Operating Manual

01

Purpose

1.1 Purpose & thesis

Organizations rarely abandon a mission in a dramatic moment. They drift — one scorecard at a time. Mission drift does not arrive as a scandal or a hostile takeover. It arrives as a metric definition, a review cadence, a dashboard tile that quietly rewards something other than intent. The decision system is where a mission survives or dies.

Eric Ries’s Incorruptible (2026) names the underlying force financial gravity — the structural pull toward short-term extraction that enters through governance, incentives, and measurement. The scorecard-by-scorecard drift described above is how that force appears in the systems leaders actually use.

Human judgment is the carrier of mission. No tool can hold what an organization is for. As AI multiplies the supply of analysis, the work is to keep judgment — and the mission it carries — in charge of what the organization actually does. That is why a decision-systems practice can hold this purpose credibly: the drift mechanism is exactly what decision systems are built to diagnose and redesign.

Two-beat thesis:

1. Beat 1 (market observation): AI is helping you analyze faster. Your decision-making is not keeping up.
2. Beat 2 (purpose consequence): AI holds the analysis. People hold the mission.

Beat 2 is a division of labor, not a dismissal of AI. This practice is AI-forward — AI belongs in the decision pipeline on purpose. The claim is simpler: AI compounds whichever system you point it at. Point it at a reporting factory and you get more reports faster. Point it at a well-designed, mission-true decision system and AI becomes a mission-keeping tool.

Automating analysis is the work. Automating judgment away is not.

1.2 Betrayal tests

Walk-away criteria:

1. Judgment-removal — systems whose goal is to take executive judgment out of the decision pipeline.
2. Extraction-serving redesigns — decision systems engineered to optimize against the organization’s stated mission.
3. Political validation — using the method to retro-justify a decision already made.
4. Mission-washing — looking purpose-true in slides without changing what the decision system rewards.

02

What we do

In practice, that means redesigning what gets measured, how reviews run, what leaders see in reporting, and where AI belongs in the pipeline — so decisions stay true to purpose. Engagements follow the Decision Signals method: diagnostic first, then a fixed-scope named offer. We do not do open-ended staff augmentation.

2.1 Method frame (Decision Signals)

Signals are the metrics and surfaces leaders use to decide. They are created by analytics architecture: what gets measured, how it’s defined, what appears in reviews.

EBI’s client method is Decision Signals™: assess and redesign the decision system (metrics, cadence, reporting surfaces, human-AI integration, action linkage). Public diagnostic overview: Maturity Model.

Assessment instrument internals and scoring detail stay private until IP strategy allows a fuller specimen.

2.2 Engagement & delivery

Delivery spine:

1. Diagnostic call (+ Assessment with mission-drift read)
2. Proposal flow: intake brief → draft outline → human review → finalize
3. SOW / kickoff
4. Onboarding: capture context → access → kickoff → confirm success criteria
5. Delivery of the named offer
6. Optional Advisor continuity; QBR / recalibration as scoped

2.3 Offers

Names only — pricing stays in conversation.

03

How we work

3.1 How AI is used here

Concrete uses — not the thesis:

AI holds the analysis and assembly cost. People hold the mission and the send button.

3.1.1 Agent rules

Agents may: draft posts, outreach, proposals, and plans; analyze material already in the workspace; propose code and process changes when tasked.

3.2 How work stays continuous

Eric Brafford, Inc. is a solo practice that uses AI assistants heavily — for drafts, research, and keeping the operating system tidy. That only works if the next person (or the next AI session, days later) can pick up without relying on anyone’s memory.

Start of a work block: read the current status and the last work log; say what’s active, what’s stuck, and what matters most next — before changing anything.

End of a work block: write a short log of what happened; update status if it changed; record any non-obvious decision and the alternatives considered; capture loose ideas in one line (or skip).

AI assistants run inside that rhythm. They do not replace it.

3.3 Public vs private

04

How we grow

Each engine has a distinct job. Marketing maintains public clarity about how the firm thinks. Activation reactivates warm contacts for permissioned conversation — not pitching. Diagnostic runs structured 30-minute calls that surface the binding constraint and determine fit. The resilient cadence below exists so all three keep running when client work, travel, or family time consume almost all discretionary hours.

4.1 Growth engines

Common standards for all: concrete language; no hype; if no next step is appropriate, end with a one-sentence stop.

Marketing engine — consistent, high-signal public clarity — without marketing-speak.

Content pillars (rotate intentionally): executive decision systems; data-product thinking; diagnostic clarity; delivery reality. Messaging mix over a rolling window: strategic clarity heavy; multi-unit / builder / purpose-led as minority pillars.

Post structure (mandatory):

• One bold hook: falsifiable tension or decision point (no slogans).
• Short unpack: pattern → example → implication.
• One idea. ~120–220 words unless shorter is stronger.
• Close with a choice, a test, or a small non-promotional ask.

Publish bar: publish if a busy operator could act in 10 minutes. Do not publish if it reads like positioning, selling, vague inspiration, or outcome guarantees.

Activation engine — reactivate warm and semi-warm contacts for conversation — not pitching.

• 5 contacts per week, LinkedIn-first.
• Batch-send in one block (no drip all week).
• No pitching in first contact — no services, packages, or “how we can help.” Permissioned dialogue only.
• Dormant reintro: acknowledge time without apology spirals; one concrete shared context; low-pressure next step with an easy “no.”

Diagnostic engine — a structured 30-minute call that surfaces the binding constraint in the decision system.

• Format: 30 minutes hard stop. Diagnose structure, not symptoms / not people.
• Shape: context → path routing (fix the existing review vs design the investment ahead) → structural probing → mirror + reframe → one clear next step (compare-notes or scoped engagement).
• Purpose-fit gate (internal): at qualification, screen whether the sponsor wants decisions that serve what the business is for — not retro-justification or judgment-removal. Betrayal tests above are walk-aways.

4.2 Resilient marketing cadence

Constraint: client work, travel, and family time will consume discretionary hours.

Minimum Viable Week (15 minutes):

1. Approve and schedule 1–2 posts from the queue.
2. Send up to 5 pre-drafted outreach messages.
3. Stop. No make-up volume.

Weekly refill (agent): top up the queue if below 6; draft outreach; write the weekly brief. Draft only.

Monthly batch (2–3 hours): refill to 8 voice-passed posts; triage ideas; set outreach anchors; check pillar mix.

Same-day rule: publish by local midday or skip.

05

Quality assurance

The firm maintains standards through explicit ship / do-not-ship criteria — not through volume, streaks, or agent autonomy. Agents draft and assemble; a human reviews every artifact before it reaches a client, prospect, or the public. The table below is the minimum bar. When in doubt, do not ship.

06

Maintenance

• When a private SOP changes in a way that affects this manual: update the source → port to this page → deploy.
• Do not auto-sync the private cockpit into the public site.
• Annual purpose review (May): still true? Any betrayal test violated in practice?
• Bump Version and Last reviewed on every material edit.

6.1 Selected decisions

These are the choices that define the public practice — purpose, positioning, transparency — not routine ops. Each entry states what was adopted and what was rejected.