Company strategy in one gated AI session
Not a consultant prompt — twelve layers with gates, Point A confirmed by leadership, strategic mode selection, and a 90-day plan that stays in project memory.
Discuss this post in AI
Send a pre-filled prompt to ChatGPT, Claude, Gemini, or Perplexity — get a summary, ask follow-ups, or compare ideas from this guide.
A typical “help me write a strategy” prompt produces fluent text with headings — and zero obligation to your numbers, your market, or your people. It is the internet average, nicely formatted. A gated strategy session works differently: sequential layers, explicit gates, and no advance to the next step until leadership confirms the previous one. Neuro OS runs that sequence as versioned roles in one project — analyst, strategist, facilitator — with Point A, decisions, and a 90-day plan that survive after the meeting ends.
Not a prompt — session architecture
The methodology collects your data first, stress-tests it for contradictions, shows you a numeric portrait of the business, and requires confirmation that it is true — only then does it discuss strategy.
You get a facilitated session where agents work in three roles:
- Analyst — gathers documents, reconciles numbers, flags gaps, compares your metrics to industry benchmarks where sources allow.
- Strategist — maps market arenas, names trade-offs, scores options, proposes a mode of action suited to predictability and resources.
- Facilitator — keeps order, prevents skipping gates, surfaces what leadership feels but has not articulated, returns focus when the room drifts.
Authority stays human. Agents propose and expose; gates require an accountable person to confirm, dissent, or reject before the repository advances.
Four outcomes in one session
1 — The owner frame
A dedicated interview block covers where your time goes, which decisions only you can make, what you still do by hand, what stops if you disappear, and which business question you want answered on demand. This is not therapy — it is operating context for what to surface, warn about, or omit later.
2 — Business understanding, tested
Not “what industry are you in” but how money is made: segments, channels, margin vs loss pools, headcount, data locations, broken processes, and their cost. Understanding is verified: five control questions where mistakes are easy — the analyst answers as an informed insider; leadership scores each 1–5. Below five-of-five, the session does not proceed. That blocks polite agreement.
3 — Point A: today in numbers
A standalone deliverable: economics with margin by direction, funnel with conversion at each step, people and payroll share of revenue, capacity and utilization, cash and obligations, where data lives, what works and what is broken. Every figure tagged: from report, from interview, calculated, estimated. Most owners see this consolidated view for the first time — not because they cannot count, but because nobody assembled one sheet.
4 — Vision, values, and measurable goals
Future picture emerges from conversation, not a “what is your mission” prompt. Values come from hard choices: what you would fire a good performer for, what you will not do for money. Annual goals are testable with numbers — yes or no in twelve months. Optional three-year horizon if you want it. Ninety-day plan by month with owners, dates, and done criteria per action.
Two beats: facts, then strategy
Beat one — facts only. No goals, plans, or advice on this beat — even if asked.
-
1
Calibration
Three questions on size and state. Depth adapts: a ten-person company is not asked about a board audit committee. Crisis mode is named explicitly — survival before elegant structure.
-
2
Data passport
List of what to export across money, customers, people, product, systems, documents — each item with why it matters and where it usually lives. Upload exports, screenshots, sheet links, or voice notes.
-
3
Interview blocks
Five blocks of three to five questions. After each block, the analyst shows what it already sees — including non-obvious patterns. Nothing already in documents is re-asked.
-
4
Verification pass
Arithmetic consistency, cross-source agreement, plausibility for the niche, period alignment, critical gaps. Conflicts are shown — not averaged — and leadership picks which source wins.
-
5
Point A gate
Full portrait presented. Session advances only when an accountable human confirms Point A is correct — or records explicit dissent in the decision log.
Beat two — strategy. After the gate: market map with boundaries and dual-sized TAM checks, competitor frame, field signals (reviews, hiring, long-running ads), diagnosis and strategic mode, one-page business model, values and goals, function/role schema for this stage, 90-day plan by function, memory digest and starter prompts for ongoing work.
Choosing how to strategize
From BCG’s strategic approach work: pick how to strategize before picking what to do. Wrong mode is worse than no mode. Three questions — predictability in twelve months, ability to shape industry rules, resource tightness — select one of five modes:
| Mode | When it fits | What it means |
|---|---|---|
| Be big | Predictable market, rules you cannot change | Planning works; scale and cost advantage win |
| Be fast | Unpredictable market | Long plans fail; run cheap rapid tests |
| Be first | Predictable and you can change rules | Window to create something new; bet on one move |
| Shape | Unpredictable but you can influence | Cannot win alone; assemble partners |
| Survive | Resources exhausted | Stabilize first; strategy second |
The strategist role documents why the mode fits Point A and market evidence — not which slide looked best.
Built-in credibility rules
- No figure enters strategy without source and date.
- Sources older than eighteen months are marked stale.
- Conflicting sources appear side by side — the gap is information.
- Paid industry reports tend to inflate market size; triangulate independently.
- For your company, truth is documents and leadership — not scraped bios.
Acceptance gates
- Business understanding: five-of-five on control questions
- Point A: explicit human confirmation in decision log
- Annual goals: leadership sign-off with dissent recorded if any
- CRM / roadmap / calendar writes: Ask before execution
The system is designed so agents cannot ship a strategy leadership did not accept.
What you get on exit
| Artifact | Purpose |
|---|---|
| Point A | Numeric portrait with source per line — valuable alone |
| Market map | Boundaries, size, dynamics, competitor table, benchmarks |
| Strategic mode | Plan vs experiment — with rationale |
| One-page business model | Who, what, how we earn, moat, what breaks at 2× growth |
| Vision and values | In your words — not template slogans |
| Annual goals | Testable numbers; optional three-year horizon |
| Function / role schema | Roles needed at this stage — not textbook org chart |
| 90-day plan | By month: owner, date, done criterion per action |
| Project memory digest | Short Layer-3 summary for every future agent run |
| Starter prompts | One per major initiative — paste to resume work in context |
| Register | Table: action, deadline, owner, status, done signal |
What happens after the session
Traditional consulting ends with a PDF in a folder. A gated Neuro OS session ends with confirmed context in the company repository — reusable by every role.
- Decisions stay in frame. Hiring, discounting, new lines — answers reference Point A, goals, and non-goals — not generic advice.
- One picture for the team. Policy and project memory in git mean sales, finance, and marketing run inside the same confirmed strategy — because the environment enforces it, not because everyone read a deck.
- Plans become work. Starters are launch prompts: new run, paste starter, agent already knows the initiative, deadline, and done criterion.
- Context compounds. Each follow-up conversation stands on verified base — compare to three instruction layers and corporate context architecture.
How to run it in Neuro OS
-
1
Create a strategy project
Evidence folder, decision log, source register — separate from day-to-day chat. One repo, scoped connectors.
-
2
Attach documents you have
P&L by month, pipeline export, headcount with payroll. Missing items are requested on the data passport step — estimates are labeled estimates.
-
3
Run beat one to the Point A gate
Do not skip to market options. Confirm or dissent in the decision log — that record is the hardest gate.
-
4
Run beat two through 90-day plan
Small portfolio: one core bet, one or two tests, explicit non-goals. Writes to CRM or calendar default to Ask.
-
5
Promote digest to project memory
Point A summary, goals, values → Layer 3. Schedule weekly review skill to compare indicators vs thresholds.
-
6
Keep assumptions that expire
Tag claims with review dates so the next session knows what needs fresh evidence.
To scope the project with Pattern Automation, get started.
Before you start
Time: forty minutes for a sub-ten-person company; two to three hours for a larger one. Split across days — gates preserve state in the project repo.
Ideal inputs: twelve-month P&L by month, customer/deal export, headcount with payroll. No documents? Session still runs — more fields marked estimated, stated plainly.
What to remember
- Strategy is a gated decision process — not a one-shot prompt.
- Point A confirmed by people before market work begins.
- Three agent roles; authority stays with leadership at every gate.
- Outputs live in git — digest, plan, register — not in a forgotten chat.
- Wrong strategic mode hurts more than no strategy; choose mode first.
Do we need documents to begin?
No. Without them, more numbers are estimates — labeled as such. With P&L, pipeline, and payroll export, the analyst guesses less and verifies more.
Can we pause and resume?
Yes. Keep one Neuro OS project so evidence, gates, and decision log persist. Return to the last confirmed gate — not a blank thread.
Why does the session refuse to continue sometimes?
That is a gate, not a failure. Strategy on a wrong Point A is entirely wrong — cheaper to stop than to execute a polished fiction.
Is this “AI wrote our strategy”?
No. Agents did analyst, strategist, and facilitator work on the org chart. People confirmed facts, chose trade-offs, and approved writes. The value is an auditable state for the next quarter — not slide filler.
How does this relate to corporate AI rollout?
This session produces the Layer-3 memory and 90-day plan that corporate context architecture expects — confirmed facts leadership owns, not chat improvisation.