Personal vs Team Agents

Every's reversal on how to deploy agents in an organization: from personal agents (one per employee, a digital replica of them) to shared team / "unit" agents that reliably perform parts of many people's jobs.

The shift

In the first flush of OpenClaw excitement, Every had every employee spin up a personal agent — "a replica of themselves." The problems surfaced fast:

• Maintenance fell on the owner. "Every time an agent broke, the person it belonged to had to fix it." Even with a stable harness, agents need upkeep. Great for tinkerers (the back-and-forth is the appeal); a burden for everyone else who "wants the benefits of an agent without the obligation of having to manage and mend it."
• No continuity. A personal agent's value is tied to whoever trained it and "disappears if that employee leaves."

The more successful model: agents as coworkers that "reliably perform parts of many different people's jobs."

• Shared maintenance — one person updates the agent's skills and the whole team benefits, vs. pushing the same update across 10 personal agents.
• Continuity / institutional memory — a team agent "retains company context and knowledge, acting more like a project manager, sales lead, or chief of staff than a private assistant."

The org design move

NLW generalizes it: instead of each person asking what an agent could do for them, map the Venn-diagram overlap of people's jobs and build agents that live in the shared space. "That seems to be where AI-native companies like Every are getting a lot of their current value."

Connections

• Every — the company whose experiments produced both the original and revised model.
• Agents as Employees vs Human-Agent Collaboration — orthogonal axis: ownership (personal vs. team) vs. mode (employee vs. collaborator).
• Agent Autonomy Spectrum — maintenance burden was a key reason mega-autonomy underdelivered.
• Build With, Not Against — cross-domain twin: Every's move from per-person agents to shared "unit" agents is the same lesson as the org heuristic — shared, composable infrastructure beats per-actor land-grabs no one can maintain or hand off.
• Agent Memory — how "retains company context" works in practice. Anthropic — The Components of Agent Memory built a shared Agent-Maintained Wiki across 128 agent sessions coordinated by a shared log, so the wiki grew from new-employee to seasoned-employee knowledge. It also found the shared-memory failure mode: several agents writing one shared page overwrote each other's work (41/57 facts survived, against 57/57 with one writer per page).
• Placement and Judgment Over Raw Execution — the shared-unit reversal is an instance of do the upkeep once at the team layer (Move 1) and compose, don't conquer (Move 3) in that synthesis.

Sources

• The AI Daily Brief — Why Agents Still Need Humans — NLW quotes a second Every essay reflecting on their first experiments (distinct from Dan Shipper — After Automation).