Every’s AI experiment suggests workplace agents need a team, not a solo act

Every co-founder Dan Shipper and head of platform Willie Williams say the company’s workplace AI experiments favoured a shared agent over personal assistants that each employee had to maintain. Their account, drawn from a conversation on the Every podcast, offers a useful counterweight to the idea that every worker needs an agent of their own.

Watch Desk analysis

What happened

In a report on the conversation, BigGo Finance describes a year of trials at Every, including the open-source OpenClaw framework and a failed product called Plus One, which aimed to make creating an agent a one-click affair. Williams said OpenClaw setups could be powerful for builders but were complex, fragile and hard to sustain without ongoing attention. Shipper’s personal agent stopped running after its Mac mini lost power, and he did not restart it.

The more durable pattern, Williams said, was colleagues adding to a shared agent that had already built up useful knowledge of the company’s work. He expects one agent per organisation to be the likelier workplace model, while personal agents may suit individual interests and ongoing relationships. Shipper offered a different possibility: a strong shared agent could eventually branch into team agents. These are practitioners’ conclusions from one company’s experience, not a verdict on every workplace. Read BigGo Finance’s account.

Why it matters

The distinction is less about whether an agent can perform a task than who keeps it useful. A personal assistant can accumulate context, but if its setup breaks or its owner moves on, that investment can disappear with it. A shared system may spread the upkeep and let colleagues build on what the agent has already learnt.

That model brings its own questions about permissions, ownership and whose knowledge the agent carries. Shipper also said OpenAI’s persistent Dots had become his main way of using ChatGPT, while describing their permission system as confusing and frustrating. The conversation makes a practical point: integration can make an agent more useful, but a tangle of approvals can quickly make it less so.

Our read

Every’s experience does not prove that one company agent is the future. It does make maintenance and shared context harder to wave away. A clever assistant that only works while one enthusiast tends it is less a workplace tool than a demanding new houseplant.

For teams weighing up persistent agents, the useful test is not how charming or capable one looks in a demo. It is whether colleagues can use it, improve it and understand what it is allowed to do without making one person its full-time keeper.

What to watch

• Whether other organisations find shared agents more durable than personal ones.
• How agent products make permissions and approvals understandable to users.
• Whether company-wide agents remain useful as they branch into team-specific systems.

Discussion spark: Should workplaces invest in one shared AI agent that builds organisational knowledge, or give each employee a personal agent they control?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.