarization, and semantic chunking before any analysis can begin — a non-trivial pipeline. Even with all files accessible, current agents lack reliable long-context synthesis across hundreds of documents without significant hallucination risk.

This is a client-facing consulting deliverable. Fabricated insights, misattributed pain points, or missed themes could damage client relationships, lead to bad product decisions, and harm the consultancy's professional reputation — none of which are easily reversible.

Identifying what is genuinely novel versus what is noise, weighting themes across dissimilar industries, and framing insights for different client contexts all require the kind of domain intuition and interpretive taste that current AI agents cannot reliably replicate.