top of page

Joon Han | Digital Diagnostics

AI as Strategic Leverage

This section documents how I integrate AI into personal branding and marketing systems.
 

My approach is not about chasing tools.

It is about using AI to accelerate structured thinking, compress iteration cycles, and enhance signal clarity.
 

Coming from a clinical background, I treat AI as an operational layer  not a replacement for judgment.

​

The Experiment

​

  • Zero Code Mastery
    Transitioning from a medical background with no prior coding knowledge to designing complex AI workflows.
     

  • Cross Platform Systems Thinking
    Connecting models, analytics tools, and structured frameworks to move from raw data to usable insight
     

  • Strategic Application
    Moving beyond beginner experimentation to using advanced systems intentionally for branding, analytics, and growth.

​

Below, I document practical experiments and what each revealed about applying AI in real marketing and measurement environments.
I use it to pressure-test and refine it

The Past: The Era of "Narrow" Intelligence

(2010s – 2024)

​

  • The Players: Dominated by early IBM Watson, Google DeepMind, and the rise of LLMs (OpenAI, Anthropic).

​

  • The Workflow: AI was a passive assistant. It required high "gravity"—manual prompts and human intervention for every task.

​

  • The Focus: Solving narrow problems: text generation, basic coding, and image creation.

​

The Present: The Agentic Shift

(2025 – 2026)

​

  • The Ecosystem: Transition from static chatbots to Autonomous Agentic Networks.

​

  • The Technology: Leveraging Antigravity to remove manual friction and A2A (Agent-to-Agent) protocols for cross-platform execution.

​

  • The Impact: Massive data synthesis via NotebookLM turns fragmented information into strategic knowledge instantly.

​

The Future: The Era of Predictive Autonomy (2027+)

​

  • The Vision: AI moves from a tool to an invisible strategic partner.

​

  • Living Brands: Personal identities that adapt

​

  • Autonomously to global sentiment in real-time.

​

  • Proactive Strategy: Shift from analyzing what happened to predicting market gaps before they exist.

The Limits of Portfolio Attention

​

A portfolio compresses complex thinking into limited space. Most visitors don't have time to fully engage, which means nuance and depth are often missed.

​

I designed a single landing card as a clear entry point for time-constrained readers.

However, I discovered key limitations:

​

Templates accelerated production but introduced constraints. Design decisions became fixed.

Flexibility narrowed.

 

AI tools aren't beginner friendly. Different platforms have different rules. What works in one tool breaks in another.

 

Tool integration requires foundational knowledge. Syncing Stitch → Figma → publishing requires understanding landing page structure and web mechanics knowledge AI alone cannot provide.

 

This taught me: focus on building foundational skills first understanding landing page structure, platform mechanics, and design systems before automating around them.

 

AI accelerated execution, but structure and expertise remained human-led.

Prompting as First Principle:
A Documented Experiment in AI Intentionality

​

Better AI output often starts before the first task is even typed.
 

This experiment documents how I shifted from casual AI use to intentional AI use by focusing on instruction clarity before automation.

Rather than jumping to agents and automation tools, I discovered that structured prompting setting context, defining roles, breaking work into stages, and maintaining human control at each step delivers more reliable results.
 

The core insight: before AI becomes an automation problem, it is usually an instruction problem.

This applies directly to analytics. Better insights require better measurement frameworks, not just better dashboards.

Better AI output requires better prompts, not just better models.
 

Read the full reflection: "Why Better AI Use Starts With Better Prompting"

From Job Search to Semi Automation:
Exploring Claude's Cowork and Projects for Practical Workflows
 

Claude's Cowork and Projects capabilities are gaining attention because they offer people a practical entry point to automation without jumping straight to full agents.
 

In my case, I'm designing a workflow to optimize my job search process using Claude's Skills: automated job searching, resume editing tailored to each job description, and application material preparation all while I sleep. When I wake up, I review and submit.
 

This approach solves a real constraint—job search is repetitive and time intensive while maintaining the human review step that ensures quality and fit.

The core principle mirrors my earlier insight: semi-automation through structured workflows beats jumping to full agents.
 

This experiment is ongoing. I'm documenting what works, what breaks, and how Claude's Skills enable this kind of practical automation that keeps humans in control while compressing iteration time.

​

​

When AI Gets Better, Iteration Gets Cheaper
 

Recent updates across ChatGPT, Claude, Gemini, and other AI platforms point to a larger shift. As models become stronger and agent workflows improve, the main beneficiary is the user.

​

For analytics, this matters because the cost of testing ideas is getting lower. AI can help analysts frame problems, explore possible causes, draft SQL or Python logic, and compare interpretations faster.

 

But speed alone is not the value. The real value is knowing what to ask, what to verify, and what not to trust too quickly.

 

As AI tools improve, the analyst’s role does not disappear. It shifts toward stronger judgment, clearer prompting, and better validation. AI can make analytical work faster, but the analyst still has to decide what is worth analysing.

​

​

bottom of page