Skip to main content

Platform

Organize

Agents arrive without an organization around them: nobody can say who owns what they do or who decides when they are wrong.

What you get

Your org chart, seats and who decides, as you run today.

A seat is a role a person and an agent hold together; the escalation matrix says who decides above each limit, seat by seat.

  • The org chart as it runs

    In ClrSlate

    Your organization modelled the way it runs today: its functions, teams and seats, not a new way of working.

  • A seat for a person and an agent

    In ClrSlate

    Every seat on the org chart is designed for two owners: a person who answers for it and an agent that does its work.

  • Decision rights, per job

    In ClrSlate

    Who decides what, up to where, and above that whom, written down before the work starts.

    The person who prepares a decision is never the person who takes it: the agent prepares, the owner decides, and the record shows both.

  • The escalation matrix

    In ClrSlate

    A decision above a limit goes to the named seat that holds it, up your chain of command; a vacant seat is skipped to the next one up.

  • How much an agent may decide is yours to set

    In ClrSlate

    Per job and per function, from asking every time to deciding within the limits you write down, with anything above going up your escalation matrix.

  • A blueprint that runs on ClrSlate alone

    In ClrSlate

    For a company with no stack yet: the same organization model, with the platform's own requests, board, knowledge and model gateway doing the work until there is a stack to connect.

  • Blueprints to start from

    In ClrSlate

    Org models, golden paths and playbooks that show how AI-native organizations are structured, to start from and grow into.

  • Procedures with an owner and a version

    In ClrSlate

    The procedures you already follow, with an owner, numbered steps and versions, run by the agent with you at the step that decides.

  • Roles and functions to start from

    In ClrSlate

    Functions with their roles, the jobs they do and who decides, ready on day one and shaped to how you run.

How it looksat work.

Cropped to what matters, at Halden and Corvane, fictional companies.

Organization change · decision record DEC-3105

The platform team, split in two

  • Kubernetes platform teamNamespaces, capacity, clustersDmitri Volkov
  • Cloud platform teamProjects and the landing zoneVacant · Dmitri covers
An organization change: seats move with their reporting lines; a vacant seat goes to the next one up.

Decision rights · version 7

Who decides what, and up to where

  • Merge to mainService owner, up to changes inside the serviceAbove: Head of platform
  • Database schema changeHead of platform, up to additive and reversibleAbove: CTO
  • Dependency upgradesService owner, up to minor versions, checks passAgent within limits
Decision rights: who decides what, up to where, and above that whom.

Blueprints

Start from a blueprint,shaped to how you run.

Organization models, approved templates (golden paths) and playbooks, each with where the decisions sit.

A golden path is an approved template; a practice is the team that owns a standard.

  • Org model · Scaling and enterpriseProduct delivery organizationAn engineering department for a software company that has adopted coding agents: product teams that own products, practices that own the standards, and every review gate a seat with its owner and its limits.
  • Org model · Scaling and enterprisePlatform and operations as self-servicePlatform, DevOps and operations offered as requests: practices own the standards, platform teams run the shared platforms, and a product team asks once for what it needs.
  • Org model · Scaling and enterpriseSecurity and identity for people and agentsSecurity, identity and model access run for agents the way they run for people: the same request, the same policy, the same owners, the same record.
  • Golden path · Scaling and enterpriseRelease golden pathFrom a change to production through gates that are seats: reviewed against the engineering standard, cleared by Quality on its evidence, approved by the owner of the release, deployed in the window, watched against its objectives.
  • Golden path · Scaling and enterpriseNew environment golden pathOne request creates an environment a product team owns: its identity, its cluster boundary, its delivery destination and its pipeline environment, each part from the practice or platform team that owns it.
  • Golden path · Scaling and enterpriseWeb API golden pathOne path from a new API to production: repository, identity, namespace and deployment, each provided by the practice that owns it.
  • Playbook · Founding, scaling and enterpriseRun a strategic initiativeOne brief becomes workstreams with owner seats from the org chart, work on the right teams’ backlogs, a daily roll-up to the sponsor, and a run captured so the next initiative starts from a playbook.
  • Playbook · Founding, scaling and enterpriseA coding agent joins a seatAn agent joins the organization the way a new hire does: a seat with an owner, the jobs it may do, the decisions it may not take, the models it may use, and the systems it may reach.
  • Playbook · Scaling and enterpriseChange the organizationTeams onboarded, merged, split or renamed and practices added as jobs with an owner, a procedure and a record, with seats, knowledge and rules moving with them.

From one team to the whole center

Start with the one piece of work that breaks at the handoffs, model the seats it crosses, run it, and let the model grow team by team from what runs.

Blueprint · Founding and scaling

Runs on ClrSlate alone

No integrations needed to begin. For a company with no stack yet, or one that does not want one: the same organization model, seats and decision rights as every other blueprint, with the platform’s own requests, backlogs and a board, knowledge, documentation and model gateway doing the work until there is a stack to connect. One way into the platform, not a second product.

With it: go-to-market for a founder-led company

How it is laid out

  • The org chart as the company runs it: the founder’s office, product, engineering, marketing and sales, each seat a person and an agent. At founding one person often holds several seats; the chart says which.
  • The work in the platform itself: requests with an owner, backlogs and a board, and a morning brief on the founder’s desk.
  • Knowledge that stays with the seat: documentation, a wiki and what each agent learns, kept with the seat, not with the person’s inbox.
  • Agents that call models through one gateway, with the models each seat may use.
  • Room to connect: when the company adopts a tool, it joins the seat that already does the work, with its rules and its record.

Where the decisions sit

  • Anything published in the founder’s nameDecides: the founder, every time——
  • Positioning and launch messagingDecides: the product marketing managerUp to: the approved positioningAbove that: the CEO
  • A piece fits the brand planDecides: the brand managerUp to: the plan as approvedAbove that: the founder
  • A new seat, for a person or an agentDecides: the founderUp to: the roles the company runs—
  • Which models a seat may use, with what dataDecides: the founderUp to: the models the company approved—
  • Work inside a seat’s limitsDecides: the seat’s personUp to: the limits written in the seatAbove that: the founder

The dial. The founder’s name asks every time and stays so. Drafting, designing, scheduling, qualifying and preparing run within limits from the first day; the founder widens a seat’s limits from the record, one job at a time.

How an organization adopts it

Start with the threads the founder is in most. Model the three seats they cross (product marketing, brand, sales development), give each its agent, and route everything in the founder’s name through one approval. When the first tool arrives, a repository host or a CRM, it connects to the seat that already does the work; nothing is rebuilt, and the record carries on.

Start with howyou run today.

Bring one piece of work that breaks at the handoffs: a pull request waiting on three reviewers, an initiative spread across twelve tools, a release nobody owns. We model the seats it crosses, give one of them an agent, and follow one decision to its record.