01 · Readiness
Without F0 measurement, every downstream AI prescription is opinion.
Garbage in, garbage forbidden. The Data Quality Scorecard refuses to ship.
In plain words: if we don't know what's actually broken, every fix is a guess.
02 · Method
SEOS, end to end.
SEOS — Scale Efficiency Operating System. One foundation (F0). Seven agentic functions (F1–F7). Five validation gates (V0–V4). Built operator-first, on five scaled companies, before it was ever offered to a client.
In plain words: F0 is the diagnostic — what's actually broken before we prescribe a fix. F1–F7 are seven engines that ask the hard questions one at a time. V0–V4 are five gates that refuse to let the system ship a guess.
F0 · Foundation Builder
The diagnostic. Before any prescription.
A physician does not prescribe before diagnosing. SEOS does not scale what it has not measured. F0 is the act of measurement — the unglamorous infrastructure that makes every downstream claim verifiable. It is also the moat.
- F0.1
Connector Library
Heterogeneous app stack → unified ingestion. CRM · billing · support · product telemetry · ads. Adapter graph that ages well.
- F0.2
AI Schema Mapper
Heterogeneous schemas → canonical model. Backbone: EPIC-27 L2 reuse · ChurnDefense.
- F0.3
Identity Resolution
Cross-source entity de-duplication. One customer, one ID, every system.
- F0.4
Data Quality Scorecard
Completeness · freshness · lineage. Output: client readiness 0–100. Below 80, no F1–F7.
- F0.5
Per-Tenant Isolation
Cortex-class architecture. Each client is its own dataset, not a row.
F1 → F7 · Agentic Intelligence
Seven engines. One scoring instrument.
Each function has a single job, a single output, and a single backtest gate against Grupo Impact's outcome receipts across five active companies.
- F1
ICP-Retention Correlation
Which customer profiles drive the high-retention slice? Anchor target: replicate Chatsac's 93%.
Source: Cortex calculators + 26 signals
- F2
Revenue Line Dissector
Current LOBs, margins, growth velocity, ceiling diagnosis. Keep · expand · kill · launch.
Source: Financial deep-dive + cohort velocity
- F3
Four-Fit Audit
Market-Product · Product-Channel · Channel-Model · Model-Market alignment. Reforge framework, automated.
Source: Balfour · AIOS-CS @analyst
- F4
Scaling Readiness Score
Composite 0–100 with 3–5 leading indicators. Single number per company. Backtested against Chatsac's trajectory.
Source: Roberge · backtested formula
- F5
AI-GTM Mapper
Which functions migrate to agentic vs stay human. Expected ARR impact + payback months per migration.
Source: AIOS-CS 12-persona catalog
- F6
Bottleneck Ranker
Top 3 interventions, ranked by leverage × probability. Numeric predictions attached to each.
Source: Reforge growth loops
- F7
Backtest Engine
Every recommendation produces numeric prediction + payback. Quarterly reconciliation against actuals. The anti-guru spine.
Source: Roberge gate · Chatsac 4yr
V0 → V4 · Validation Gates
If a gate fails, no productization.
GateDayNameCriterionFail consequence
V0D7F0 Foundation GateData quality score ≥ 80, identity resolution complete, schema canonicalized.No F1–F7 execution. Engagement stops.
V1D14Backtest GateF7 reconciles all recommendations against historical actuals ± 25%.Recommendations are opinion. Revoke prescription.
V2D21ReCity Case-0 GateDiagnostic locked + Case-0 documentation begun.Cannot productize the engagement.
V3D45Hit-Rate GateNumeric predictions exceed naive baseline by ≥ 15 percentage points.Models reset to F4 + F6 re-run.
V4D90Absence TestCould the client have arrived here without SEOS? Honest answer must be no.Engagement does not graduate. No Welsh-rung promotion.
Loss aversion
What you cannot do without it.
02 · Willingness to change
Without V1 Backtest Gate, recommendations never reconcile against actuals.
Theater compounds. Quarterly hit-rate scorecard pre-empts it.
In plain words: if we never check whether the AI's predictions came true, we're selling fortune-telling.
03 · Leadership + processes
Without F3 Four-Fit Audit, a perfect AI rollout lands on a misaligned GTM.
Market-Product · Product-Channel · Channel-Model · Model-Market. All four, in order.
In plain words: a perfect AI on the wrong market still loses. Sequence matters.
Engagement
Three rungs. Layered, not stacked.
Welsh rung #1
$150
Diagnostic kit
F0 dry-run on your data. Quality score 0–100. Single-call walk-through. Outcome: do you pass V0 or not.
Welsh rung #2
$1.5K–$2.5K
Diagnostic sprint
F0 build + F1–F7 once on real data. Lock diagnostic v1. Outcome: a SEOS-scored map you can take to a board.
Welsh rung #3
$15K–$50K
Engagement
Quarterly cycle: F7 backtest + F6 ranking + intervention plan + V3/V4 gates. Outcome: receipts you reconcile.
The difference was never the AI model. It was always the organization. Its readiness, its processes, its leadership, its willingness to change and fail.
Brynjolfsson Digital Economy Lab · April 2026 · 51 enterprises