Case study
Client
UltraSuperNew (internal initiative)
Services
AI Transformation
Year
2025–2026 (ongoing)
Challenge
In late 2025 AI was everywhere in the conversation and almost nowhere in the workflow. Agencies ran pilots and posted manifestos; nobody had an operating model. The Anthropic Creative Agency pitch exposed the gap: “using AI” and “operating as AI-native” are different problems.
Decision
I chose to build a system, not a tool stack: 14 role-specific playbooks, a 12-lesson curriculum, a Red/Amber/Green role-impact map across 15 functions, bilingual Responsible AI guidelines, a shared Claude skill for SoW drafting, an automated request-intake flow, and a quarterly measurement layer. I built governance before a client incident forced it, and designed learning around how busy senior people actually learn — not around the tools.
Outcome
In production at Tokyo HQ, with phased rollout to Colombo, Singapore and Amsterdam. Leadership uses the role-impact map for hiring and capability investment. RFQ responses dropped from 3–5 days to about 5 hours.
Key metric
86% modeled cost reduction in voice-over localization (5 languages, AI tools vs. internal production rates)
The work
The bet: A 35-person creative agency can move faster on AI than any global holding company — if someone builds the system.
Context
This initiative traces back to Anthropic's Creative Agency pitch: a competitive process I contributed to as part of the winning USN team. The brief exposed a gap: "using AI" and "operating as AI-native" are entirely different problems. Closing that gap became the work.
In late 2025, AI was everywhere in the conversation and almost nowhere in the workflow. Agencies were running pilots, posting LinkedIn manifestos, and quietly hoping their clients wouldn't ask hard questions about productivity.
UltraSuperNew is a Tokyo-based independent creative agency working with Anthropic, Mozilla, Amazon, Salesforce, Adobe, Porsche, Spotify, Daimler, and others, with offices in Tokyo, Singapore, Amsterdam, and Colombo. The conditions were right for actual transformation: small enough to move, senior enough to know the work, independent enough to make decisions without a global capability deck.
What was missing was an operating system. I built one.
What I built
USN AI OS — an agency-wide capability platform that turns AI from a tool conversation into an organizational capability. Designed and built independently, in production across Tokyo HQ, with phased rollout to Colombo, Singapore, and Amsterdam.
It is not a tool stack. It is a system: curriculum, governance, measurement, automation, and change management — engineered to make AI fluency the default, not the exception.
The components
14 role-specific AI guides — From Junior Designer to Creative Director to Business Development, each function gets a tailored playbook showing what AI changes, what it doesn't, and what to do differently starting Monday.
12-lesson Claude Playbook — A self-paced curriculum for Account and Producer roles. Built for the way agency people actually learn: short, applied, with real USN project examples.
AI Role Impact Analysis — A strategic capability map across 15 agency functions, classifying each role by AI replacement risk (red), transformation scope (amber), or value increase (green). Used by leadership to inform hiring strategy and capability investment.
SoW drafting & request intake — SoW drafting runs as a shared Claude skill; client requests submitted via Tally become Notion projects with six dated standard tasks automatically.
Responsible AI Guidelines (EN/JA) — Bilingual governance covering client confidentiality, IP handling, and disclosure standards. Built before clients started asking — not after.
AI Pitch Playbook + Tools for Creative Production — Operationalizing AI for RFPs, competitive pitches, and production workflows, with the rate card to back it.
Measurement layer — Quantitative time-allocation survey + structured 1-on-1 interviews establishing baseline adoption, capability gaps, and rollout priorities. Tracked quarterly against defined KPIs.
The numbers (Selected Cases)
86% modeled cost reduction in voice-over localization 5-language benchmark of AI tools against internal production rates (modeled, not yet realized)
15 roles mapped in the AI Role Impact Analysis Red / Amber / Green tier classification
14 functions with role-specific AI playbooks 12 lessons in the structured curriculum 4 offices in the phased rollout roadmap
What's harder than it looks
The technology is the easy part. The hard parts are:
Getting senior people to admit what they don't know
Designing learning that respects how busy people actually are
Translating "AI productivity" into a billing model that doesn't collapse on itself
Building governance before a client incident forces it
Maintaining momentum after the launch announcement fades
USN AI OS is engineered around these problems, not around the tools.
What's next
Value-based pricing model (Phase 2)
Multi-office rollout: Colombo → Singapore → Amsterdam
External AI capability positioning for new business
LMS-backed certification via TalentLMS + Credly digital badges
Role
Initiative Lead: Strategy & leadership. Drives adoption across teams, designs personalized AI Guidelines, conducts 1on1 interviews, reports ROI to leadership, coordinates with all Directors.
