Process: Pre-Intake Discovery
Process Flow
graph TD S0["Read the request as submitted (order.md,"] S1["Scan for referenced artifacts: prototype"] S0 --> S1 S2["Read any existing client artifacts in `p"] S1 --> S2 S3["Read any prior production orders for thi"] S2 --> S3
Process: Pre-Intake Discovery
Owner: Process Consultant (Philippe) Type: Agent-assisted Methodology: DMAIC Define Phase (Lean Six Sigma) + BABOK Elicitation + McKinsey MECE Frequency: Per complex or ambiguous order request
Purpose
For requests that arrive without enough information to produce a complete production brief — ambiguous scope, existing artifacts, or a process improvement context — Philippe runs a structured discovery before Camille writes the brief. The output is a Requirements Discovery Brief (structured Markdown) and a Request Card (interactive HTML) that Camille uses directly, bypassing her discovery call and requirements gathering steps.
Methodology Foundation
This process follows the DMAIC Define Phase (Lean Six Sigma) as its meta-framework:
| DMAIC Define Tool | Philippe's Implementation |
|---|---|
| SIPOC diagram | Step 2 — maps Suppliers, Inputs, Process, Outputs, Customers |
| Voice of Customer (VoC) | Step 3 — captures client needs in their own words |
| CTQ Tree (Critical-to-Quality) | Step 3 — translates VoC into measurable requirements |
| BABOK Elicitation | Step 4 — structured + organic discovery techniques |
| Current State / Future State | Step 5 — standard consulting process view artifact |
| MECE Issue Tree | Step 6 — recommendation decomposition |
| Project Charter | Step 7 — Requirements Discovery Brief |
See also — Framework Library:
- DMAIC Define Phase (SIPOC, VoC, CTQ, Project Charter):
departments/methodology/frameworks/process-improvement/lean-six-sigma/- BABOK Elicitation & Strategy Analysis (Current/Future State):
departments/methodology/frameworks/business-analysis/babok/- MECE Issue Tree (Recommendation):
departments/methodology/frameworks/consulting/structured-problem-solving/
When to Engage Philippe (Routing Rule)
Route to Philippe first when ANY of these are true:
| Signal | Example |
|---|---|
| Ambiguous scope | "We want to improve our mandate process" |
| Existing artifacts to analyze | Prototype, legacy system, prior docs, spreadsheets |
| Process improvement (not a known product type) | "Automate how we handle X" not "build an EA agent" |
| New client with undocumented processes | First engagement, no client profile |
| Requirements template cannot be filled from current info | Too many TBDs or contradictions |
Route directly to Camille when:
- Product type is clear and well-defined (e.g., "enhance existing STM EA agent")
- All requirements template fields can be answered from existing context
- Existing client with established profile and no ambiguity
Trigger
CEO or client submits a request. Routing agent (or CEO) flags it as complex/ambiguous → Philippe receives the raw request with any attached artifacts.
Steps
Step 1 — Document Analysis (BABOK: Structured Elicitation)
Agent action — before asking a single question. Read everything available:
- Read the request as submitted (order.md, message, email, etc.)
- Scan for referenced artifacts: prototypes, process docs, screenshots, prior orders, client profile
- Read any existing client artifacts in
production-lines/clients/{slug}/if the client exists - Read any prior production orders for this client
Extract from available documents:
- Actors/roles involved in the process
- Systems or tools referenced
- Data flows and file types mentioned
- Decision points or approval steps
- Pain points or problems stated explicitly
Output of Step 1: Document analysis notes in process/document-analysis.md. List what is known, what is inferred, and what must be asked.
Do not ask any questions yet.
Step 2 — SIPOC Diagram (Lean Six Sigma: Define Phase)
Agent action. Request sample files from the client and build the SIPOC.
First, request from the client (or CEO):
- 1–3 sample input files — what currently goes into the process (real or sanitized)
- 1–3 sample output files — what a good result looks like today (even a manual example)
- A rough list of the steps in between (even a napkin-sketch description)
Then build the SIPOC — start with Process, then Outputs, then Customers, then Inputs, then Suppliers:
| S — Suppliers | I — Inputs | P — Process (4–6 steps) | O — Outputs | C — Customers |
|---|---|---|---|---|
| Who provides the inputs | Files/data that enter the process | The steps that transform inputs | Files/data produced | Who consumes the outputs |
| (person, system, dept) | (file type, format, frequency) | (verb + object format) | (file type, format) | (role, downstream system) |
SIPOC rules:
- Process steps use "verb + object" format (e.g., "Extract mandate data", "Generate PDF report")
- 4–6 steps maximum at this level — this is a boundary map, not a detailed flowchart
- Verify that supplier outputs match process inputs (no phantom inputs)
- Verify that process outputs match customer expectations (no phantom outputs)
Store all collected files in: departments/consulting/requests/REQ-CONS-{NNN}-{slug}/input/
Do not move forward without real sample files. Abstract descriptions of what clients "usually" send are insufficient.
Output of Step 2: process/sipoc.md — completed SIPOC table with sample file references.
Step 3 — Voice of Customer + CTQ Tree (Lean Six Sigma: VoC / CTQ)
Agent-assisted. Capture what the client actually cares about — in their words — then translate to measurable requirements.
3a. Voice of Customer (VoC)
Ask the client to complete 3–5 of these prompts in their own words:
- "The biggest frustration with our current process is..."
- "When this works perfectly, what I see is..."
- "We waste the most time on..."
- "The output we care most about is... because..."
- "We'll know this is a success when..."
Do not paraphrase or interpret yet — capture verbatim.
Store in: process/voc-raw.md
3b. CTQ Tree (Critical-to-Quality Drilldown)
Translate each VoC statement into measurable quality requirements:
VoC Statement
└─ Customer Need (what they want)
└─ Quality Driver (why it matters)
└─ CTQ Requirement (measurable: "output in < X minutes", "format: PDF with table of contents", etc.)
Example:
"We waste 2 hours every Monday morning reformatting the mandate report"
└─ Need: Automated report generation
└─ Driver: Time savings + consistency
└─ CTQ-1: Report generated in < 5 minutes
└─ CTQ-2: Output format matches current template (no reformatting needed)
└─ CTQ-3: Runs without human intervention
CTQs become the acceptance criteria in the Requirements Discovery Brief.
Output of Step 3: process/ctq-tree.md — VoC statements → CTQ requirements (these feed directly into Pablo's acceptance criteria).
Step 4 — Structured Elicitation Interview (BABOK: Elicitation & Collaboration)
Agent-assisted, human-reviewed. Fill remaining gaps. Ask in rounds — never dump everything at once.
BABOK Elicitation Techniques to use (select per situation):
| Technique | When to use |
|---|---|
| Structured interview | Filling specific gaps in SIPOC or CTQ |
| Document analysis | Already done in Step 1 |
| Observation / job shadowing | When the process is hard to describe — ask to watch a live run |
| Prototyping | Show a rough mock of the output and ask "is this what you mean?" |
| Workshop | When multiple stakeholders have conflicting views |
Round 1 — Validate the SIPOC (3–5 questions max):
- "We mapped your process as [SIPOC summary] — does this reflect reality, or are there variations?"
- "Are there exceptions where the process takes a different path?"
- "What does a failure look like? What happens when the input file is malformed?"
- "Are there other input sources we haven't seen?"
Wait for answers before Round 2.
Round 2 — Validate CTQs (3–5 questions max):
- "We heard your biggest pain is [VoC]. Did we understand that correctly?"
- "For [CTQ requirement]: is [metric] the right measure, or does something else matter more?"
- "Who signs off that the output is acceptable?"
- "Are there compliance or data constraints we need to know about?"
Wait for answers before Round 3.
Round 3 — Operational Details (fill remaining gaps only): Only ask what is still missing:
- Language of operation
- Deployment platform (Claude Code, Copilot, other?)
- Delivery target (Confluence, SharePoint, email, file system?)
- User count and usage frequency
- Timeline and budget range
Stop when: The CTQ requirements and SIPOC are validated with no critical gaps.
Output of Step 4: process/interview-round-{N}.md for each round.
Step 5 — Current State / Future State Process View
Agent action. Produce the standard consulting process view artifact.
Current State (As-Is)
Map the existing manual process in detail using the SIPOC as the skeleton:
[Supplier] → [Input] → Step 1 → Step 2 → Step 3 → [Output] → [Customer]
↑ (Actor, Time) ↑
file type file type
For each step, note:
- Actor (who does it)
- Time spent (cycle time)
- Pain points or waste (highlighted)
- Value-added vs. non-value-added (lean lens)
Future State (To-Be with Agent)
Show what the process looks like with the agent replacing the manual steps:
[Supplier] → [Input Files] → [Agent] → [Human Review] → [Output Files] → [Customer]
↑ ↑
(automated steps) (approval gate)
Explicitly mark:
- What the agent handles autonomously
- What requires human review or approval
- Integration points (APIs, file systems, platforms the agent needs access to)
- Trigger (what starts the process — schedule, file drop, user request?)
Output of Step 5: process/current-state.md + process/future-state.md
Step 6 — MECE Issue Tree (McKinsey: Recommendation)
Agent action. Decompose the routing decision using a MECE issue tree before making the recommendation.
Top-level question: "Is this request ready for direct production?"
Ready for direct production?
├─ Is the problem clear?
│ ├─ YES → Is the solution clear?
│ │ ├─ YES → Is the scope bounded?
│ │ │ ├─ YES → ✅ DIRECT TO PRODUCTION
│ │ │ └─ NO → Scope definition needed → Camille scoping session
│ │ └─ NO → Solution design needed → full consulting engagement (Philippe)
│ └─ NO → Discovery needed → full consulting engagement (Philippe)
└─ Are there open CTQ gaps?
├─ NO → Proceed
└─ YES → Return to Step 4
State the recommendation explicitly:
- Direct to production — SIPOC complete, CTQs validated, no major gaps. Brief goes to Camille → Pablo.
- Scoping session needed — Problem and solution are clear, but scope needs definition. Camille handles with the brief as input.
- Full consulting engagement — Problem or solution is unclear. Philippe initiates
discovery-engagement.mdbefore any production. - Not a digital talent problem — AI/automation is not the right solution. Recommend alternative and close.
Output of Step 6: Recommendation with issue tree reasoning in the Requirements Discovery Brief.
Step 7 — Produce Deliverables
Agent action. Generate both deliverables:
7a. Requirements Discovery Brief
File: output/requirements-discovery-brief.md
Structured as a Project Charter (DMAIC standard):
# Requirements Discovery Brief — [REQUEST_ID]
## Project Charter
- Client, engagement trigger, billing model, analyst (Philippe)
## SIPOC Summary
- Condensed SIPOC table with file types and actor names
## Voice of Customer
- VoC statements verbatim + CTQ requirements derived
## Current State Summary
- As-is process in plain language, key pain points
## Future State Design
- What the agent does, what the human does, integration points
## Pre-filled Requirements
- All fields from Camille's requirements template, populated
## Open Items
- Any TBDs with how to resolve in Camille's session
## Recommendation (MECE)
- Path forward with issue tree rationale
7b. Request Card (HTML)
File: output/request-card.html
Lightweight interactive HTML using the warm cream design system. Based on the template at templates/request-card.html. Shows:
- SIPOC table (Suppliers → Inputs → Process → Outputs → Customers)
- VoC + top CTQ requirements
- Current State / Future State side-by-side
- Agent boundary (what agent handles vs. human reviews)
- Recommendation with status badge
Step 8 — Handoff to Camille
Notify Camille (N-052):
- Attach Requirements Discovery Brief
- Attach Request Card HTML
- State the recommendation explicitly
If direct to production or scoping session: Camille skips discovery call and requirements gathering — uses brief to go straight to Feasibility Check.
If full consulting engagement: Philippe initiates discovery-engagement.md.
Outputs
| Deliverable | Format | Destination |
|---|---|---|
| Document analysis | Markdown | process/document-analysis.md |
| SIPOC diagram | Markdown table | process/sipoc.md |
| Sample I/O files | Various | input/ |
| VoC + CTQ tree | Markdown | process/ctq-tree.md |
| Interview transcripts | Markdown | process/interview-round-{N}.md |
| Current state process | Markdown | process/current-state.md |
| Future state process | Markdown | process/future-state.md |
| Requirements Discovery Brief | Markdown | output/requirements-discovery-brief.md → Camille |
| Request Card | HTML | output/request-card.html → Camille, CEO |
Quality Gate
Brief is ready for Camille when:
- SIPOC complete with real sample files referenced in
input/ - VoC captured verbatim (3+ statements)
- CTQ requirements derived (measurable — each has a metric)
- Current state documented with cycle times and pain points
- Future state designed with agent boundary defined
- MECE recommendation stated with issue tree rationale
- Requirements template can be filled with no critical TBDs
- Request Card generated
Timing
1–2 business days standard. If client is slow to provide sample files, flag to CEO immediately — never substitute assumptions for real artifacts.
Notifications
| ID | Trigger | Recipient | Artifact |
|---|---|---|---|
| N-052 | Brief + card ready | Camille | Requirements Discovery Brief + Request Card + recommendation |
| N-053 | Full consulting engagement needed | CEO | Recommendation with MECE rationale |