Video summary

DSAI HDA AVS Technical Skills 2: Generative AI and Its Application in Industry

Main summary

Key takeaways

Business

Business-Focused Summary (Generative AI + Payment-Industry Application)

1) Program / Organizational Context (BINUS Student Engagement + Hiring Pipeline)

  • BINUS runs TRK “independent self-study” series events that equip students with:
    • Technical skills
    • Soft skills
    • Emphasis on company exposure/voice and graduation preparation
  • E4TP (I4TP / IORTP — inconsistent spelling in subtitles) is presented as a potential:
    • Industry partner
    • Recruiting channel
  • Student engagement timeline
    • Students in Semester 6 (and some Semester 7) are included as part of graduation preparation / waiting periods.
  • Continued engagement plan
    • Via LinkedIn/Instagram
    • Via internships
  • Career outcomes mentioned
    • Internship → potential full-time conversion (example shared by BINUS alumnus David)
  • Open roles referenced
    • Backend developers
    • QA (test automation context implied)

2) Company Strategy & Market Positioning (E4TP as Payment Gateway + Ecosystem)

E4TP operates in the payment gateway sector as a “broker”:

  • Aggregates payment channels from bank partners
  • Sells/serves those capabilities to B2B merchants

High-Level Timeline / Evolution

  • 2015: Started as “MC payment”, with an early focus on tourism/travel agencies
  • 2023: After ~8 years, acquired / integrated under IORTP
  • Broader group positioning (as mentioned):
    • Upstream: “N/needle” (unclear in subtitles)
    • Downstream examples include telecommunications, energy/solar, FTTH, and payments (E4TP)

Scale Metrics (Stated)

  • ~6,000 merchants
  • ~70 million transactions processed
  • ~20 payment methods
  • 4TP mentioned (likely scope/company count/operations—unclear), alongside emphasis on continued expansion and international capability

Product / Operations Scope

  • Supports both online and offline payments (end-to-end solution)
  • Ongoing merger/integration with “Nova” devices to unify online + offline processing
  • Emphasis on faster service by becoming an acquirer, reducing dependency on the TP-side for some SLAs (“faster”)

3) Go-to-Market / Merchant Enablement (How Merchants Integrate)

Merchant integration paths

  • API-based integration
    • Merchant/app consumes E4TP capabilities via API
  • Non-platform integration (no merchant platform required)
    • E4TP provides Instapay
      • Merchant creates an invoice
      • Shares a payment link through WhatsApp / Telegram / email
      • Customer selects from multiple payment options and completes payment

Sales / activation examples

  • E4TP pushes usage for:
    • Retail
    • E-bazaars
    • Events
  • Example mentioned: Excel Paris (spelled “Excel” in subtitles)
    • Integration in progress
    • Merchant can purchase data packages and make payments
    • E4TP processes “behind the scenes”

4) Payment Product Capabilities (Business Use Cases)

“Pay in / Pay out”

  • Pay in: receive funds from customers
  • Pay out: distribute/transmit funds to many destinations

Concrete scenarios

  • Payroll
    • One source account → distribute to multiple destinations
  • Commissions / agent payouts
    • Companies pay agents through the platform

Compliance positioning

  • Licensed by Bank Indonesia
  • Registered with relevant associations (e.g., “KOMDigi” mentioned; partially obscured)

Generative AI: Frameworks & Industrial Application (Fraud Detection Example)

A) AI Evolution Playbook (IBM “Phases” Narrative)

The presenter explains AI maturity using a timeline and differences between AI types:

  • ~1940s: conceptual neural networks (brain-neuron inspiration)
  • AI term/phase appears later (a conference is referenced where the term emerged)
  • AI Winter
    • Funding reduced due to limitations (insufficient compute/resources)
  • 1990s onward
    • Renewed progress (faster computers, more software; example: chess program vs world champion)
  • Progression: Machine Learning → Deep Learning → Generative AI
    • Machine Learning: categorization/recommendations (e-commerce)
    • Deep Learning: image recognition / face detection
    • Generative AI: predicts/generates responses (chat-style outputs)

B) Traditional AI vs Generative AI in Fraud Detection (Key Operational Contrast)

Use case: A payment gateway must determine whether a transaction is fraud or not fraud using a Fraud Detection System (FDS).

  1. Traditional AI (rule-based / ETL / fixed rules)

    • Requires preparation work similar to ETL:
      • Extract → Transform → Load
    • Decisioning depends on pre-defined rules and/or fixed patterns
    • Example anomaly logic:
      • “You always shop once a month,” but anomalies like “shop daily from different locations” → flagged

  2. Generative AI (adaptive learning / less manual rule engineering)

    • Contrast presented:
      • Less need for ETL-style manual rules
      • Better handling of changing context
    • Example:
      • Movement between locations may not imply fraud
      • The model can learn context like “you moved from B to A,” rather than treating “location changed” as automatically suspicious

Claimed FDS advantages

  • Adaptive learning
  • Potentially less frequent manual rule updates as fraud patterns shift
  • Better responsiveness to evolving fraud behaviors

C) Advantages & Disadvantages / Governance Needs

Advantages mentioned

  • Adaptability vs fixed rule systems
  • Continued learning from existing data
  • Flexibility when fraud patterns evolve

Disadvantages / risks

  • Hallucinations / wrong predictions
    • AI is not guaranteed 100% correct
  • Requires human oversight for verification/precision
  • High compute cost
    • Larger models require more resources (RAM/compute referenced)

Generative AI “Agentic” Use Case (Action-Taking Systems)

The presenter introduces agentic AI:

  • Traditional chat: user asks → AI answers
  • Agentic AI: user sets a goal → AI plans → takes actions → evaluates results → repeats/improves

Example flow (e-commerce)

  • “Find the cheapest Nike shoes below $X and buy them”
  • AI searches, selects, and orders
  • Human verifies (e.g., confirms correct price/item)

Key risk noted

  • Verification remains difficult:
    • How to ensure the agent is real/not malicious
    • How to ensure it follows human instructions

Conclusion emphasized

AI/agentic systems are tools, not full replacements for humans.

Actionable Recommendations (From Q&A + Employer Perspective)

For Developers / Students

  • Don’t rely on AI as a shortcut
    • Learn fundamentals first; use AI as a “study partner” to explain concepts
  • Choose a track and master fundamentals
    • Backend vs frontend require different toolsets
    • For testing/QA, toolsets depend on whether testing is for web UI or APIs

For Internal Career Planning

  • E4TP is open to roles such as:
    • Backend developer
    • QA (API/testing focus referenced)
  • Current engineering work described as building:
    • APIs
    • QA-related responsibilities

Concrete “AI in engineering” application ideas discussed

  • Improving efficiency and implementation workflow
    • Use AI to speed up coding processes and improve documentation
  • Debugging via agents and error summarization
    • Feed error logs repeatedly to an agent
    • Agent summarizes:
      • error patterns
      • when/where errors spike (time-based, component-based)
      • higher-level conclusions to speed remediation

Key Metrics / KPIs Mentioned (Business Execution)

  • E4TP operational metrics:
    • ~6,000 merchants
    • ~70 million transactions
    • ~20 payment methods
  • Adoption / enablement
    • Mentions payment channels including:
      • card payments
      • virtual accounts
      • direct debit (explicit types mentioned)
  • Targets / timelines
    • No explicit revenue/CAC/LTV/churn targets stated
    • Timeline mainly used for company history:
      • 2015 start
      • 2023 acquisition/positioning
      • ~10-year ecosystem framing

Presenters / Sources (As Stated)

  • Dinda — Talent Acquisition, E4TP/EORTP (spelling varies in subtitles)
  • Mas Rizki Isnand Putra — Engineering Manager; BINUS alumnus (BINUS CS, class of 2010); presenter on Generative AI
  • Mr. Irfan Yuliandi — Marketing Manager; presenter on E4TP overview
  • David Ekoputra / David Putra — BINUS alumnus (class of 2019); shared internship → full-time experience story
  • Mrs. Jurike — moderator / event lead (host role referenced)
  • Teacher Mrs. ST / Mrs. Ajang — mentioned during introductions (role unclear in subtitles)
  • External referenced source:
    • IBM (basis for the AI “six phases” framing)

Original video