06-reference

every agent launch shared slack coworker

2026-10-06·reference·source: Every·by Dan Shipper

Introducing the Every Agent -- Dan Shipper (Every)

Every's general-availability launch of the Every Agent: a single, shared AI coworker living in the company's Slack, built on Claude Managed Agents, that any employee can tag (@Every) to delegate work -- PRs, contracts, reports, video clips -- in the open thread. Pricing: $15/person free credits (up to $500/workspace), then $30/month per workspace on top of pass-through token costs (explicitly "zero markup -- what we pay is what you pay"). Connects to 1,000+ tools (Drive, Gmail, Notion, Figma, GitHub, HubSpot...), ships pre-loaded with Every's own skills (Compound Engineering, Compound Writing), and includes "Frontier Alerts" that proactively flag new model/tool releases relevant to a user's specific workflows.

The core argument is "prompt in public": most AI use is invisible (a person alone in a terminal), so coworkers never see the request, the corrections, or the iteration that produced a finished result -- only the magic trick's output, which nobody can copy. Running delegated work in a shared Slack channel makes the process itself the training material; a colleague who'd never touch a terminal can read the thread and think "I could ask for that." Every says this is why their earlier "Plus One" experiment (April, one personal agent per employee) is now fully retired in favor of one shared company agent: individual pet agents got personalized and then abandoned once the novelty wore off, while the single shared agent -- "AI-pills a company faster than a horde of personal ones" -- has been Every's daily-use internal tool for almost six months. The piece closes with internal usage anecdotes (contract drafting, PR review, exec reporting, time-off requests, even the company fantasy football league) and one beta-customer quote on "a shared capability that improves as we learn how to use it together."

Why this is in the vault

This is the shipped, GA version of the pivot Every first announced in May ([[2026-05-15-every-team-agents-vs-personal-pets]]): personal agent-pets don't scale as an org-wide deployment shape, a single shared role-owned agent does. That note was already load-bearing for RDCO's harness-engineering thesis; this one is the proof that the pivot held for five months of real internal use and is now a priced, GA product built explicitly on Claude Managed Agents.

Mapping against Ray Data Co

Strong mapping. The "prompt in public" mechanism -- visible delegation in a shared channel is what spreads AI adoption through an organization, not private 1:1 chat use -- is a sharper, more concrete articulation of the same shared-agent argument Every made in May, and it names the adoption mechanism (visibility → virality) that the May piece only implied. That mechanism is directly exploitable in RDCO's own agent-deployer pitch to phData-style prospects: the pitch isn't "give your team personal copilots," it's "stand up one visible, correctable, shared agent and let its thread history do the training." It also validates Ray's own always-on channels-agent setup ([[project_channels_agent_setup]]) as the right shape -- one persistent shared agent the founder corrects in-channel, not a fleet of personal instances -- though Ray's deployment is iMessage-only/single-user rather than a multi-seat Slack rollout, so the "spread virally across coworkers" half of the argument doesn't transfer directly; it's evidence for the deployment-shape thesis, not for RDCO's specific channel choice. Self-promotional origin noted below: Every is the vendor selling the thing it's describing, so the "almost indispensable after six months" framing should be read as a case study with an obvious incentive to be flattering, not neutral evidence.

Related

⚠️ Sponsorship

Every is announcing and selling its own new product (the Every Agent, $30/month + token pass-through). The entire piece -- including the "prompting in public" argument and the internal-usage anecdotes -- is written by Every's CEO to sell that product. Treat the adoption claims (six months indispensable, beta-customer testimonial, "AI-pills a company faster") as vendor case-study framing, not independent validation. No third-party sponsor is involved; this is self-promotion/structural sponsorship.