Aimeice
AI-Native Delivery · Embedded Client
UA
Client AI executes Gate Auto-release

The client, inside the pipeline

SDLC 2.0 Policy: clients create tasks directly in the pipeline, and low-risk work ships to production without waiting for a human — the human gate isn't removed, it's replaced by two AI gates. Risk stays contained by access boundaries, hard exclusions and auto-rollback.

Status · Draft Owner · Delivery Lead Model · Embedded Client
Navigate with or press Space
What changed

Baseline standard → Embedded Client

The pipeline is unchanged — seven phases, human accountability. What changed is who files the task and what replaces the human gate on the low-risk track

Baseline · low-risk waits for a human at three gates days intake waits for PM build waits for review QA waits for release Fast-track · the human gate is replaced by two AI gates hours client Loops Cyrus AI Review AI QA prod
Schematic: segment length = phase duration · waiting on a human · the gain comes from removing queues, not from skipping checks
AI-Native Delivery · baseline
Embedded Client · this edition
The team files tasks; the client writes in chat
The client files tasks directly, against a Definition of Ready
Every PR waits for a human at the gate
risk:low — two AI gates instead of a human gate, auto-merge and auto-deploy
Risk is judged by whoever writes the task
Loops applies the risk:* label within 15 min; the requester sees it but cannot change it
Repository access is all-or-nothing
Repo classes: client-editable / guarded / restricted
Multi-repo work is coordinated in the lead's head
Parent + sub-issues with repo:* labels and blocking relations
Policy principles

Four rules that keep the speed safe

The goal is to make the client a full participant in delivery without handing over control of risk

01

The client defines WHAT, the pipeline defines HOW

The requester brings the problem and the acceptance criteria. Route, assignee and speed are decided by the system — not negotiated in chat

02

Automation never changes scope

Automation moves a task inside approved scope and only changes its state. A new requirement is a new task, not an agent quietly widening the current one

03

Speed replaces the gate — it doesn't skip it

Fast-track drops no checks. It swaps a human gate for an AI gate: code review and QA stay mandatory, they just run in minutes

04

"Cannot harm production or data"

Client access is bounded by that principle. Anything where a mistake breaks data, contracts or payments stays behind the pipeline

Overview · routes

One intake. Three tracks by risk

Every task enters the same way — and splits after the risk label. Click a track to see its rule

risk:low Fast-track UI, copy, styling, content, feature config without data, isolated bugs
ClientLoopsCyrusAI ReviewAI QAAuto-mergeProd30 min telemetry
There is no human gate as long as both AI gates return APPROVE. Deploys happen only inside the guard window, with auto-rollback on an error spike
risk:medium Standard track Business logic, API changes within contract, new components
ClientLoopsBuildAI ReviewHuman gate on PRQARelease
AI review still runs, but a human authorises the merge. This is the default track for anything not explicitly low
risk:high Human plan + staged release Migrations, auth, payments, external integrations, cross-repo contracts
ClientLoopsHuman planBuildAI ReviewHuman gateHuman QAStaged release
A human writes the plan before any code. The release is staged, with its own rollback plan. Everything caught by hard exclusions lands here automatically
Loops labels risk within 15 minutes of a task passing DoR — that is the pipeline SLA
Pipeline

Seven phases — pick one

The route is identical for every client and repository. Colour shows who owns the phase; on fast-track, the gates of phases 04 and 06 are executed by AI

Roles

Who owns what

Six roles, four of them automated. Humans stay where a decision is needed, not where execution is

RoleTypeResponsibility
Client (Requester)humanCreates the task, provides description, files and context, accepts the result
Linear + LoopsAIAssesses task risk, applies risk:low / medium / high, routes the assignee and the affected repositories
Cyrus · AI EngineerAIImplements risk:low tasks, opens a PR within a single repository
AI Code Review · CodeRabbitAI gateGate one: code quality, security, regressions
AI QAAI gateGate two: acceptance criteria verified on staging
Delivery Lead / Tech LeadhumanHuman gate for risk:medium+, escalations, fast-track incident retros, repo class changes
Rule The generator and the reviewer are different agents with different lenses. Cyrus never reviews its own PR: the same agent carries the same blind spots
Intake phase

Definition of Ready — the entry ticket

A task is valid for the pipeline once all five conditions are met. Try it: toggle the items and watch what happens

0 / 5
Task is not ready
pngjpgmp4pdf mdcsvjsonloghar ziprar7z
Archives are not allowed: an agent must read attachments directly, without unpacking
Consequence A task failing DoR is automatically labelled needs-info and returned to the client with a list of what is missing. Loops does not assess risk for tasks in needs-info — the SLA clock has not started
Risk assessment

How a task earns its route

Loops analyses the description, the attachments and the areas of the codebase involved. Hard exclusions are checked before scoring — and override any verdict the model produces

New client task DoR passed? no needs-info → client amended yes Hard exclusion zone? yes never risk:low auto-escalation no Loops: risk score risk:low fast-track · 2 AI gates risk:medium human gate on PR risk:high human plan + staged
Dashed line — forced escalation: a hard exclusion overrides the model's score
Appeal The requester sees the label but cannot change it. Disagreement means a comment on the task, decided by the Delivery Lead. That removes bargaining over speed from chat
Interactive

Try it: which track would your task get

Pick what the change touches — the label, the route and the reasoning appear on the right

What the change touches
Where
Hard exclusions

Six zones that are never low-risk

Regardless of what Loops scores, and regardless of how small the change looks

01

Data and schemas

Migrations and any data schema change

02

Access

Authentication, authorisation, roles, permissions

03

Money

Payment flows and billing

04

Data deletion

Deleting or bulk-modifying data

05

Infrastructure

Secrets, environment variables, CI/CD config

06

Contracts

API contract changes between repositories

Safeguard The rule also applies after scoring: if a PR from Cyrus touches files in these areas, the pipeline automatically raises the task to risk:medium and removes auto-deploy. Risk is decided by the diff, not only by the description
Fast-track · risk:low

A task's path from client to production

Press Play to watch the task move. Switch on "Gate fails" and the track hands it back to a human

8 steps · both AI gates must return APPROVE
CHANGES REQUESTED Cyrus self-fix · up to 5 iterations Human gate · Delivery Lead
Ready to run
Fast-track · rules

Four conditions the track cannot run without

01

Both gates say APPROVE

Any CHANGES REQUESTED from AI Review or FAIL from AI QA immediately moves the task to a human gate. There is no partial pass

02

Maximum 5 self-fix iterations

Cyrus gets five attempts to address AI Review findings. The next failure escalates to a human. The cap catches tasks where the problem is the brief, not the code

03

Guard window

Deploys happen only during the team's working hours. A PR ready at 23:40 waits until morning — so there is someone to respond to an incident

04

Automatic rollback trigger

An error spike in Sentry or a failing health check rolls the release back and opens an incident task. The rollback decision does not wait for a human

Cyrus · self-fix after AI Review findings 1 2 3 4 5 escalation to a human no longer fast-track
Safeguards

When we deploy, and what happens when it breaks

Guard window · the team's day PR at night → waits for morning working hours · deploys allowed 000408 24 Outside the window the PR is not blocked — only its release is
Telemetry · first 30 minutes after a fast-track deploy alert threshold · error rate deploy auto-rollback + incident task The rollback does not wait for a human · the post-mortem is its own task in the pipeline
Consequence Every fast-track release has an address to roll back to and 30 minutes of monitoring with Slack alerts. Speed without a rollback isn't speed, it's exposure
Multi-repo

The client writes the problem. The pipeline cuts it by repository

Cyrus works within one repository per task. Cross-repo work is therefore decomposed into sub-issues carrying repo:* labels — the client never has to think about repositories

Parent · coordination and e2e acceptance
The client's problem
Never assigned to Cyrus · holds the contract, e2e QA and acceptance
repo:shopify-plugin sub-issue own risk label · own PR
repo:frontend-app sub-issue own risk label · own PR
repo:demo-frontend sub-issue own risk label · own PR
LabelRepositoryClass
repo:marketing-siteMarketing website (static)client-editable
repo:shopify-pluginShopify plugin + backend APIrestricted
repo:frontend-appFrontend app (dashboard)guarded
repo:demo-shopify-pluginDemo shopify pluginguarded
repo:demo-frontendDemo frontendguarded
Risk is scored per sub-issue, but a fast-track sub-issue only ships when it is not blocked by dependencies
The contract rule

The one human gate that is never delegated to AI

Cyrus in two different repos shares no context. Without a fixed contract it will generate two incompatible halves. Switch the scenario:

Demo repos and acceptance

Where we can go faster, and how a parent closes

demo Demo ships faster

demo-shopify-plugin and demo-frontend hold no production data and no real payments
Their sub-issues are fast-track candidates by default, even when the mirrored change in the main repo scores risk:medium
Syncing demo with the main product is filed as separate sub-issues of the same parent, not as "and also fix the demo while you're there"
Hard exclusions still apply: demo does not lift the ban on auth, payments and data schemas

QA Two-level acceptance

A sub-issue closes on its own acceptance criteria — AI QA verifies within its repo
The parent closes only after e2e AI QA on staging, once every sub-issue is released
If e2e fails while all sub-issues are green, it is almost always a contract error
Such a parent escalates to a human and is not handed back to Cyrus: the agent will fix its half and break the other
Fast-track boundary A parent with a cross-repo contract change never runs fully automatically — at least one human gate is always present. Full auto-release is possible only for multi-repo work that leaves contracts untouched
Client access boundaries

Least privilege, generated together with the pipeline

Access isn't handed out "on request in chat" — it is part of the project configuration

ResourceClient accessRationale
Linear (client project)Create / comment / readThe primary channel of participation
risk:* labelsRead onlyRisk is decided by the system, not the requester
StagingFull UI accessVerification before and after AI QA
Preview deploymentsReadEarly inspection of a PR
ProductionAs a product user onlyNo deployment admin panels
GitHub / repositoriesOwner; write level is per-repoThe client owns the repos; direct-change boundaries are set per repository
CI/CD, infrastructureNo direct access to configurationAimeice's area of responsibility, including rollback mechanisms
Sentry / telemetryDigest in SlackThe data may contain PII
Project Slack channelMemberStatus and release notifications
Per-repo model

The question isn't ownership — it's whether direct edits are safe

The client owns the repositories. Each repo is assigned a class when the pipeline is generated. Pick a class:

Operations · flow rules

Fast-track only removes the queue for low-risk

Medium and high tasks still hit human gates. Without WIP limits the pipeline simply piles up queues in front of them — and the fast-track gain is eaten by everything else waiting

No limits WIP limits + aging alert 8 6 9 In Review QA Deploy 2 3 4 In Review QA Deploy
Tasks queued in front of human gates · dashed line — WIP limit (≤ 2 / ≤ 3 / ≤ 5)
01

Limits only on human phases

In Review ≤ 2 per dev · QA ≤ 3 per engineer · Ready to Deploy ≤ 5 per team. AI phases carry no limits

02

Cap per agent

Cyrus carries at most 3–5 build tasks at once. "Queue 30 tasks overnight" means a review backlog and shallow gates

03

Aging alert instead of standup

A task sitting over 24h in front of a gate pings the gate owner in Slack. The queue is a metric you can see

Evidence

What we measure on fast-track

A weekly metrics review is an Aimeice commitment under this policy. You see the same numbers we do

% auto-released
Share of tasks reaching production with no human touch. The direct measure of the track working
autonomy
Rollback rate
Share of fast-track releases rolled back automatically. Above 5% a month triggers a review of the risk:low criteria
safety
Escalation rate
How many risk:low tasks fell back to a human gate. A high number means Loops under-scores risk, or DoR is weak
calibration
Time to risk label
From passing DoR to the Loops label. SLA — within 15 minutes
sla
Low-risk cycle time
From the client filing the task to release. The headline predictability measure
end-to-end
Escape defect rate
Defects reaching production despite both AI gates. The main argument for or against widening the track
quality
Commitments

What we guarantee, and what the client takes on

Aimeice Tech

Pipeline response SLA: Loops scores a task within 15 minutes of it passing DoR
Transparency: the client sees the status of every task and the result of both AI gates in Linear
Weekly fast-track metrics review: % auto-released, rollback rate, escalation rate
The rollback mechanism for client-editable repos is set up by us — and rolling back is free

Client

Follows the DoR and does not bypass the pipeline with direct requests to the team
Accepts that the system assigns the risk label. Appeals are a comment on the task, decided by the Delivery Lead
Understands that fast-track applies only to risk:low
In client-editable repos owns the consequences of their own changes; in guarded and restricted, changes go through the pipeline only
Why Bypassing the pipeline doesn't break bureaucracy, it breaks telemetry: a task outside Linear has no risk label, no gates and no address to roll back to
Policy review

The policy carries its own safeguard

Fast-track is not a permanent arrangement. Drag the slider — it is the monthly rollback rate:

2.0% fast-track rollback rate · per month
0%5% threshold12%
01

Quarterly review

The policy is reviewed every quarter regardless of metrics. The risk:low criteria narrow or widen based on actual data

02

Review after an incident

Every production incident caused by a fast-track release triggers a retro and a policy review. No exceptions, no "it was a one-off"

Principles · 1 / 2

Eight rules holding it all together

Phases are the mechanics. Principles are the reasons. They separate a system from "we sprinkled AI everywhere"

01

Spec-first

The spec matters more than the code. DoR and acceptance criteria are approved before work starts. A vague spec means AI filling gaps with guesses about a business it doesn't know

02

AGENTS.md in every repo

One file with conventions, stack and do/don't for the agent. A de-facto standard in 2026. Cheap, and it raises quality immediately

03

Route by risk, not by effort

A one-line change in billing is high-risk. Auth / payments / migrations / PII always go to a human, however few the lines

04

AI reviews, but never itself

Generator and reviewer are different agents with different lenses. The same agent carries the same blind spots

Principles · 2 / 2

Eight rules holding it all together

Plus the one clients value most: you see the process, not only the result

05

Gates don't disappear — they change hands

On fast-track, review and QA are executed by AI, but they remain mandatory. A skipped gate isn't a fast track, it's the absence of process

06

Least privilege by repo class

Access is decided not by job title or ownership, but by "what happens to production if this goes wrong here"

07

Every auto-release has an address to roll back to

Rollback trigger, tag of the last stable release, incident task. Automation without a rollback isn't automation

Transparency by default

AI gate results, deploys and metrics are visible to the client in Linear and Slack in real time. You watch the task move through the pipeline

Difference

A typical services company vs Aimeice

This is not "the same process, but faster". It is a different operating model

Traditional IT services
Aimeice · Embedded Client
The client writes in chat; the task gets lost between people
The client files the task directly in the pipeline; status is always visible
A small fix waits for the next sprint
Low-risk ships to production in hours through two AI gates
"Fast" means "without review"
Fast means "review runs in minutes, executed by AI"
Access is handed out ad hoc
Repo classes and least privilege are part of the project configuration
Releases are manual and stressful
Guard window, auto-rollback, 30-min monitoring
The process policy lives in the lead's head
The policy has metrics and an automatic review trigger
Rollout · 1 / 2

Week 1 — boundaries, access, intake

Goal: the client can file a task, and the system knows what to do with it

Mon Tue Wed Thu Fri 1 2 3 4 5 6
Indicative week schedule · row numbers match the plan items below
1
Repository classification Delivery Lead · 1 day
Every repo gets a class — client-editable / guarded / restricted — and a repo:* label. Branch protection on guarded, direct changes blocked on restricted
2
Rollback mechanism for client-editable Team Lead · 0.5 day
Tag of the last stable release + rollback workflow, or instant revert through the host. Plus a build check before deploy so a broken build can't take the site down
3
Least-privilege access Team Lead · 1 day
Linear, staging, previews, Slack — per the access matrix. risk:* labels read-only for the client. Any extension is recorded in the pipeline README
4
DoR task template + needs-info automation PM · 1 day
Template with mandatory AC, the list of allowed attachment formats, and an auto needs-info label listing exactly what is missing
5
Loops: risk classification + hard exclusions Team Lead · 1–2 days
Rules for risk:low / medium / high, the list of hard-exclusion paths in every repo, auto-escalation based on the PR diff
6
AGENTS.md + Slack structure Devs + PM · ~2h/repo
Conventions, stack, high-risk zones, how to run the tests. Channels #deploys, #ai-pipeline, and a separate client channel
Rollout · 2 / 2

Week 2 — the first auto-release

Goal: a client task runs the track end to end, and fast-track metrics start accumulating

Mon Tue Wed Thu Fri 1 2 3 4 5 6
Indicative week schedule · row numbers match the plan items below
1
Pilot: 3–5 risk:low tasks from the client Client + Devs
Real tasks written by the client themselves — to test the DoR against live wording, not against examples
2
Two AI gates on the PR Team Lead
CodeRabbit as gate 1, AI QA on staging against the acceptance criteria as gate 2. Both must APPROVE for auto-merge
3
Guard window + auto-rollback Team Lead
Deploy window, Sentry trigger on an error spike, health check, automatic incident task creation
4
A multi-repo case with a contract Devs
One parent task with sub-issues and a fixed contract — to exercise blocking relations before an urgent case shows up
5
Fast-track metrics PM · 30 min/week
% auto-released, rollback rate, escalation rate, time to risk label. The baseline we later automate into a dashboard
6
Retro and risk:low calibration Delivery Lead + Client
What fell back to a human gate and why. Narrow or widen the fast-track criteria on the facts of week one
Deferred Automated client digests, a metrics dashboard, canary / blue-green — the next iteration. Two weeks give a working fast-track on the low-risk flow; we widen it on metrics, not on feelings
What you get

You're inside.
Not in the queue.

You file the task yourself Low-risk in production within hours Gates and rollback stay in place

We don't sell developer hours. We sell a measurable delivery cycle in which the client is a participant, not a customer in a queue.

Aimeice

Ready to build together?

Let's work out which of your repositories could go fast-track this month

Model 01 / 29