01
Fragmented marketing knowledge
Product context, brand tone, terminology, examples, SOPs, and reference materials were spread across different documents, folders, and conversations.
Case 05 / AI Marketing Team
AI Marketing Team
Turning recurring marketing work into agent-assisted workflows.
- Project
- Independent Practice / POCSTARS
- Period
- 2025–2026
- Role
- Workflow Designer / Global Marketing Manager
- Focus
- AI workflows, agent roles, skills, SOPs, references, templates, human review
01 / Context
AI needs structure before it can support marketing work.
In lean B2B marketing, recurring tasks from blogs to campaign briefs need to be produced quickly but still require judgement. The challenge was not faster drafting but building a workflow structure that preserves context, terminology, and review quality.
AI WorkflowAgent RolesSkillsSOPsTemplatesReferencesB2B ContentHuman Review
02 / Challenge
Why one-off prompting was not enough.
02
Repetitive tasks with judgement
Blogs, EDMs, social content, and campaign materials may look repetitive, but each task still needs judgement around accuracy, claims, tone, channel fit, and market context.
03
AI output needed boundaries
Without roles, skills, references, templates, and review logic, AI output can look complete but still be inaccurate, generic, or unsuitable for B2B use.
03 / System Built
A local AI Marketing Team built around roles, skills, and knowledge layers.
The system separates marketing work into layers: tasks are mapped to agent roles, agents call relevant skills, skills are supported by SOPs, templates, references, and context files, and final output remains subject to human review.
01
Task
- Blog
- Social
- EDM
- Campaign
- Research
- Sales Enablement
›
↓
102
Agent
- Market Researcher
- Content Creator
- Data Analyst
- Creative Designer
- Campaign Strategist
›
↓
203
Skill
- Blog Writer
- Social Copy
- Campaign Brief
- Data Visualisation
- Branded Deck
›
↓
304
Knowledge Inputs
- Context Files
- SOPs
- Templates
- References
- Best Examples
›
↓
405
Output + Review
- Draft Output
- Human Review
- Final Asset
04 / Output
Selected workflow artefacts.
A local folder-based workflow separating tasks, agents, skills, SOPs, templates, references, and outputs so recurring marketing work is easier to brief, reproduce, and review.
Blueprint
01
AI Marketing Team blueprint
A design logic mapped recurring marketing tasks into agent roles, skills, and a structured workflow.
Agent Roles
02
Agent role system
Five agent roles map to practical functions: research, content creation, data analysis, creative support, and campaign strategy.
Folder System
03
Local knowledge structure
Folders separate context, SOPs, templates, references, agents, skills, scripts, and outputs, keeping the workflow modular and easier to maintain.
Blog Workflow
04
Blog writing workflow
Blog writing became the most mature applied workflow, with a clearer SOP, skill definition, references, and repeatable structure for B2B technical content.
Skills
05
Expandable marketing skills
Built to expand beyond blog writing into social copy, EDM, case studies, campaign briefs, sales enablement, and reporting.
05 / Impact
Making AI usable inside real marketing work.
The system turned AI from a one-off drafting tool into a more controlled workflow layer, making recurring marketing work easier to brief, reproduce, and adapt, while final judgement remained human-led.