Make context usable
Turn project background, rules, live status and previous decisions into readable working memory—not scattered chat history.
Selected case / 10
AI-enabled marketing systems / ongoing practice
Build the conditions for people and AI to think, decide and deliver better together.
An applied practice in designing AI-assisted operating systems for complex marketing work: improving speed, preserving context, strengthening judgment and turning work into reusable assets.
01Contextgoals · evidence · constraints
→02AI-assisted workretrieve · structure · test
→03Human judgmentchoose · challenge · approve
→04Reusable outputbrief · handoff · decision
Not a chat prompt.
A reviewable operating system.
01 / Why this matters
The problem is not simply storing information.
Create a working system where human expertise and AI each improve the other.
Complex marketing work carries more than files: market context, client constraints, platform language, working assumptions, incomplete evidence and decisions made along the way. When these live only inside chat sessions, the work becomes slow to restart, hard to delegate and easy to distort.
I design the connective layer around AI: the context that enters the task, the role each tool plays, the points where a person must intervene, and the evidence that remains after delivery. The goal is not maximum automation. It is faster progress with better-informed, more defensible decisions.
02 / Design principles
Efficiency, judgment and continuity must be designed together.
Four principles turn AI use into a system that can survive real work.
Route work by strength
Use AI for retrieval, synthesis, comparison and structured drafts; use the human for goals, trade-offs, evidence and approval.
Keep judgment visible
Build source trails, evidence gaps, stop conditions and review gates into the process so a polished output never hides uncertainty.
Compound what is learned
Convert completed work into briefs, handoffs, templates and reusable capabilities that make the next project faster and better.
03 / Working systems
Workflow design that makes judgment inspectable.
Three systems show how I move complex marketing work from ambiguity to a useful, accountable next action.
Validated
Market signal → decision-ready action
A marketing-intelligence loop designed to answer a real business question. AI structures the evidence and options; the human decides whether an action is justified.
01Frames the business questionDefine the decision, audience, timing, commercial goal and constraints that the work must support.Human
→02Sets the evidence planSpecify valid sources, comparison criteria, known gaps and the boundary of what the system may claim.AI + workflow
→03Organises bounded evidenceStructure market, consumer, platform and internal inputs with source trails—without treating all signals as facts.AI + tools
→04Builds the decision mapSeparate fact, inference, uncertainty and consumer friction; surface options and trade-offs.AI prepares · human challenges
Decision gateIs the evidence relevant enough to justify a marketing action?
↓Yes → Approve a route or testNo → Research more / revise scope / hold
Business-ready outputTurn the chosen route into a strategy note, content brief, test plan or stakeholder-ready recommendation.Execution and observed results become the next decision inputHuman decides · AI prepares
Validated
External material → reliable knowledge
A governed research branch that keeps a saved reference from becoming assumed knowledge, a recommendation or a database entry before it has earned that status.
01Names the research purposeA link, reference or field observation enters with a question to answer—not as an automatic collection task.Human
→02Captures the usable traceRecord accessible evidence, context and source link; do not fabricate a summary when the material is partial.AI + tools
→03Builds an evidence ledgerKeep source, visible clue, confidence, unknowns and possible relevance legible for later review.AI prepares · human reviews
Evidence gateDoes this material warrant deeper use in a live decision?
↓Yes → Extract / compare / investigateNo → Archive as reference / hold
Reusable evidence packProduce a source-linked note or comparison that can inform a brief, research question or decision meeting.No approval means no automatic conclusion, recommendation or knowledge claimHuman approves · AI formats
Active
Project complexity → continuous delivery
A continuity loop for long, multi-tool work: make the project—not a single conversation or person—the holder of intent, decisions and next actions.
01Creates a shared briefCapture objective, constraints, source locations, live decisions and the definition of a useful deliverable.Human + AI
→02Splits work by strengthSeparate research, judgment, execution and communication rather than burying every task in one chat.Human routes · AI assists
→03Maintains the live stateRecord decisions, open questions, evidence status, risks, files and the next meaningful action.AI prepares · human verifies
Continuity gateCan the work continue safely, or does it need an explicit handoff?
↓Yes → Write a decision-aware handoffNo → Continue with the live brief
Delivery that can moveResume from the same constraints and decision history, then leave a clear trace for the next transition.Continuity makes cross-functional collaboration faster without losing accountabilityNext human / model / session
These are evolving working systems, not claims of autonomous execution. The design makes evidence boundaries, human accountability and the business use of each output visible.
04 / Business translation
AI is useful when it makes strategic conversations clearer.
Translate market complexity into choices an international team can actually discuss.
In cross-market work, local platform tactics can sound like disconnected activities to overseas stakeholders. I use AI to organise verified research, field observation and platform evidence—then translate them into a shared decision language without inventing facts.
01 / Local signalVerified platform or consumer inputResearch, observable behaviour and field context with an evidence boundary.AI organises · human validates
→02 / Decision frictionName the consumer barrierWhat is stopping attention, trust, consideration or conversion in this market?Human frames · AI compares
→03 / Business logicConnect action to marketing purposeMake audience, channel role, commercial relevance and expected signal explicit.AI prepares · human challenges
Team decision gateIs the commercial logic clear enough to resource and test?
↓Yes → create an executable briefNo → refine evidence or the market question
04 / Shared actionStakeholder-ready recommendationA testable strategy, content or channel action that an international team can evaluate, execute and learn from.Human approves · AI documents
This is where AI supports judgment rather than replacing it: it makes assumptions visible, separates evidence from inference, and helps teams challenge the commercial logic before committing resources.
05 / Human control
Useful systems retain accountability.
AI can accelerate the work. It should not quietly take ownership of the decision.
Search and retrieval, source organisation, structured synthesis, comparison, draft formats, handoff preparation and repeatable document work.
Problem framing, market judgment, source credibility, business trade-offs, brand sensitivity, public-disclosure boundaries and final approval.
06 / What this proves