AESOP Studio — Executive Intake to AI Adoption Roadmap

Complete process map: 19-field executive form → 4-stage pipeline → 5-page founder-ready roadmap. Every heuristic, knowledge base, rubric, and guidance document is shown at its point of injection.

Pipeline at a Glance

Intake Fields

Processing Stages

Knowledge Docs

~1,900

Target Output Words

Pipeline Flow

Stage 1

Intake Processing

19 fields → 4 org_context rows
Confidence scoring + gap analysis

smb_intake_executive_service.py

▸

Stage 2

Prompt Engine

~220-line system prompt
11 tenets, 5 principles, 14 rules

adoption_roadmap_agent_executive.txt

▸

Stage 3

LLM Generation

Claude Sonnet 4.5
Validation loop (up to 2 retries)

adoption_roadmap_agent_executive.py

▸

Stage 4

Post-Processing

Confidence merge + phase extraction
Source manifest + email notification

adoption_roadmap_executive_service.py

detailed stage breakdown

Stage 1 — Intake Processing

19 Fields Across 4 Sections

About Your Business (5): business_name, industry, employee_count, what_they_do, what_makes_different

What's Working, What's Not (5): biggest_bottleneck, time_consuming_tasks, success_definition, top_priorities, business_impact

How Work Gets Done (4): founder_role, team_description, delegation_gap, decision_style

Tools & Guardrails (5): essential_tools, document_storage, compliance_requirements, investment_comfort, non_negotiables

4 org_context Rows Created

company_profile — Business Snapshot
strategic_goals — Goals & Ambition
smb_operations — How Work Gets Done
smb_constraints — Constraints & Guardrails

Confidence Scoring Engine
4 sections × 25 points each = 100 max (capped at 95)
19 individual field-level completeness checks
Field fill ratio → section score → total score
Level: high ≥75, medium ≥50, low <50

gap analysis fires for every missing field

Gap Analysis — 19 Field-Specific Impact Statements + 19 Cohort Exercises

Plain-Language Impact

Each missing field gets a specific, plain-language statement of what it affects. Example: "Without the founder's role, the roadmap cannot identify delegation opportunities." Every one of 19 fields has a unique, hand-authored impact — not a template.

Cohort Exercise

Each missing field maps to a concrete, week-numbered exercise the founder can do with their team. Example for "Delegation gaps": "Week 2 exercise: For one week, every time you do something someone else could handle, write it down."

Confidence Summary

Three-tier plain-language summary based on score tier. High: "The roadmap is well-grounded." Medium: "Direction is sound but recommendations may shift." Low: "A starting point for conversation, not a finished plan."

org_context + confidence → prompt engine

Stage 2 — The Prompt Engine adoption_roadmap_agent_executive.txt v3.0

Core IP: A ~220-line system prompt containing embedded knowledge frameworks, design principles, output schema, and generation rules. This is not a generic "write a roadmap" prompt — it encodes a complete methodology for honest, evidence-grounded AI adoption guidance.

Embedded Knowledge Frameworks

AI Maturity Model (5 × 5)

5 Levels: AI Aware → AI Ready → AI Capable → AI Mature → AI Native
5 Dimensions: Information Readiness, Process Clarity, Decision Governance, Team Capability, Tool Foundation
Bottleneck Principle: Overall maturity = lowest dimension score. Prevents cherry-picking the strongest dimension to inflate assessment.
Backed by: ai-maturity-model.md with full scoring protocol

The Eleven Tenets (AESOP Methodology)

Philosophical operating principles from aesop-methodology.md:
1. State Change 2. Problems Before Solutions 3. Evidence Over Eloquence
4. Know What You Need 5. People Are the Center 6. Humans Decide
7. Multiple Perspectives 8. Context and Brevity 9. Guardrails Not Gates
10. Trace the Connections 11. The Questions Stay the Same

5 Non-Negotiable Design Principles

CLEAR

Skimmable at 6am. Short words, short sentences. Under 20 words per sentence. Business language, not academic.

CONCISE

5 pages maximum. Every paragraph must earn its place. Substantial, detailed, not padded.

ACTIONABLE

Every section points to something they can DO. No passive observation without a practical implication.

NO JARGON

Banned: LLM, API, vector database, embedding, token, agent, platform, ecosystem, solution. Write for a business founder.

NO ASSUMPTIONS — The Hardest Rule

Only say what the intake data supports. If they didn't tell you about their customers, don't describe their customers. If the data is thin, output shorter, more tentative guidance. Before every claim: "Did they tell us this?" If not, don't write it or explicitly place it as a confidence gap.

Writing Discipline — Banned Words List

significant particularly exceptionally demonstrates comprehensive sophisticated leverage synergy paradigm automation solution platform ecosystem integration implementation

Output Validation Schema (The Rubric)

maturity_snapshot

• current_level + current_level_number
• bottleneck_dimension + bottleneck_why (2-3 sentences)
• bottleneck_what_this_means (2-3 sentences)
• 4 stage_gaps (1→2, 2→3, 3→4, 4→5)
— each: whats_missing + what_it_unlocks
• 3 confidence_gaps
— each: gap name, why_it_matters, how_to_close

movements[] + Insights

• 3 movements: name, rough_timeframe, focus, why_first
  — guidance: 4-5 practical bullet points
  — what_to_avoid: 2-3 specific anti-patterns
  — risk_signal: one observable warning
• the_honest_truth: 2-3 uncomfortable sentences
• the_one_move: single highest-leverage 30-day action
• what_changes: 3-4 tangible outcomes

14 Generation Rules (Self-Enforced)

Quality Constraints

1. Target ~1,900 words (1,600–2,200 range)
2. Maturity snapshot gets real word count — each gap is 2-3 sentences, not bullets
3. Exactly 3 movements, each with detailed guidance
4. Exactly 3 confidence gaps ranked by roadmap impact
5. No assumptions: 80%+ claims must be directly traceable to intake fields

Voice Constraints

6. Mirror without announcing — never say "you told us" or "from your intake"
7. The honest truth: direct, not cruel, grounded in their specifics
8. What changes: [Person] can [do thing] without [current bottleneck] pattern
9. What to avoid: concrete anti-patterns, never generic advice
10. Risk signals: observable warnings, not abstract risks

Guardrails

11. Banned: specific products, AI model names, internal scaffolding language
12. Thin intake (<12 of 19 fields) → shorter, more tentative output
13. Self-check before output: count claims, scan for jargon, check sentence length, verify all sections present
14. why_first must never begin with "Why this phase comes" — substantive content first

prompt + context → LLM generation

Stage 3 — LLM Generation + Validation Loop

Model Configuration

Model: Claude Sonnet 4.5 (via Anthropic API, streaming)
Temperature: 0.2 (low — factual, consistent output)
Max Tokens: 32,000 (room for full roadmap)
Provider: Configurable — Anthropic or DeepSeek
Agent: AdoptionRoadmapAgentExecutive

User Message Construction

Builds structured prompt from:
• 4 org_context sections (preferring content_summary, falling back to full content)
• Confidence assessment (score, level, known gaps)
• If total chars > 30,000: warns about model input budget
• Appended: "Produce an executive AI adoption roadmap as JSON. See the system prompt for the exact output format and generation rules."

Structured Validation Loop (Up to 2 Retries)

Attempt 1

LLM generates full roadmap JSON

→

Parse JSON

Extract JSON (handles markdown wrapping, regex fallback)

→

Validate

Check 20+ schema rules:
maturity fields, 4 gaps, 3 confidence gaps,
3 movements, honest truth, one move,
what changes (3+), word count range

→

Failed?

Specific violation list returned to LLM with retry prompt

→

Retry (×2 max)

Full multi-turn conversation. Each retry: "$0.02-0.04" additional cost.
Warnings attached if still failing after 2 retries.

validated JSON → post-processing

Stage 4 — Post-Processing & Delivery

Phase Extraction

Movements array → structured phase objects with phase_number, name, timeframe, focus, why_first, guidance (preserved as bullet list), what_to_avoid, risk_signal. Timeline computed from first/last movement timeframes.

Confidence Merging

Heuristic score (field completeness) adjusted downward by LLM-identified gaps. Each LLM gap deducts 5 points (min 5, max 40). Produces base_score + adjusted_score + adjusted_level + merged summary.

Executive Insights Extraction

the_one_move, what_changes[], the_honest_truth, stage_gaps, confidence_gaps → executive_insights JSONB column. Separates v3.0 fields from legacy columns for forward compatibility.

Source Manifest — Full Provenance

org_context: document_count + document_types + provenance ("org_context table, is_active=true, all types")

confidence: score, level, gap_count, provenance ("compute_confidence_executive() — field completeness from 19-field executive intake")

agent: name, temperature, prompt_file, max_tokens, input_tokens_used, output_tokens_used, cost

validation: warnings list + warning_count

Output — What the Founder Receives

Maturity Snapshot

Current level + bottleneck + 4 stage gap maps

3 Movements

Practical phases with guidance, anti-patterns, risk signals

Honest Truth

2-3 uncomfortable sentences grounded in their data

One Move

Single highest-leverage action for the next 30 days

persisted to DB + email notification sent

Knowledge Base — All Documents That Inform the Process

11 reference documents in backend/engine/knowledge/ provide the methodological foundation. The prompt engine synthesizes these into a coherent, context-aware output — but the documents exist independently as versioned, auditable knowledge assets that can be updated without retraining or rebuilding the pipeline.

aesop-methodology.md

The Eleven Tenets, autonomy classification (4 levels), build summary artifacts (5 types)

ai-maturity-model.md

5 levels × 5 dimensions, bottleneck principle, full scoring protocol with evidence requirements

prd-decision-framework.md

Assistant vs Agent classification, 8-dimension scoring model, decision tree, hard gates

feasibility-rules.md

Platform connector registry, constraint rules, trigger compatibility matrix, file format support

hybrid-architecture-patterns.md

9 architecture patterns for combining deterministic + non-deterministic logic, anti-patterns

discovery/triage-questions.md

7 triage questions, red flags, complexity signals (simple/medium/complex), routing rules

discovery/capability-thresholds.md

7 capability domains with quantitative breaking points, migration paths, cost implications

discovery/project-failures.md

7 common failure patterns with early warning signs and red flag signals

discovery/trade-off-frameworks.md

Decision frameworks for scope, quality, and speed trade-offs in AI adoption

discovery/question-banks.md

Structured question banks for discovery interviews across multiple domains

discovery/platform-capabilities.md

Platform-specific capability matrices and constraints for build targeting

Why the Output Is Quality — 10 Rigor Mechanisms

Multi-layered confidence: Heuristic field completeness score → adjusted by LLM-identified gaps → merged confidence with provenance. No single source of truth about quality.
Bottleneck principle: Overall maturity = weakest dimension. Prevents cherry-picking the strongest area to inflate assessment. Backed by explicit scoring protocol in ai-maturity-model.md.
Structured validation with retries: 20+ schema rules checked programmatically. Up to 2 automatic retries with specific violation feedback. Warnings attached if still failing.
Evidence traceability: 80%+ claims must be directly traceable to specific intake fields. Self-check enforced in the prompt. "Did they tell us this?" before every claim.
No jargon enforcement: 15+ banned words prevent generic AI consulting language. Write for someone who runs a business, not a technology department.
Mirror without announcing: References their specifics naturally without analysis-speak ("you told us," "from your intake"). They should feel seen, not analyzed.
Compass not GPS: Direction with enough texture for Monday morning, but not false precision. 3 movements, not 12 steps. Risk signals, not risk matrices.
Gap → impact → exercise chain: Every missing field produces: (a) plain-language impact, (b) concrete cohort exercise with week number. Actionable, not diagnostic.
Full source provenance: Source manifest records every input: org_context types and counts, confidence provenance, agent config, token usage, cost, validation warnings.
Eleven Tenets foundation: Philosophical rigor from aesop-methodology.md — People Are the Center, Evidence Over Eloquence, Guardrails Not Gates, Trace the Connections — ensures every output embodies a coherent methodology, not ad-hoc heuristics.

Architecture note: This is the executive path — a lightweight variant of the full AESOP Build Pipeline. The full pipeline adds: readiness assessment, project portfolio with strategic prioritization, department-level context, multi-agent quality analysis (instructions, ethos, bias, safety, legal, infosec), interview phase, custom rubric + scenario generation, adversarial red-team (Cave of Shadows), functional testing execution, and synthesis reporting. The executive path is designed for small business founders who need a focused, honest, actionable roadmap — not an enterprise-scale analysis.