Roster

What it is

Every enterprise bought AI seats. None of them have a record of what those seats produced. Roster is a Claude-native plugin that turns AI access into a record of work — a personal team of specialists for every employee, dispatched from chat, tracked in a ledger, earned through real work.

It’s built on Claude Code’s plugin primitives: skills, subagents, slash commands, MCP. Deployed to your tenant. Anchored to Microsoft 365.

The problem

Generic AI gives every employee the same blank box.

Prompts without portfolios

Your seats are deployed. Your usage dashboard climbs. Nothing accumulates. The output of a thousand sessions evaporates between Mondays, and the per-seat invoice still arrives.

Role-blind tooling

A consultant, a designer, an engineer, and an executive all open the same chat box. None of it knows who they are, what they ship, or what their company expects from them. The AI is generic because the product is generic.

Training on a separate track

Your compliance and governance programs run on chase emails and quarterly nags. Your AI program runs on a separate budget. The two never reinforce each other. You pay for both.

How Roster works

A specialist team, dispatched from chat, accumulating in a ledger.

Personal roster

Every employee runs /build-roster once. Roster pulls identity, role, manager, and team from Microsoft 365 — no setup forms, no role pickers — and assembles a starting roster of three Claude subagents: Researcher, Editor, Coordinator. The hub opens in their browser the same day.

Role-aware. M365-native. Zero-config.

Dispatch and ledger

Work is routed to named specialists by intent. /dispatch "review this MSA" goes to the Reviewer. /dispatch "draft my Friday update" goes to the Coordinator. Every artifact the roster ships lands in a personal ledger — title, summary, tags, link, timestamp — that persists across sessions, projects, and roles.

Intent routing. Auditable output. Institutional memory.

Progression on real signals

Specialists unlock through work, not points. Reviewer comes online when the employee completes contract-risk training. Sentinel unlocks at the data-privacy module. Strategist unlocks after five items shipped. Compliance and capability run on the same engine.

Governance-wired. Composable. No badges.

Why it compounds

Two things make the work add up: Roster is wired into where the work already lives, and every roster is personalized to the employee it serves.

Connected, not copy-pasted

Specialists act on Microsoft 365 — calendar, mailbox, SharePoint, Teams, the documents your team already touches. A Coordinator drafts the Friday client update from the week’s meetings, the deck deltas, and the open action items — no template, no prompt engineering. A Reviewer flags a contract clause against the firm’s playbook and pastes the comment into the Word doc in-place. A Researcher pulls the last three engagements with a client and assembles the kickoff brief before the call.

Roster doesn’t return a chat reply for someone to paste. It takes the artifact off your plate and produces the next one.

Personalized by identity

Every roster is shaped by who the employee is. Role, manager, team, calendar, the work product they actually ship — pulled from Microsoft 365, not from a setup form. Two consultants on the same engagement get two different rosters, tuned to what each of them does. The specialist library is the same; the person it serves is not.

That’s why the ledger compounds. The longer an employee runs Roster, the more the system knows about how they work — and the more of their week the specialists can absorb.

Lab origin

We built Roster because we kept watching enterprise AI rollouts produce dashboards instead of artifacts. The work was happening — somewhere — but no organization could see it, audit it, or carry it across employees. Identity-aware, ledger-backed, governance-wired AI tooling was the missing layer.

We built it for one professional-services firm, then for the next, and now it ships as a product.

Status

Live and available. We onboard new customers with a four-week setup engagement and a 30-day pilot. Bring one department; we’ll show you the ledger.