Ida Wahyuni, M.Kom., Ph.D - Sesi Praktek: AI untuk Profesional & Akademisi - Umum 2026-06-09 S2

Main Ideas, Concepts, and Lessons

• AI adoption is not “new,” but capability has changed rapidly

  • The speaker explains that earlier generations often learned and completed assignments without AI.
  • While AI existed during their doctoral period, it was limited (e.g., grammar correction) and did not yet perform deeper tasks like:
    • data analysis
    • complex coding
  • Building a foundation in basics/theory helps professionals interpret AI outputs and verify them.

• AI outputs must be understood and validated (not blindly trusted)

  • Even with claims like “~80% accuracy,” there’s still a ~20% risk of:
    • errors
    • hallucinations
    • fabricated details
  • Analogy: autonomous vehicle models trained in one environment may fail in another because real-world conditions differ (e.g., traffic light visibility).

• To get useful AI results, prompts need structure

  • The session introduces a prompt engineering framework referred to as R O C E X / “ROCEX.”
  • The key message: context must be engineered—AI should not guess your internal assumptions.

---

Methodology / Instruction List (Prompt Frameworks and Practical Workflow)

A) “ROCEX” Prompt Structure (for Better AI Responses)

Used to tell the AI who it is, what to do, and what constraints apply, so outputs better match your needs.

• R = Rule

  • Specify the AI’s role/expertise.
  • Example: “Act as an expert SEO agent with 10 years of experience.”
  • Tailor to the user’s needs.

• O = Objective

  • State the specific goal.
  • Example: “Create an SEO strategy for the website’s intended purpose.”

• C = Context (described as the most important part)

  • Provide background and “engineering context” so the AI doesn’t guess your internal assumptions.
  • Include target details such as:
    • target audience age group
    • monthly budget
    • business/assignment background
  • Include constraints/limits (the second part of “C”), for example:
    • prevent the AI from using materials “from anywhere”
    • focus only on “white-hat SEO” and Google Indonesia only

• E = Example

  • Provide an example of the desired output format.
  • Example: SEO strategy shown as a table with columns like keyword, volume, difficulty, etc.

• X = Extension

  • Additional specifications; interpretation can vary by platform/tool.
  • The speaker notes that examples may be shown later depending on the workflow.

---

B) Templates for Reuse Across Chats

• If you’re using tools like ChatGPT (or CJPT) or other AIs:
  • Use templates early to get correct structure on the first try.
  • Then revise with additional commands without re-sending the whole template every time.
• Two template styles discussed:
  • Longer structured template (ROCEX-like)
  • Shorter “case” template
    • Includes the core context/background + objective/actions
    • Emphasizes keeping context consistent across prompts

---

C) Practice Workflow: Create HTML Outputs Directly From Prompts

The instructor demonstrates a hands-on “homework/project” approach:

1. Use a prompt to request output in HTML format (e.g., a one-page executive summary).
2. Generate an HTML file and open it in a browser (e.g., Chrome).
3. Customize placeholders:
   • replace case/business details
   • adjust branding/theme instructions (e.g., dark mode, gradients)
4. Review/edit if needed, then present.

---

D) AI-Made Presentations as Alternatives to PPT

Instead of relying only on slides/PPT, the instructor recommends:

• Using landing pages (HTML) as presentation formats
• Adding interactivity, such as:
  • hover effects
  • click-to-show sections
  • animated elements
  • menus

Goal: reduce dependence on web developers for simple but presentable pages.

---

Tooling Guidance (What Was Emphasized)

• Use whichever AI app you already use most (e.g., ChatGPT, Gemini, “Cloud”, “Codex”, etc.).
• Many examples can be reused via the same templates; different tools may vary in output quality/detail.
• A “cloud” approach was mentioned as a way to:
  • connect data sources (spreadsheets/files)
  • run jobs like an agent/workflow
  • schedule tasks

---

Data Privacy and Safety Points (Explicit Concerns Raised)

• Confidential business data may be risky to upload to some AI tools.
• The speaker advises:
  • review terms & conditions for uploaded files
  • check whether uploaded files are used for training
  • be especially cautious with free AI versions (less control)
• Example setting: in Gemini, there may be an option/button to disable file usage for training.
• Key caution: even if files aren’t publicly shared, the AI could still “learn” from them depending on configuration—creating risk if competitors later ask similar questions.

---

Q&A Highlights (Summarized)

• Can AI retrieve marketplace/sales data using links?

  • Depends on the AI—some can collect data if you provide marketplace links and/or use connectors.

• If there is a lot of Excel data, must you summarize first?

  • Some AIs allow only around ~10 file uploads for analysis.
  • If more than that, summarize in batches, then ask again.

• Do you need to summarize first, then create the HTML page?

  • Sometimes you can do it directly in one step, but multi-step workflows can also work.

• Why did one tool generate irrelevant industry content when fields were left blank?

  • Some tools infer/guess missing context, causing domain mismatch (a hallucination/context-mismatch risk).

---

Speakers / Sources Featured (As Identified in the Subtitles)

1. Ida Wahyuni, M.Kom., Ph.D (main instructor; referenced repeatedly as “Mrs. Ida”)
2. Mr. Rendi
3. Mr. Tanadi
4. Mr. Ari (mentioned regarding an AI agent/agent creation material in a later module)
5. Ms. Yuli (questioner; “Ms. Yuli Oke” mentioned)
6. Mr. Michael Abimayu
7. Mr. Denis Natawijaya
8. Mrs. Sari
9. Mrs. Yun Kusumawati
10. Mrs. Michel (referenced as reserved/asking later)
11. Mrs. Tias Sarj / Tias Sarjadd (presented; “Mrs. Tias”)
12. Mr. Rudi
13. Mr. Budi Mazuki
14. Mr. Hendi (present in multiple turns; also thanked/asked by others)
15. Mr. Nata (presented; refers to analyzing a financial report)
16. Mr. Budi (asked in Q&A about internal data/privacy)
17. Mr. Yuli(s) (mentioned in the discussion about modules/rewards; also “Yulis leaked it early”)