Executive Briefing · Wasteology Engineering

Agentic Developer Workflow

Building software with AI, under human control.

Wasteology

The idea in one line

AI agents do the building. People approve every change.

  • Local-first. It runs on flat, unmetered developer subscriptions before any metered cloud spend.
  • Human-gated. Every change ships as a draft pull request that a person reviews and merges.

Why it matters

More built per engineer, without losing control

We get more output from the same team, without giving up control of quality, cost, or security.

And it compounds. Every project that adopts it makes the next one faster, because the company's knowledge is shared.

The process

The ADW loop

A task becomes working software through the same repeatable steps every time. It runs on demand or overnight.

Classify Plan &review plan Implement Independentcheck Commit Draft PR a human merges A task enters the queue the only step that is not an agent

Control

What keeps it safe

Independent checkA separate maker-checker verifies the work against the spec.
Evidence gateNothing is called "done" without proof it actually works.
Human mergeA person merges every change. Agents draft; people sign off.

Every run is traced end to end, so there is a full audit trail of what each agent did.

Maker-checker review Evidence gate Ontology guardrail Human merge on prod

Who decides what gets built

Work enters one queue

Tasks come from two places: an estate watcher that proposes work, or a person who adds one.

Nothing builds unless it enters that queue, so the team stays in charge of what gets worked on.

Why the output is trustworthy

Grounded in company knowledge

Agents read each project's own documented intent, plus a company-wide knowledge atlas, and build to the spec instead of guessing.

A broken or missing reference cannot pass the gate, so the documentation stays honest.

The next step for us

One terminal, many projects

A single hub oversees many projects at once. One place to watch every queue and every pull request, and to dispatch work to the right repository.

wg-orchestration one terminal for every queue and PR Project A · runs its own ADW loop Project B · runs its own ADW loop Project C · runs its own ADW loop Project D · runs its own ADW loop many builds in parallel · the hub picks the right execution model for each repo

Cost is a choice

Local-first by default

Day-to-day work runs on flat, unmetered subscriptions. The metered cloud path is opt-in, only for unattended work.

Even then, the throwaway steps use a cheap model, and spend is tracked from day one.

~98%

less cost on the routine commit-message step by using a cheaper model for it

The plan

From one engineer to a software factory

Three earned stages, gated by readiness, not by dates. Each stage is proven before the next begins.

1 · Learn and pilotOne team learns the loop and proves it on real work.
2 · Orchestrate and fan outOne hub runs many projects at once, across teams.
3 · Atlas and software factoryShared company knowledge makes every new project faster than the last.

Full plan and visual timeline live in the Academy roadmap.

Today

The foundation is already built

  • The Academy that teaches this workflow.
  • The standards and CI gates that enforce it.
  • The knowledge atlas that grounds it.
  • This reference orchestration hub, proving "many projects, one terminal."

Not a new bet

It's already on your roadmap

The Program Roadmap board tracks all 45 initiatives in this year's plan. This platform is already the delivery engine behind it.

27 of 45already building or deployed on this platform today.
8 initiativesalready have a named IT lead tagged – for example TJ, on the warehouse and semantic layer.
Every changeon the board is tracked the same way this briefing describes: drafted, reviewed, human-approved.

This is not a separate initiative to fund. It is how the roadmap already approved gets built.

The ask

Two decisions

  1. Which teams pilot first?
  2. At what pace?

We prove each rung before climbing to the next. Everything else is ready.

Local-first · Human-gated · Grounded · Auditable

Wasteology