The Intelligence System Case Study
Turning scattered client knowledge into compounding strategic intelligence.
Work done as Director of Strategy & Growth (contract) at Monumental, a Shopify-focused growth agencyThe Challenge
While working at Monumental, I identified a recurring failure point in client services: strategic context.
Everything learned in meetings, audits, and data reviews lived in scattered notes and individual memory, and evaporated when accounts changed hands or time passed between touchpoints.
Business reviews were being rebuilt from scratch each quarter, taking 9+ hours of deck work, and the insight that made a client feel understood in Q2 was gone by Q3.
Objectives
Convert every client interaction into durable, compounding strategic intelligence — so each new conversation builds on everything that came before.
Cut business-review and brief prep time without cutting depth.
Build a verification layer that catches AI overconfidence before it ever reaches a client.
Snapshot
Industry: Ecommerce growth agency — Shopify Plus client portfolio
Role: Director of Strategy & Growth (contract) — system design, prompt engineering, quality assurance, portfolio rollout
Tech: Claude, Shopify Plus data, meeting transcripts, multi-year sales exports
Duration: Q4 2025 - Q2 2026 then in active use across the portfolio
Results: $14M+ misattributed channel revenue surfaced · reviews -75% (~100–150 hrs/yr saved portfolio-wide) · briefs 2–3h → ~30 min · 8 client accounts
Services: System design, iterative prompt engineering, domain-expert QA, live validation, templatized for scale
Results
Business-review prep time dropped ~75% (an estimated 100–150 hours saved annually across the portfolio)
Strategic briefs too ~83% less time to produce
The system was in active use across 8 client accounts, generated from 44 working conversations and roughly 200 supporting documents.
Deliverables
One prototype with a repeatable framework.
It started as a single hand-built prototype for one client in November 2025: one long conversation, three meeting transcripts, and a business-review deck, producing a master strategic document plus four supporting ones. Within a day, this was generalized it into a repeatable framework.
A fixed 11-section schema.
The second deployment forced unification of two diverging prototypes into one fixed 11-section schema, the point where it became a system rather than a one-off document.
A business-model reframe.
One engagement proved the system could turn 22 source documents and a 3-year sales export into a client-ready strategic document in a single session and surfaced a reframe: the client's site was a research channel feeding 85% in-store revenue, not an underperforming sales engine.
The verification discipline.
The system's most defensible piece. The system caught the AI's first-pass analysis fabricating a "$25 first order / $12 profit" crisis narrative built on an unrepresentative 9–20% cohort sample. The real number was closer to $52. That correction became a standing rule: benchmark every claim, check sample representativeness, and log every correction before it reaches a client.
The $14M finding.
The same scrutiny was applied to a finding for an outdoor gear brand that turned out to expose a genuine $14M+ attribution gap in their email channel: five years of email-driven revenue that broken UTM tagging had been crediting to direct and organic traffic. The discipline is what let me tell the difference between a real finding and a plausible-sounding wrong one.
Scaling in both directions.
The system extended to a brand-new relationship, generating institutional memory before the engagement even started and has been adapted for a lean, low-budget clients, proving the model scaled down as well as up.
The Growth Playbook.
By Q2 2026, the system matured into nine scored, productized 90-day growth initiatives, ranked by a weighted rubric (revenue impact, effort/ROI, client readiness) and delivered as one-page, yes/no-able recommendations turning the intelligence directly into a revenue-generating deliverable.
Why it matters
Everyone has the same AI tools now. The advantage is the system around the tool: designed, not just prompted; verified by a domain expert before anything reaches a client; iterated through real, messy, high-stakes use rather than in the abstract; and matured into a productized, revenue-facing offering.
Applied AI methodology, not just AI usage.