⬤ Kuala Lumpur, Malaysia
AI that fits the way your team actually works
Noctal works with finance, admin and operations teams across the Klang Valley to redesign the parts of their workflows where language models can do real work — without requiring a technical department to run them.
About Noctal
Practitioners, not vendors
Noctal started from a simple observation: most teams in the Klang Valley already spend a large portion of their week on repetitive document tasks — re-keying data from PDF forms into spreadsheets, routing emails to the right person, chasing approvals through WhatsApp. The technology to reduce that load is here. Getting it deployed sensibly is a different matter.
We work across four weeks to six weeks with finance, logistics and administration departments, mapping what is actually happening before proposing anything. The result is a specification your IT team or vendor can build from, and staff who know how to use it.
Based at Wisma Central on Jalan Ampang, we take on a small number of engagements at a time. That keeps the work close and accountable.
What we address
Three problems worth solving
Manual re-keying
Finance and admin staff spend hours copying figures from PDFs into ERP systems. Small errors compound. Language models can extract structured fields reliably — if the flow is designed correctly.
Find out more →
Choosing the wrong model
Public benchmarks rarely match a company's actual tasks. A model that performs well on English reasoning may struggle with Malay-English mixed text or local document formats.
Find out more →
Staff using AI carelessly
Unreviewed outputs copied into client reports or HR correspondence are a real risk. A structured literacy programme establishes checking habits before they are needed.
Find out more →How we work
Principles behind every engagement
These are not aspirational statements. They are the specific commitments that shape how each programme is scoped and delivered.
01 / Current-state first
Map before proposing
We spend the first week understanding what is actually happening, not what the process diagram says. Surprises at that stage are far cheaper than surprises after build.
02 / Staff-legible outputs
Written for the people running it
Specifications and exception-handling guides are written in plain English, not technical shorthand. The team that will operate the process should be able to read and question every step.
03 / Fixed scope, fixed fee
No billable-hour creep
Each programme is priced as a flat engagement. If something takes longer than expected, that is our problem to manage, not an invoice line item to negotiate.
Document workflows
The goal is fewer keystrokes, not a showcase
In a typical trading firm, an accounts assistant might re-enter the same invoice figures four times across different systems before a payment clears. That is not a technology problem — it is a workflow design problem that technology can address once the fields and exceptions are properly defined.
Our Document Handling Workflow Design engagement produces a redesigned flow diagram and an extraction specification your vendor can build against, not a pilot that ends with a slide deck.
Discuss your document flow →
Model evaluation
Test on your material, not someone else's
A company considering whether to self-host a smaller open model or stay on a hosted service needs data, not opinion. The Model Evaluation Bench gives your engineering team a repeatable test harness built on your own documents — purchase orders, support tickets, internal reports — with scoring scripts they can run again whenever a new version is released.
Cost and latency figures are included alongside accuracy. The written comparison sits in your own repository, not ours.
Set up a bench →Why teams trust Noctal
Practical grounding, not theoretical enthusiasm
A few specifics that matter when you are choosing who to work with on something this close to your operations.
Written deliverables
Every engagement ends with documents your team can read, critique and hand to an implementation partner.
Fixed timelines
Programmes run four to eight weeks. Both parties know at the start what is in scope and when it ends.
Bilingual delivery
Team literacy sessions are available in English with supporting materials in Bahasa Malaysia — important for mixed-language departments.
Your data stays yours
Evaluation sets and extracted material are built under the client's own repository. We do not retain operational data after handover.
Our approach
Three things we hold to
Scepticism about hype
We do not sell an AI strategy or a transformation roadmap. We look at a specific workflow, identify where extraction or classification is actually feasible, and help you implement that narrow thing well.
Respect for local context
Klang Valley businesses handle Malay-English mixed text, SST documentation and BNM reporting formats. Our evaluation benches and workflow designs account for these specifics, not generic Western enterprise examples.
Transferable knowledge
After an engagement, your team should be able to run the process, repeat the literacy sessions and re-run the evaluation bench without us. The deliverables are designed with that in mind.
Programmes
Three structured engagements
Each programme has a fixed scope, a fixed price and a defined delivery period. All prices are in Malaysian Ringgit and exclude SST where applicable.
Document Handling Workflow Design
A working session series for finance operations and admin managers in trading, logistics and services firms. We map how incoming forms, purchase orders and scanned correspondence currently move through your company, then redesign the flow with language models placed only where they reduce genuine manual work. Scope covers current-state mapping, a redesigned flow diagram, extraction field definitions, exception-handling procedures and a ready-to-implement specification pack. Three workshops across four weeks.
- ✓ Three facilitated workshops
- ✓ Redesigned flow diagram
- ✓ Implementation-ready specification
Model Evaluation Bench
A repeatable test bench so your engineering team can compare candidate models on your actual tasks rather than public leaderboards. Suited to technical teams choosing between hosted and self-run options for summarisation, extraction or classification work. Scope includes task definition, a labelled evaluation set built from your own material, scoring scripts, cost and latency measurement, a written comparison of results, a handover session with your engineers and the full test harness committed to your repository. Six weeks.
- ✓ Evaluation set from your own data
- ✓ Cost & latency measurement
- ✓ Full harness in your repository
Team AI Literacy Programme
A structured learning programme for non-technical staff in departments of 15–80 people. Covers how current language models behave, where their answers can be wrong, how to write clearer instructions and how to record and review outputs before they are used. Marketing, HR, administration and account management teams across the Klang Valley have gone through this programme. Scope includes six live sessions, printed reference cards in English and Bahasa Malaysia, a company-specific prompt collection, a short written assessment and a facilitator guide for internal leads to repeat future intakes. Eight weeks.
- ✓ Six live sessions included
- ✓ English & Bahasa Malaysia materials
- ✓ Facilitator guide for repeat intakes
How it works
From first call to final handover
Step 01
Scoping call
We spend 45 minutes understanding your current situation — what the document or workflow looks like, who is involved, what has already been attempted. No pitch deck. If the programme is not a fit, we say so at this stage.
Step 02
Engagement delivery
Workshops, evaluation sessions or literacy sessions run on the agreed schedule. Your staff attend in their normal work context — we do not ask for extended off-sites or large blocks of calendar time.
Step 03
Handover & close
All written outputs, scripts and guides are transferred to the client. A final review session confirms everything is legible and usable. The engagement closes cleanly — no retainer, no follow-on dependencies unless you choose them.
Client feedback
From teams we have worked with
"Our accounts team re-keyed invoice data manually for years. After the workflow mapping sessions, we had a clear spec to give our developer. The extraction accuracy on our PDF forms has been very high since deployment."
Siti Azura
Finance Manager, trading firm · Petaling Jaya
July 2025
"The evaluation bench was exactly what we needed. We had been going back and forth on whether to self-host for months. Having the scoring scripts run on our own documents made the decision straightforward."
Reuben Chin
Engineering Lead · Shah Alam
June 2025
"Our HR and marketing teams were using AI tools with no real understanding of when to trust the output. The literacy sessions changed that. Staff now know what questions to ask before passing anything along."
Norhaiza Karim
Head of People, services firm · KL Sentral
July 2025
Questions
Frequently asked
Do these programmes require us to buy any AI software?
Our IT team is small. Can we still run the evaluation bench?
How many people need to attend the literacy sessions?
Can we run more than one programme at the same time?
What happens after the engagement ends?
Are sessions held at our office or online?
Contact
Get in touch
Noctal
We take a limited number of engagements at a time. If you have a specific workflow or evaluation challenge in mind, the scoping call is the right starting point.
147 Jalan Ampang
50450 Kuala Lumpur
Monday – Friday, 9:00 AM – 6:00 PM MYT
Closed on Malaysian public holidays
Our location