Cortex
The company's AI brain

Ask the company a question. Get the answer in seconds — not days.

A 24/7 employee that sits on top of your existing systems — always watching for problems, always ready when you ask, and handling the routine tasks on its own. Not a fixed dashboard.

8 steps
collapse into one plain-language question
24/7
watching for problems before you have to ask
0
changes to your system of record — it stays the truth
You ask

Revenue is down 15% this month. The drop is concentrated in two estates — and it traces to delayed delivery recognition, not lower production.

Source: ERP · Sales ledger Confidence: high Recommended → review delivery timing first
The problem today

One simple question can cost a week of work.

Your data already exists — it's just locked inside systems, spreadsheets, and people's experience. Getting one clear answer means a chain of manual steps, repeated every time leadership asks a follow-up.

The way it works now

Manual, slow, and hard to repeat

  • Log into the ERP and find the right module
  • Search and export the data into Excel
  • Clean it, combine it, and build the charts
  • Prepare a report for management
  • Start over when the follow-up question comes
Hours to days, every time
The way it could work

Ask in plain language, get a sourced answer

  • Type the question the way you'd say it out loud
  • The AI finds the right data across departments
  • It checks related causes and explains the result
  • You get a summary, chart, and recommended action
  • Follow-up questions are answered instantly
Answers in seconds
The core idea

Think of it as a 24/7 employee.

Not a tool you log into when you remember to. A team member who never clocks off — always watching, always ready, and handling tasks the way you like them done.

Always on duty

Watches the business around the clock and flags an abnormal case the moment it appears — overnight, weekends, holidays.

Always available

Ask any time of day and get an answer straight away — no waiting for someone to be free or to pull the numbers.

Runs your routines

Scheduled jobs generate the daily, weekly, and monthly reports and run the workflows on their own — on time, every time.

Personal to you

Adapts to what each person actually wants to see and how they want it framed — not one fixed template for everyone.

Holds a conversation

Ask follow-up questions and keep drilling deeper — instead of staring at a static dashboard and guessing.

Draws its own charts

Builds the right chart or table itself for each answer — so it works even if you have no BI tool at all.

On duty · always
06:00Daily cost monitor — checked 6 categoriesclear
06:00Anomaly scan — procurement cost1 flagged
06:05Alert sent → Procurement Managerdone
08:00Weekly CEO report — generated & emaileddone
NowWatching 12 metrics across 5 estateslive
Ask the company anything — it's already here.
No BI tool? It still works.

It's personal, not a fixed dashboard.

A fixed BI dashboard
  • One template — the same view for everyone
  • You bend your question to fit what it shows
  • No follow-ups — only filters and clicks
  • Has to be built and maintained first
Your AI employee
  • Answers your exact question, framed for you
  • Takes follow-up questions and digs deeper
  • Draws the right chart itself for each answer
  • Works even with no BI tool in place
What it does

Six things it does for you every day.

Not another dashboard to read. A working assistant that retrieves, explains, watches, and remembers — so decisions get faster and more consistent.

Answers business questions

Ask anything in plain language and get a short answer with the numbers, the source, and what to check next.

"Why did revenue drop this month?"

Writes the reports for you

Daily, weekly, and monthly reports generated automatically — with the summary, the key changes, and the charts already done.

"Generate this month's management report."

Catches problems first

It watches the business around the clock and alerts the right person when costs spike, stock runs low, or a supplier slips.

"Supplier A was late five times this month."

Explains the why

Not just that a number is low — it checks production, procurement, manpower, and logistics to find the most likely cause.

"Which estate caused most of the drop?"

Holds company knowledge

SOPs, policies, and leadership experience become searchable. New staff get the same answer the best manager would give.

"What's our SOP for a supplier delay?"

Remembers past decisions

When a problem looks like one you've handled before, it recalls the cause and the decision you made — so you don't start from zero.

"This is like last March — check delivery timing."

See it in action

What each of these actually looks like.

Tap through six real examples — an answered question, a finished report, a problem alert, and how it reasons across departments instead of just showing a chart.

AI Assistant · Chat
Show manpower cost by estate this month.

Manpower cost is RM 4.2m this month, up 6% on last month. East estate is the outlier — overtime there rose 28% while output stayed flat.

Source: ERP · Payroll + Attendance Confidence: high Next → check East overtime approvals

Plain question in. Summary, the number, a chart, the source, and a next step out.

Monthly Management Report — May 2026

Auto-generated

Revenue came in 15% below last month, concentrated in two estates. Production and stock held steady, so the gap traces to delivery recognition timing rather than operations. Procurement cost ticked up on one supplier. Two items need attention.

Revenue
RM 12.8m
▼ 15% vs Apr
Cost
RM 9.1m
▲ 4% vs Apr
Margin
28.9%
▼ 3.1 pts
Abnormal findings
  • East & South estates account for 80% of the revenue gap
  • Supplier A fertiliser price up 18% — drove the cost increase
Recommended actions
  • Confirm delivery & sales recognition timing for East and South
  • Procurement to review Supplier A pricing before next order

A full report, written and charted automatically — ready before the meeting.

Anomaly detected SEVERITY: HIGH
Procurement cost up 23% vs 3-month average
Detected today, 06:00 · daily monitoring
Likely driver
Supplier A unit price ▲ 18%
Affected
Fertiliser category
Estimated impact
≈ RM 210k / month
Compared against
Feb–Apr average
Assigned → Procurement Manager · due Friday
● Open

It finds the spike, explains why it's abnormal, sizes the impact, and routes it to the right person.

Revenue −15% this month — why?
Production
Normal
Inventory
Normal
Procurement
Normal
Manpower
Normal
Delivery timing
Delayed
Root cause
It's a delivery-recognition delay, not an operations failure. Four areas checked out normal; the gap is in when sales were recognised.

This is the difference from a dashboard: it pulls data from five systems to rule causes in and out — not just plot one number.

"What do we do when a supplier delivery is delayed?"
Procurement SOP
Section 4.2 — supplier delays
Past incident log
Supplier A · March 2026
Manager playbook
Escalation rules
Action plan
  1. Log the delay against the supplier's record and notify the buyer.
  2. Check safety stock; trigger the backup supplier if cover is under 7 days.
  3. If it's the 3rd delay this quarter, escalate to a supplier review (per playbook).
Built only from approved company knowledge — not the open internet
Now · May 2026
Revenue dropped 15%, production steady
Two estates underperforming on revenue while output and stock look normal.
92%
match
Remembered · March 2026
Same pattern — resolved in 2 days
Root cause was delivery-recognition delay, not an operations issue. Fixed by correcting recognition timing.
Applying what we learned
"This looks like the March case. Recommended first check: delivery timing and sales recognition — last time that resolved it without touching operations."

No starting the investigation from zero — it brings back the cause and the decision you already made.

Why this isn't ChatGPT

It knows your company.

A general chatbot has never seen your ERP, your estates, your suppliers, or the way your leaders make calls. This is built on all of it — a company-specific decision system, not a generic assistant.

Same easy way to ask. Completely different depth of answer.

How it works — in plain terms

Four steps from raw data to a clear answer.

You don't need to understand the plumbing. But here's the whole idea in four moves.

1

It connects to your data — safely and read-only

The AI reads from the ERP, dashboards, and documents you already have. In the first phase it can only look, never change. Your ERP stays the official record.

2

It cleans and understands the numbers

Messy codes, mismatched names, and missing values get tidied up. It learns what your business means by "revenue," "cost," or "abnormal" — so it answers the way you would.

3

It analyses, explains, and double-checks itself

It pulls related data across departments, forms a likely cause, and tests it against the numbers — always showing where the answer came from.

4

It delivers an answer you can act on

A summary, a chart, the source, and a recommended next step — delivered as a report, a chat reply, a dashboard, or an alert to the right person.

Guardrails & trust

Built so you can trust what it tells you.

The point of this system is better decisions — which means it has to be safe, governed, and honest about what it knows.

The ERP stays the source of truth

The AI helps you use the data better — it never replaces or quietly rewrites the official record.

It doesn't make things up

Answers are grounded in real data, with the source shown. If it isn't sure, it tells you the confidence level.

People decide, not the AI

It recommends and flags. Any high-impact action still needs a human to approve it.

People only see what they should

Access follows roles and permissions, sensitive data is masked, and every query is logged for audit.

How we'd roll it out

Start small. Prove value. Then scale.

We begin with one high-impact use case for management, prove it works, and expand only once it has earned trust.

Start here Phase 1 · First 90 days

Management AI dashboard

A focused pilot for leadership
  • Read-only link to finance, procurement & inventory
  • Ask-anything assistant for management
  • Automatic monthly report
  • First anomaly alerts & charts
Phase 2 · After the pilot

Across the departments

Once the pilot proves itself
  • Manpower, HR, production & maintenance
  • Supplier performance analysis
  • Task follow-up & tracking
  • Teams / Slack & email alerts
Phase 3 · The vision

A company-wide AI layer

The long-term destination
  • Sensor, drone & vision data
  • Forecasting & predictive maintenance
  • Approved write-back to the ERP
  • AI assistant for the whole company
What it costs to run

About the cost of an office and its electricity.

Two simple lines, nothing hidden. You rent the house (the place it lives) and you pay for the fuel (the work it does). A normal month — one small server, the database, the scheduler running your reports and scans, around 15 people, a few hundred tasks handled.

The house · rent

Infrastructure

Runs the app, stores the data, runs the daily schedule. Hosted on AWS.
US$120–250 / month
Fixed cost. Barely moves whether 5 people use it or 50 — like renting the office.
The fuel · per task

AI model usage

Every task the employee completes — reports, scans, investigations, answers. ~300–600 tasks/month on a low-cost model.
US$15–35 / month
Variable cost. Grows gently with how many tasks you run — like electricity.
Total to operateNormal monthly operation
≈ US$200 / monthroughly RM 950 · range US$135–285
The house is ~85–90% of the bill and stays steady — so this is a predictable monthly line, not a surprise.
The fuel is the small variable part that scales with tasks. Routine work runs on a cheap model; only hard tasks use a pricier one.
What we need from you

A few decisions get us moving.

The technology is ready. To start the pilot, leadership only needs to point us at the right first targets.

1

Which three questions does management spend the most time chasing answers to?

2

Which reports take the most manual effort to prepare each month?

3

Which abnormal cases hurt most when we catch them late?

4

Which data is sensitive, and who is allowed to see it?

5

Who approves company knowledge before the AI relies on it?

6

What would make this pilot a clear success in 90 days?

The vision

From manual reporting to a real-time decision brain.

One intelligent layer connecting your ERP, dashboards, knowledge, and leadership experience — so the company can finally answer, in seconds:

What happened?
Why did it happen?
What should we do next — and who should act?
Has this happened before, and what did we decide?