ai-six-sigma-property-os

AI Six Sigma Property OS

Build a practical operating model for property service quality using:

Ontology defines the business world.
Agent Team executes and audits workflows.
Six Sigma DMAIC continuously reduces errors, delay, rework, and cost.

This skill is for designing the management system before building software. It should produce executable operating structure: ontology, roles, CTQ metrics, work-order flow, database tables, dashboards, and MVP scope.

Usage Template

Prompt

Use ai-six-sigma-property-os for my Property Agent OS.
Design an AI + Ontology + DMAIC Black Belt operating model for property work orders, worker dispatch, environmental testing, quote generation, CRM follow-up, evidence upload, and quality dashboards.

Use Case

• Founder wants to turn messy property maintenance operations into a measurable AI workflow.
• Operator needs CTQ metrics, root-cause analysis, dispatch rules, quote controls, and evidence gates.
• Product team needs a first-stage MVP plan before building a full property SaaS.

Expected Result

• A practical operating memo with pyramid model, DMAIC loop, ontology objects, agent roles, CTQ scorecard, dashboard design, core tables, work-order states, control plan, and MVP roadmap.

Output Example

• MVP Stage 1: classify work orders, recommend workers, generate quote draft, require evidence upload, and track response time, completion time, rework rate, complaint rate, quote error, gross margin.

Verification Case

• Every module maps to at least one CTQ metric, data field, owner, human confirmation point, and control check.

Verified Effect

• A service workflow becomes a measurable quality flywheel instead of ad hoc manual coordination.

Success Metrics

• Defines the business objective and first-stage operating scope.
• Produces a DMAIC workflow tied to real property operations, not generic quality jargon.
• Names ontology objects, required fields, agent roles, CTQ metrics, dashboards, and work-order states.
• Separates AI recommendations from human approval for quotes, dispatch exceptions, safety, compliance, and customer-impacting decisions.
• Includes a narrow MVP roadmap focused on work orders, workers, quotes, evidence, and quality dashboard before expanding.

When to Use

• "Design my Property Agent OS."
• "Build an AI Six Sigma model for property maintenance."
• "Use DMAIC to improve dispatch, quote, and service quality."
• "Create CTQ metrics and dashboards for my work-order business."
• "Design agent roles for property, environmental testing, and CRM operations."

Operating Pyramid

Use this as the top-level model:

Business goals
  Reduce cost / raise speed / stabilize quality / make repeatable / support financing
      ↓
Six Sigma Black Belt layer
  DMAIC / data analysis / root cause / control plan
      ↓
Ontology semantic layer
  Customer / property / asset / work order / worker / route / quote / rule / evidence
      ↓
Agent Team execution layer
  Classify / dispatch / quote / audit / review / control
      ↓
Field operations
  Repair request / service / environmental test / payment / review

Core rule:

Ontology clarifies.
Agents execute and audit.
DMAIC improves the system after every work order.

Step 1: Define the Operating Scope

Classify the case before designing:

| Field | Options | |---|---| | Business type | property repair, maintenance, environmental testing, cleaning, inspection, CRM follow-up | | Stage | idea, manual pilot, spreadsheet MVP, internal tool, SaaS product | | First workflow | work-order classification, dispatch, quote, evidence, quality dashboard | | Human approval level | all decisions, quote only, exceptions only, mostly automated | | Data maturity | no data, sheets, CRM, database, integrated system |

Default MVP scope:

1. work-order classification
2. worker dispatch recommendation
3. quote draft generation
4. evidence upload and audit
5. quality dashboard

Do not expand into a full ERP, marketplace, payroll system, or finance system before this loop works.

Step 2: DMAIC Workflow

Map Six Sigma to the property workflow:

| DMAIC | Property OS use | |---|---| | Define | Define customer pain, work-order types, SLA, service standards, CTQ metrics | | Measure | Track response time, dispatch time, completion time, quote error, rework, complaint, evidence completeness | | Analyze | Find root causes for delay, wrong dispatch, missing evidence, wrong quote, low rating | | Improve | Update dispatch rules, quote rules, worker matching, SOPs, customer scripts | | Control | Use dashboards, alerts, approval gates, SOP audits, agent review, weekly Black Belt review |

Every work order should become a learning event:

Work order creates data
Data reveals problems
Problems trigger root-cause analysis
Root causes improve rules
Rules train agents
Agents improve speed and quality
More volume creates better data

Step 3: MECE Quality Domains

Score quality across seven non-overlapping domains:

| Domain | Controls | Core metrics | |---|---|---| | Customer quality | experience, response, satisfaction | first response time, satisfaction, complaint rate | | Work-order quality | classification, dispatch, completion, acceptance | first-time fix rate, rework rate, timeout rate | | Worker quality | skills, location, reliability, rating | on-time rate, completion rate, customer score | | Quote quality | accuracy, margin, approval | quote error rate, gross margin, close rate | | Process quality | end-to-end flow | cycle time, bottleneck, wait time | | Data quality | completeness, accuracy, traceability | missing field rate, missing photo rate, missing location rate | | Knowledge quality | SOPs, rules, lessons | SOP hit rate, rule update frequency, case review rate |

If a module has no metric, it is not ready for automation.

Step 4: Ontology Objects

Define the business world before defining agents.

Minimum ontology:

| Object | Required fields | |---|---| | Customer | id, name, contact, address, priority, consent flags | | Property | id, address, building, unit, access rules, manager | | Asset | id, property_id, type, model, location, maintenance history | | WorkOrder | id, customer_id, property_id, type, priority, SLA, status, description, evidence_required | | Worker | id, skills, service area, location, availability, rating, capacity | | Quote | id, work_order_id, labor, materials, travel, margin, approval_status | | Evidence | id, work_order_id, photos, video, signature, location, timestamps, checklist | | Rule | id, trigger, condition, action, owner, version | | Review | id, work_order_id, issue, root_cause, countermeasure, owner, due_date |

Add environmental testing fields only when needed:

| Object | Fields | |---|---| | TestResult | work_order_id, pollutant_type, device_id, value, unit, threshold, calibration_status | | Device | id, model, calibration_date, operator, status |

Step 5: Agent Team Roles

Use agents as a Black Belt team, not as unbounded autonomous decision-makers.

| Agent | Responsibility | Human gate | |---|---|---| | Define Agent | define problem, SLA, CTQ, work-order type | new service standard | | Measure Agent | calculate metrics and missing data | metric definition changes | | Analyze Agent | run 5 Why, fishbone, bottleneck analysis | root-cause acceptance | | Improve Agent | propose SOP, quote, dispatch, route improvements | policy and pricing changes | | Control Agent | monitor dashboard, alerts, control plan | closure of exceptions | | Dispatch Agent | recommend worker by skill, distance, availability, SLA | high-risk or low-confidence dispatch | | Quote Agent | draft quote and margin calculation | customer-facing quote approval | | Review Agent | detect wrong orders, missing evidence, abnormal quotes, risk | case closure and disciplinary action |

Step 6: CTQ Metrics

Use CTQ (Critical to Quality) as the operating scoreboard:

| Goal | Metrics | |---|---| | Fast | first response time, dispatch time, completion time | | Accurate | classification accuracy, dispatch accuracy, quote accuracy | | Stable | rework rate, complaint rate, timeout rate | | Cheap | cost per order, empty-run rate, material waste | | Profitable | gross margin, close rate, repeat purchase | | Controlled | evidence completeness, SOP execution rate, data completeness |

Set thresholds:

Green = acceptable
Yellow = watch
Red = root-cause review required

Step 7: Dashboard Design

Return four dashboards:

| Dashboard | Required widgets | |---|---| | Business | new orders today, completed orders today, revenue, gross margin, satisfaction, alerts | | Process | avg response time, avg dispatch time, avg completion time, overdue orders, rework, waiting materials | | Quality | classification accuracy, dispatch accuracy, quote error, evidence completeness, SOP hit rate, complaint rate | | Worker | location, active orders, completed orders, on-time rate, score, empty-run rate |

Every dashboard metric must trace to a table field.

Step 8: Root-Cause Analysis

For each red metric, use fishbone categories:

| Category | Questions | |---|---| | People | wrong skill, unclear customer service note, customer unavailable | | Machine | tool missing, test device failure, no system reminder | | Material | stockout, wrong part, unavailable consumable | | Method | unclear SOP, bad dispatch rule, quote approval too slow | | Environment | route too far, weather, access restriction | | Measurement | missing start time, no acceptance standard, incomplete data |

Then run 5 Why until the countermeasure changes a rule, SOP, field, training item, or control threshold.

Step 9: Work-Order State Flow

Use this default state machine:

submitted
  -> classified
  -> data_needed
  -> dispatch_recommended
  -> quote_drafted
  -> quote_approved
  -> assigned
  -> in_progress
  -> evidence_uploaded
  -> quality_review
  -> customer_acceptance
  -> closed

Exception states:

waiting_customer
waiting_material
worker_rejected
quote_rejected
rework_required
cancelled
escalated

Step 10: Output Format

Return:

## AI Six Sigma Property OS Memo

Business objective:
MVP scope:
Human approval boundaries:

### Pyramid Model
...

### DMAIC Workflow
| Phase | Action | Data | Owner | Gate |

### Ontology
| Object | Required fields | Why it matters |

### CTQ Scorecard
| Goal | Metric | Formula | Green | Yellow | Red |

### Agent Team
| Agent | Input | Action | Output | Human gate |

### Work-Order State Flow
...

### Core Tables
...

### Dashboards
...

### MVP Roadmap
Phase 1:
Phase 2:
Phase 3:

### Risks and Controls
...

Quality Gates

• [ ] MVP is limited to work orders, workers, quotes, evidence, and quality dashboard unless the user explicitly asks to expand.
• [ ] Each CTQ metric has a formula or data source.
• [ ] Each agent has a bounded input, output, and human gate.
• [ ] Quote, dispatch exception, safety, compliance, customer privacy, and payment-impacting steps keep human approval.
• [ ] Root-cause analysis ends in a rule, SOP, field, training item, threshold, or owner; not just commentary.
• [ ] The output separates current facts from assumptions and missing data.