Guest Lecture · E-Business Concept · BINUS

From
Analytics
To AI

How digital business turns data into decisions

Luthfi ArifinProduct Engineer @ Tech in Asia · Founder @ Abadikan.id

15 Jul 2026

Luthfi Arifin

Product Engineer · Founder · building since 2019

Your speaker

Luthfi Arifin

Product Engineer @ Tech in Asia · Founder @ Abadikan.id. I don't just teach this shift — I run it every day.

2019→ now

Building real products for 6+ years

Everything today comes from my own stack — not a textbook

My day job

Product Engineer @ Tech in Asia

I ship product features on top of analytics — data is my daily input, not an afterthought.

My own business

Founder @ Abadikan.id

The live business you'll meet later today — real revenue, real dashboards, bootstrapped by me.

How I work now

AI is my analyst

I ask my own data in plain language — the exact shift this whole talk is about.

Three numbers from 2026

The analyst's job changed this year

91%

Businesses now use AI

Up from 55% in 2023 — near-universal in three years.

Master of Code, 2026 ↗

From <5% to

40%

Enterprise apps with an AI agent

Within 2026 alone — roughly an 8× jump in one year.

Gartner, 2026 ↗

43%

Workers fear AI takes their job

Within the next two years — up 5 points from 2025.

Founder Reports, 2026 ↗

Source

Gartner press release: 40% of enterprise apps will feature task-specific AI agents by 2026, up from less than 5% in 2025

Gartner, Aug 2025 — the headline behind the 40% stat

Something shifted fast. Before I show you my own data — one question.

One question first

How does AI
show up in
your day?

You ask ChatGPT to fix your code. Claude to summarize a reading, draft the email you keep dodging. That's AI already doing your work — today we flip to the side that builds it.

You are here

2:41 ☰☰!

Claude

Which Abadikan product is losing momentum — and what do I do about it?

Claude

Queried Metabase · sales_monthly

Suro is down 34% over 3 months — your biggest risk. Cut ad spend here and shift it to Nastar (+18%)

Reply to Claude…

AI in my day — my own Claude app, this afternoon

Today, in four moves

01

The ladder you already know

A quick recap of the analytics your lecturer taught.

02

Real case: my data

Live numbers from running Abadikan.id.

03

The shift to AI analyst

From SQL-by-hand to asking in plain language.

04

What it means for you

The skills that stay valuable.

Heads up: there's a screenshot demo in the middle — hang on for it.

1

Part One

The ladder
you already
know

Why data, not vibes

Business is dynamic. Every move is steadier when it's backed by data. You learned this already — I just want to prove it from my own business, not a textbook.

Understand customers

Who buys, what they want, when they come back.

Sharpen campaigns

Spend where it returns; stop where it leaks.

Driving forward at dusk, city lights ahead

Move faster

Adapt strategy to the market before it's too late.

Five levels of analytics

05 Cognitive Can a machine analyze & decide the way a human would? Remember this one ←

04PrescriptiveWhat should we do about it?

03PredictiveWhat is going to happen next?

02DiagnosticWhy did it happen?

01DescriptiveWhat happened?

2

Part Two · The heart of today

The data
I actually
have

Meet Abadikan.id

A digital invitation platform, expanding into gifts & anniversaries. Bootstrapped by me.

Sources

Transactions · ad spend · web sessions

→

Data
Warehouse

One place, cleaned & joined

→

Metabase

Dashboards I read every day

A small business, a real data stack — not corporate-only anymore

Abadikan.id · live Abadikan.id storefront — an interactive digital wedding invitation and the new Interactive Gift product page

My live storefront — invitations + the new gifting line

Level 01 · DescriptiveWhat happened?

Total Revenue

Rp ▓▓▓

Ad Spend

Rp ▓▓▓

Blended MER

3.25×

Live · Metabase Abadikan Blended Performance dashboard — revenue vs ad spend, monthly

My Blended Performance dashboard · rupiah figures masked

The dashboard answers this in a second. This used to be a manual spreadsheet rebuild, every week.

Level 02 · DiagnosticWhy did it happen?

MER by month · invitation line

Mar 26.5× Spike — barely any ad spend that month

Apr 3.83× Spend climbs back on

May 3.67× Revenue peaks — but spend rose faster

Jun 1.91× ⚠ The cliff — revenue fell off

Jul* 2.13× Partial month (14 of 31 days)

A chart shows WHAT.
Not WHY.

Seasonality? A campaign gone wrong? A tracking bug? Finding out used to mean hours of digging — row by row, by hand.

Hold that question →

Revenue didn't slowly fade — it peaked in May, then fell off a cliff in June. The chart makes that obvious. The why is the hard part.

Blended MER

3.25×

Every Rp 1 of ads returns Rp 3.25 in revenue.

One number,
one decision

Rule > ~3× & healthy margin → scale up · < 1× → cut

Healthy ecommerce MER sits at 3 to 5×. Abadikan sells a digital product — near-zero cost of goods — so 3.25× is genuinely strong here.

Same number, different decision — because of business context. Hold that thought. So — who runs all this analysis?

3

Part Three

From human
analyst to
AI analyst

The Bottleneck Moved

Same data, new flow — before: analyst bottleneck with manual query, dashboards and slow insights; now: AI connects via MCP, queries, analyzes and explains in plain language for a faster decision

What disappeared is the bottleneck, not the person. Gartner: 75%+ of enterprises moving to augmented analytics by 2025 — an industry shift, not just me.

DemoAI answered the "why" — in one chat

Claude + Metabase MCP Claude diagnosing Abadikan's revenue decline via the Metabase MCP server

Read-only · scoped to my permissions — I wrote zero SQL

What it found — from my data, one conversation

🐛 A bug I couldn't see

A case-sensitive filter (status='PAID') was hiding gifting — 24 orders in 14 days, my fastest-growing line, invisible on the dashboard.

📉 The real cause

Not the ads — Suro / Muharram wedding seasonality, ≈60–70% of the drop. "No amount of spend fixes it."

✂️ The decision

Cut the cold campaign (0 purchases all July) · scale gifting + the warm pool — with the exact campaign names.

Descriptive → diagnostic → prescriptive, in one conversation. Official Metabase MCP, scoped to my permissions. Not a hack.

Claude + Metabase MCP · unedited Full Claude conversation diagnosing Abadikan's revenue decline through the Metabase MCP server

Watch it work

One prompt.
Five steps.

1I ask in plain language — no SQL

2It loads Metabase MCP tools · finds dashboard id 27

3Queries my warehouse · pulls the monthly series

4Ranks 3 hypotheses by evidence (Suro #1)

5Ends with a decision — cut this, scale that

One AI. All five levels.

The pyramid didn't change — what changed is who climbs it.

05Cognitive Can a machine analyze & decide like a human? That demo. This level. ←

04PrescriptiveWhat should we do?1 AI · 1 chat

03PredictiveWhat will happen?1 AI · 1 chat

02DiagnosticWhy did it happen?1 AI · 1 chat

01DescriptiveWhat happened?1 AI · 1 chat

byShape — a digital product studio born from an Abadikan.id project: one client's request (Hotel Alana's interactive catalog) became a business

4

Part Four

What this
means
for you

Three things to keep

01

AI is only as good as your data

Messy data = garbage insight. Data management stays the foundation — that's your Data & Information Management class.

02

AI gives insight, you give judgment

It says "MER 3.25×". Whether to scale or cut needs margin, cashflow, context — which only you have.

03

Speed is the real shift

Data to decision, far faster. That's the whole title of today.

What stays valuable

Not "master SQL." These:

Team planning at a whiteboard of sticky notes

Understand the business

What question is the data answering?

Ask the right question

A good question beats a good answer.

Chess board with one dark pawn among light pieces

Judgment + context

The part AI still doesn't have.

Data literacy

Sanity-check the AI — it can be confidently wrong.

AI won't replace you — but a person using AI will replace someone who doesn't.

“

Digital business wins not by having the most data — but by turning data into decisions the fastest.

Luthfi Arifin · laam.my.id

Let's argue about it

How much of your
data work is still
manual - and what
could AI do instead?

Think about the last data task you did — cleaning a sheet, building a report, pulling numbers. Which part is still by hand, and which part would you hand to AI tomorrow?

Slides + connect

QR code — open the slides and connect with Luthfi Arifin

laam.my.id/talks/
from-analytics-to-ai

From Analytics To AI

Luthfi Arifin

Product Engineer @ Tech in Asia

Founder @ Abadikan.id

AI is my analyst

The analyst's job changed this year

Businesses now use AI

Enterprise apps with an AI agent

Workers fear AI takes their job

How does AIshow up inyour day?

Today, in four moves

The ladder you already know

Real case: my data

The shift to AI analyst

What it means for you

Why data, not vibes

Understand customers

Sharpen campaigns

Move faster

Five levels of analytics

Meet Abadikan.id

A chart shows WHAT.Not WHY.

One number,one decision

The Bottleneck Moved

🐛 A bug I couldn't see

📉 The real cause

✂️ The decision

One prompt.Five steps.

One AI. All five levels.

Three things to keep

AI is only as good as your data

AI gives insight, you give judgment

Speed is the real shift

What stays valuable

Understand the business

Ask the right question

Judgment + context

Data literacy

How much of yourdata work is stillmanual - and whatcould AI do instead?

From
Analytics
To AI

How does AI
show up in
your day?

A chart shows WHAT.
Not WHY.

One number,
one decision

One prompt.
Five steps.

How much of your
data work is still
manual - and what
could AI do instead?