GoodLeads.Club Platform
BRIEF The agent organization STATUS Live in production AS OF 2026-07-10
System Brief · How GoodLeads Ships

One equation.
An org that optimizes it.

GoodLeads ships product through an organization of AI agents — general managers, an engineering manager, specialists, rituals, memory, and budgets. One rule holds all of it together.

For the operator running outbound: records delivered before Apollo has them indexed, contacts who can actually authorize a purchase, and a quality floor that rises without a corresponding rise in cost. For anyone evaluating the architecture: what follows is how.

business growth = (sold × price) (build + sell + serve + support)

For the operator: fewer dead dials, more first conversations, leads nobody else has called yet. For the business: a cost structure that doesn't multiply with markets.

Every agent must name the variable it moves before work begins. If it can't, we don't take the work. That's how companies align thousands of employees they can't individually supervise — a P&L, not a rulebook. We applied it to agents from day one.

The Lever Gate

Work is admitted by variable

The equation isn't a poster on the wall. It's the admission gate. A proposal that can't name its variable gets sent back — by other agents. The moves that pass, by variable:

Variable · Sold

Sell more records

  • Fill the missing phone or email → a record a buyer can act on today
  • Improve the match rate from a filing to a real person → more of the inventory is sellable
  • Open a new market → more inventory in front of more buyers
Variable · Price

Make each record worth more

  • Add a classification or a score → one record now serves more use cases
  • Unlock a new use case for a group of records → the same inventory earns again
  • Verify deeper than anyone else will → the record commands a premium
Variable · Build

Build each record for less

  • Sweep a model agent's learnings into owned rules → the next thousand records classify for free
  • Find a cheaper source — or get closer to the original one → same record, lower build cost
  • Qualify with free data before spending on enrichment → paid effort only where it pays back

"Which variable does this move?" is the first question every agent asks — and the last gate before work ships.

The Org Chart

A classic matrix, run by agents

Business units across the top — one General Manager per market, owning how often its records sell. Departments down the side — one craft per row, serving every unit. A specialist sits at each intersection, answering to both. Every seat is an identity: a system prompt, a skill library, tool permissions, a memory scope — not whichever model happens to back it.

Context Infrastructure
Founder · above the matrix

One interface: an inbox and a doc

The entire matrix reaches him through a single authored briefing line — and that line doesn't grow when the matrix does.

Units →Departments ↓
Business unit · State GM

Colorado

Business unit · State GM

Florida

Opening · req posted

Next market

Platform Engineering
the EM — one department, every unit
Reviews every change cross-model · merges what's safe · ships to production · verifies & reverts · briefs the founder. One engineering department, amortized across all markets.
Classification
right code → right buyer
Identity & Names
the person who can authorize a buy
Address Intelligence
a filing becomes a reachable person
Enrichment & Scoring
channels found, verified, ranked
The Harness
memory · rituals · review · deploy
The Toolkits
direction, laddered to outcomes & KPIs
The Equation
every seat answers to a variable
The substrate every seat lands on — step in, inherit everything

Filled dot = the department serving that market · open dot = the posted seat.
Departments carry the craft across units; units carry the market. Specialists answer to both — the classic matrix trade, made cheap.

Add a market: a column. Add a craft: a row. Neither adds a line to the founder's day.

A seat opens like a job req

A new market — or a new craft — starts as a hiring spec, written against shared context infrastructure: the memory stores, the wiki, the toolkit ladder, the harness itself. Because all of that already exists, the spec is short and the start date is immediate.

Seat requisition · Business unit

State GM — Next Market

Identity
The State GM role: standups, findings, experiment proposals — the same seat, pointed at a new market
Context
A memory store seeded with the market's archetype, baselines, and filing patterns
Tools
Pipeline access · a monthly experiment envelope · the toolkit ladder
Onboarding
Reads the ladder, inherits the rituals — first standup the same morning
Filled by instantiation, not recruiting
Seat requisition · Department

Specialist — New Craft

Identity
A craft prompt plus its department's skill library and owned pattern library
Context
The wiki view for its domain — what every market has already taught us about the craft
Tools
An experiment envelope pointed at its variable of the equation
Onboarding
Same ladder, same rituals, same review gates — day one, every unit
Filled by instantiation, not recruiting
The Rituals

The org has a heartbeat

We didn't invent new ceremonies for agents. We gave them the ones that already work — every ritual a software company would recognize, run by agents, on a cadence.

The RitualThe Agent Org's Version
Standup Every State GM runs a morning and evening standup — reads its market, files findings, emails what moved.
Daily sync The EM authors a daily narrative to the founder's inbox — what moved, the calls it needs, what it's watching. Written, not templated.
Review & Retro Sprint reviews and retros are written artifacts, published to the same docs the founder already reads.
Board update The weekly recap reads like a VP of Engineering briefing the board — because the EM writes it in that voice.
Quarterly planning The toolkit ladder gets re-scored against the equation — bets re-ranked, scope re-cut.
The offsite Dedicated harness-debt paydown sessions: the org steps off roadmap work to fix how it works — retiring hardcoded rules it has outgrown.
Dailystandups ×2 · EM briefing
Weeklyboard recap · sweeps
Sprintreview · retro
Quarterlytoolkit refresh · offsite

Every ritual is a write to memory. Standups write what happened. Retros write how we work. Reviews write what we know. Cadence is how experience becomes an institution.

It's also why the founder can be out of the loop without being out of touch — synthesis reaches him on a schedule, instead of him going looking for it.

Four Kinds of Memory

Most stacks bolt on a vector store. We built four memories.

Because that's how organizations — and people — actually retain knowledge. Each layer has its own persistence, its own write cadence, and its own ritual that writes it.

← volatile · the sessionpersistent · indefinite →
Workingthe task at hand
Episodicwhat happened
Semanticwhat we know
Proceduralhow we do things
Layer 01

Working

Current task
Holds
briefing binder + session events
Written by
the session itself
Lifespan
minutes
Layer 02

Episodic

What happened
Holds
findings · founder comments · experiment results
Written by
standups & daily briefings
Lifespan
indefinite, timestamped
Layer 03

Semantic

What we know
Holds
the wiki — state × domain
Written by
weekly synthesis
Lifespan
living, versioned
Layer 04

Procedural

How we do things
Holds
skills — job-shaped playbooks
Written by
retros & reviewed evolution
Lifespan
durable, evolves by PR

What we know is a live table, not a build artifact. Edit the wiki and the agent's next session works from the new knowledge — versioned, auditable, no redeploy. Changing an agent's mind takes seconds, not a release cycle.

Sessions are ephemeral. Memory is durable.

Every agent session runs in a container that vanishes when the work ends. The agent's memory doesn't. Each seat carries its own persistent store — market archetype, baselines, patterns, every finding it ever filed — and it compounds. Fire the session, keep the employee.

Every week it works, an agent gets cheaper to run — and harder to replace.

The Experiment Economy

We didn't give the agents a to-do list. We gave them a budget.

Any agent can propose an experiment against its variable — and spend real vendor dollars running it, inside a monthly envelope enforced by the server, not by trust.

$ monthly envelope per agent · enforced server-side Propose agent names its variable Budget gate fits the envelope, or stops Simulate forward-replay on live context Ship win → config row Verify daily, against production Auto-revert retreat + file a finding on regression

The agent proposes, spends, ships, verifies — and retreats when production disagrees.

Real dollars

The envelope is real money against real vendor credits — skip-trace, gap-fill, validation — not simulated cost. An agent that wants to test a hypothesis pays for it like anyone else.

Simulation without drift

Variants forward-replay through the production context-builder — test and production are one code path, so experiments predict production. There is no separate test harness to drift.

Ships — and retreats — itself

A win writes an append-only config row. A daily verifier watches production metrics and auto-reverts regressions, filing an auditable finding for each retreat. No one has to notice.

Owned Assets

Models build their own replacement

When we use a model as a pipeline worker, the model call is a cost. What we sweep out of it is an asset.

Records every new filing Owned rules handle the known · free Model · rented resolves the uncertain tail Owned library rules · crosswalks · mappings uncertain tail only the sweep grows the rules next month: fewer calls

Deterministic rules handle what's known, free. The model gets only what they can't resolve.
Every resolution is captured — recurring patterns become owned rules that handle those cases forever.

The concrete example

Name intelligence and industry classification route only their uncertain tail to a model. Resolutions are captured; recurring patterns are swept into rules, crosswalks, and per-state mappings.

The falsifiable claim

Each pattern should cost roughly one model call, ever. The call rate on known patterns falls; per-record cost declines while quality holds. That's a curve we watch, not a slogan.

The same sweep, everywhere

Every brain gets the pattern — and so does the harness itself. Every hardcoded rule in the org is a marker for an answer an agent will eventually learn to discover on its own.

The model is rented. The library is owned.

The Founder's Surface

The whole org fits in an inbox and a doc

The founder never opens an engineering tool. His entire interface is email and Google Docs — the tools he'd be using anyway.

Inbound → to him

The EM authors a daily narrative in a VP's voice — today, what moved, the calls it needs, what it's watching. Decisions arrive as one-click buttons. A click records a durable, auditable decision; the agent executes the instruction it authored for that option. One click. The agent handles the rest — and auto-reverts if production disagrees.

Outbound ← from him

He comments on docs. That's it. The org mines those comments, clusters them, and promotes the recurring ones into formal judge scorers that grade future agent work. Management judgment, compiled.

The org meets the founder in his tools — and learns from what he says there.

The Toolkit Ladder

Direction is a ladder, not a backlog

Who decides what the agents work on? Strategy lives in a stack of toolkits, laddering from durable commitments down to this week's evidence. An idea descends only by earning it — and every level answers to the equation above it.

scope descends

Tenets

Years

Thirteen durable commitments — speed from public availability, confidence not certainty, unit economics. The constitution's articles.

Platform toolkit

Quarters

The strategy: what we're building, why it wins, what we're explicitly not doing.

Feature toolkits

Months

One per bet — the learning loop, commerce, the intelligence surfaces. Each carries its own outcomes and kill criteria.

Sprints

Weeks

Scoped, reviewed, retroed. The rung where ideas become shipped work.

Findings & experiments

Days

The evidence layer. What the org learned today — feeding everything above it.

evidence ascends

Every layer carries the same anatomy: the job to be done · the outcomes it targets · the KPIs that say it's working

Agents are first-class readers of the ladder. A proposal must cite the tenet or toolkit it serves — the equation gate, one level more specific. And new seats inherit the ladder on day one, which is how the org absorbs more agents without absorbing more chaos.

Governance

Autonomy is a gradient with a P&L

Every rung of autonomy the org climbs safely deletes a category of human cost from the equation. Here's how far it goes — and where everyone else stops.

1

Agents write the code.

2

Agents review each other — two model families, both green to merge. No same-family blind spots.

most agent deployments stop here
3

Agents merge without a human — tier-gated by what the change touches.

4

Merged code deploys itself to production. The merge is the only gate — keyless, immutable releases, self-healing rollback.

5

Agents verify their own shipments against production metrics, daily.

6

Agents revert themselves on regression — and file the finding that says why.

7

The org compiles the founder's judgment into formal evaluation criteria.

8

Agents propose changes to their own instructions — a structured case (why · evidence · target outcome), approved on the trust ladder.

The trust ladder, by tier

Tiers are computed mechanically from what a change touches. Fixable mechanical gaps don't block — the EM repairs them itself and re-reviews. The founder sees only what's worth his judgment — in plain English, in a doc.

C

Low risk

Auto-merge when both reviewers agree. The founder never sees these unless he asks.

B

Internal

Auto-merge, logged in the daily sweep.

A

Hot path

Auto-merge with ownership gates satisfied.

S

Product risk

Routes to the founder as a plain-English risk summary. A comment approves it. Rare by design.

Licensable · The Deliverable

The data platform

Everything a licensee's contract covers — pipeline, brains, API. The agent layer mounts behind a single feature flag: switched off, the org disappears and the data plane runs as if it never existed.

backend/pipeline/classification-service/
CI-enforced line
Retained IP · GoodLeads only

The organization

The GMs, the EM, the rituals, the memory, the learning loop. A licensee build literally subtracts this directory. A CI test fails the build on any import that crosses the line.

agents/

The commercial architecture: the deliverable is licensable because the org was never fused into it.

Scaling Economics

A new market is a config, not a hiring plan

Expansion means posting a seat req and filling it the same morning: a seeded memory store, a state configuration, a budget envelope. The horizontal layer — review, deploy, verification, the founder's briefing — is already amortized. It doesn't grow with markets.

5
Markets live · fetched from the production API on every page load
~$50
Monthly experiment envelope per agent · the marginal cost of curiosity
1
Horizontal layer, amortized across every market · it never multiplies
0
Engineers paged to keep the org running · quiet by design
Operational Evidence

Working today. Reload the page.

Not aspirations. The live-market count below is fetched from the production API on every load; the rest is stamped from the repository at publish time — never hand-typed history.

500+
Merged changes on main · majority agent-authored, auto-merged
24
Sprints shipped · reviews and retros written by the org
5
Markets live right now · fetching…
Standups per market per day · plus the EM's daily briefing

Repository figures as of 2026-07-10 · market count live from /api/v1/leads/states

What's hard to copy — and why

A competitor copying the surface ships in a sprint. The rows below each presuppose a commitment made months earlier.

ComponentDifficultyWhy
Alignment by business equation High Requires real per-record unit economics. Most agent stacks have no variable to optimize — their agents do plausible work; ours move a number.
Budgeted experiment economy High Real dollars, server-enforced envelopes, and a verifier that retreats. Most teams stop at "the agent suggests."
Forward-replay simulation High Variants share the production context-builder — not a simplified copy. Compromise this once and experiments stop predicting production.
Founder-comment → judge-scorer pipeline High Needs a founder who writes on a stable surface, a mining pipeline, and a trust gate. Feedback forms don't produce a corpus; a manager does.
Four memories + the rituals that write them Med-High The schema is a week of work. It's worthless without the cadence that writes to it — and the cadence is a culture, not a cron job.
Model-to-library sweep Med-High Anyone can call a model. Capturing its resolutions into an owned library that lowers next month's bill is a discipline, applied per brain.
Founder-from-inbox governance Medium Technically simple; culturally rare. Plain-English narratives and one-click decisions instead of dashboards is a commitment most teams won't make.
The deliverable boundary High A commercial-architecture choice enforced by CI. Most teams discover their IP is fused to their deliverable at licensing time — too late.
Cross-model review Low-Med Two integrations and an aggregator. Defensible only because it's shipped and load-bearing.