# If we want them to do Knowledge Work, design them as Knowledge Agents — Benjamin Clavié, Mixedbread

Source: [AI Engineer](https://www.youtube.com/watch?v=O84lhGc1OOI)  
Feed7 permalink: https://feed7.dev/p/if-we-want-them-to-do-knowledge-work-design-them-as-knowledge-agents-ben-1f64i2y  
Published: 2026-09-16T15:00:39.000Z  
Trust: Source Linked (source_linked)

## Why Included

Knowledge agents need intent decomposition, specialist researchers, and multiple retrieval primitives; copying a coding-agent loop fails when evidence is ambiguous and context-dependent.

## Source Summary

Clavié argues that non-code knowledge lacks durable identifiers and explicit definitions: a phrase such as “30 days” can represent several unrelated obligations. Search therefore begins with intent and conditional context, not a grep-ready target.

## Practical Implication

Design the main agent like a professional lead: decompose the question, send bounded research to **specialist subagents**, and combine their memos. Give the harness multiple primitives—**BM25, semantic search, and multimodal PDF search**—and teach it when each applies.

## Agent-Ready Context

Clavié argues that non-code knowledge lacks durable identifiers and explicit definitions: a phrase such as “30 days” can represent several unrelated obligations. Search therefore begins with intent and conditional context, not a grep-ready target.

Design the main agent like a professional lead: decompose the question, send bounded research to **specialist subagents**, and combine their memos. Give the harness multiple primitives—**BM25, semantic search, and multimodal PDF search**—and teach it when each applies.

The cited benchmark result says improved search orchestration reduced the human-agent mistake gap by **40%**, but the talk omits enough setup detail to generalize that number. Better orchestration also cannot recover evidence absent from the corpus or retrieval tools.

## Connected Context

Feed7 judgment across 807 accumulated Signals:

This sharpens knowledge-agent design around ambiguity: the lead agent must infer intent, decompose research, and choose among retrieval modes rather than merely distribute queries. It supports bounded specialist subagents while conflicting with fixed multi-agent chains that lose shared context. The reported mistake-gap reduction is evidence for orchestration, but remains too underspecified to generalize.

- [Where RL Will Take Search — Maximilian-David Rumpf, SID.ai](https://feed7.dev/p/where-rl-will-take-search-maximilian-david-rumpf-sid-ai-1jeb1sw) — Both isolate retrieval in bounded specialists to protect the lead agent’s context; this Signal adds that those specialists need multiple search primitives selected according to intent and evidence type.
- [Why We Killed Our Multi-Agent Pipeline — Subbiah Sethuraman and Abhilash Asokan, ZS Associates](https://feed7.dev/p/why-we-killed-our-multi-agent-pipeline-subbiah-sethuraman-and-abhilash-a-0fmz3z3) — The failed fixed pipeline supplies the architectural boundary for this proposal: one lead retains end-to-end ownership while subagents handle bounded investigations rather than passing the whole task through a chain.
- [It’s Tokens All The Way Down: How RLMs are Different — Kevin Madura, AlixPartners](https://feed7.dev/p/it-s-tokens-all-the-way-down-how-rlms-are-different-kevin-madura-alixpar-1w7pcym) — RLMs provide a complementary way to inspect and partition large inputs programmatically; this Signal focuses the same bounded-delegation principle on ambiguous knowledge retrieval across heterogeneous search tools.

## Context Map

- Layer: agent
- Domains: research
- Topics: harness-engineering, subagents, retrieval

## Uncertainty

- The cited benchmark result says improved search orchestration reduced the human-agent mistake gap by **40%**, but the talk omits enough setup detail to generalize that number. Better orchestration also cannot recover evidence absent from the corpus or retrieval tools.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.