Maggie Appleton says agents start work knowing too little about you

Maggie Appleton identifies a split that shapes how agents fail: the human's world and the agent's world are not the same place. The human carries goals, constraints, relationships, and history that do not fully transfer. The agent works from what it received, not what the human meant. That gap is where plans go wrong. Appleton borrows the term 'boundary object' from sociology. She uses it to name a plan, a brief, or a task list that both sides can work from at once. Each holds enough shared meaning that the human and the agent can act without having the same full picture. Most interfaces make that object thin: a text box, a prompt, a chat turn. The richer the shared object, the less each side has to guess. Appleton calls her prescription 'thicker interfaces'. It is a design target, not a feature. It means building the space between human and agent so that assumptions surface before the agent runs. Disagreements show up as edits rather than surprises. The plan becomes a negotiated record both sides can inspect. For product teams, this reframes the handoff from a moment into a medium.