Frame
Convert the brief into a sharp problem statement, hypotheses, and decision rights. Pick the right framework set.
Run by · Strategist Agent Pass 1
02 Methodology
MBB rigor as a guard rail.
Three AI agents move work through five MBB phases bundled into three agent passes, each gated by a 12-criterion scorecard and a human checkpoint. Every claim is traceable to a framework, a source, and a reviewer. No hourly billing, no junior analyst guesswork, no proprietary black box.
5 phases · 3 agent passes3 specialised agents23+ frameworks12-criterion scorecardHuman-in-the-loop
Five phases. Three agent passes.
Decompose
Issue tree built MECE. Each branch tied to a question, an analytical method, and the data needed to answer it.
Run by · Strategist Agent Pass 1
Quantify
Financial models, scenarios, and sensitivities. DCF, unit economics, market sizing — fully traceable to inputs.
Run by · Financial Modeler Agent Pass 2
Synthesize
Pyramid Principle write-up. Lead with the answer, support with three pillars, evidence the bottom row.
Run by · Strategist + Modeler Agent Pass 1
Review & Ship
12-criterion scorecard, Red Team challenges, Eval Suite checks, board-ready exhibits. Reviewer signs off before delivery.
Run by · Quality Reviewer + Human Agent Pass 3
Three agents. One brief.
Strategist
Owns framing, issue trees, and synthesis. Picks the right consulting frameworks for the question and writes the storyline.
- MECE issue trees
- Pyramid Principle synthesis
- Framework selection & rationale
Stack · Model Selector (Sonnet 4.6 · Opus 4.7 · GPT-5.2/5.4/5.5) + RAG
Financial Modeler
Builds the numbers. Three-statement models, DCF, unit economics, scenario sensitivities. Every cell tied to an assumption.
- DCF + sensitivity tables
- Market sizing (TAM/SAM/SOM)
- Audit-traceable assumption ledger
Stack · Model Selector + Math Verifier + Wolfram
Quality Reviewer
The Red Team. Scores every deliverable against an MBB rubric, challenges the storyline, audits citations, signs off — or rejects.
- 12-criterion quality scorecard
- Citation & framework audit
- Counter-argument generator
Stack · Model Selector + Responsible-AI Eval Suite
Every deliverable scored.
Before anything ships, the Quality Reviewer runs a 12-criterion scorecard against the brief, the framework set, and the data. Below 80, work loops back. Above 80, a human reviewer signs off in writing. No deliverable leaves without both.
Problem framingHypothesis · scope · MECE
96
Framework fitRight tool · right depth
90
QuantificationModels · sensitivities · units
88
Source qualityCitations · recency · weight
84
StorylinePyramid · so-what · stickiness
92
Risk & counter-viewsRed Team · failure modes
86
Executive clarity40-second readout
91
Analytical rigorLogic chains hold
89
Financial credibilityAssumptions defensible
87
Implementation readinessPath to action
83
Citation auditEvery claim backed
88
DifferentiationUnique to this client
94
Composite89✓ Passes review
23+ frameworks, indexed and citable.
Strategy
Positioning & competitive edge
Porter's 5F · Blue Ocean · Wardley · 7S · Ansoff
Financial
Numbers, modelled
DCF · Unit economics · TAM/SAM/SOM · LTV/CAC · Profit pools
Operations
Cost & throughput
Lean · TOC · Activity-based costing · OPS diagnostic
Change
Make it stick
Kotter 8-Step · ADKAR · Bridges · McKinsey 7S
Innovation
Where to play next
Jobs-to-be-Done · Three Horizons · Stage-Gate
Customer
Insight & loyalty
CX journey · NPS drivers · Segmentation matrix
Risk
What could break
Monte Carlo · Scenario planning · Pre-mortem
Synthesis
How we communicate
Pyramid Principle · SCQA · MECE · Storylining
AI does the work. Humans hold the keys.
The pipeline is automated; the governance is not. Every engagement has a named human reviewer who can pause, reroute, or kill any stage. You see the audit trail in real time — every framework chosen, every source cited, every assumption made.
Live audit trail
Every prompt, framework call, source, and revision is timestamped and visible to you and your reviewer.
Reviewer approval gate
Phase 05 (Review & Ship) cannot complete without a named human reviewer signing off. We capture their notes alongside the deliverable.
Revision loops
If something feels off, push back. Each tier includes a defined number of revision loops; the agents adjust, the reviewer re-scores.
Your data, your scope
Documents you upload are scoped to your engagement. Nothing trains a model. Encryption in transit and at rest.
Why this beats a smarter chatbot.
01
Structured method, not a prompt
The frameworks are encoded as tools the agents must call. There's no “creative” path that skips the issue tree or the scorecard.
02
Specialised agents, not one model
Different brains for different jobs. Strategy synthesis, financial modeling, and quality review run on different model + tool stacks tuned for each role.
03
Human governance, not AI confidence
Every stage gates on a quality score and a human signoff. Confidence is earned by audit, not asserted by the model.
Method-led. Outcomes-priced.
Pick a tier · Minutes per stage · Hours to full reviewed deliverable