Malaysia's public health system has been digitising for decades, yet paper is still part of daily work in many hospitals and clinics. Malay Mail's recent analysis notes that the Health Ministry is aiming for electronic medical records (EMR) at all 150 public hospitals and more than 3,000 primary care facilities by 2029. That is a long programme, and it is about public facilities.
Private clinics and group practices face a similar gap now, without a national rollout to wait for. Referral letters, lab reports, supplier invoices and panel claim forms still arrive on paper, as scanned PDFs or as photos on a phone. Someone has to read each one and type it into another system. This article looks at why that paper persists, and how Document AI can take on the reading and data entry while a clinic keeps the systems it already uses.
Why paper survives in clinics
Paper is not a sign that a clinic is behind. It usually survives because it still solves a problem that the software around it does not.
Documents arrive from outside your control
A clinic can choose its own clinic management system, but it cannot choose what other parties send. A specialist writes a referral reply on letterhead. A laboratory returns results as a PDF. A pharmaceutical distributor sends a printed invoice with the delivery. A panel company or insurer has its own claim form. Each comes in a different layout, and many come in more than one layout over a year.
Paper is quick at the moment of care
Writing a note by hand or ticking a box on a form takes seconds. When the waiting room is full, staff choose the fastest route, and they deal with the typing later, often at the end of the day.
Systems do not talk to each other
Many clinics run a clinic management system for appointments and billing, accounting software for suppliers, and spreadsheets for claims follow-up. When these do not connect, paper and email become the link between them. A printout is easy to pass from the front desk to the accounts clerk.
Change is risky during opening hours
Replacing a core system means training, migration and a period where things can go wrong. For a clinic with a small team, that risk is real, so the safer choice is often to carry on as before.
What the gap costs a clinic
The cost of paper is mostly time and error rather than paper itself. Staff retype the same details more than once: patient name, IC number, test codes, invoice totals. Each retyping is a chance for a mistake. Lab results can sit in a tray before they are attached to the right patient record. Supplier invoices can be paid late, or twice. Claims can be returned because a field was missed, which delays cash coming in.
There is also a records risk. Loose paper and PDFs in shared folders are hard to control. Who has seen a document, where it is stored and when it should be destroyed are all harder questions to answer when records are scattered.
How Document AI bridges the gap
Document AI is software that reads documents the way a trained clerk would, then passes the data on. It does not ask a clinic to change how documents arrive. It works on what comes in.

Step 1: Capture
Documents are collected from where they already land: a shared email inbox, a scanner at the front desk, or photos taken on a phone. Nobody has to learn a new routine beyond putting the document in one place.
Step 2: Read and extract
The system reads the document, works out what type it is (a lab report, an invoice, a referral letter, a claim form) and pulls out the fields that matter. For an invoice, that may be supplier, date, items and total. For a lab report, it may be patient identifiers, test names and results.
Step 3: Check
Extracted data is checked against rules you set. Does the IC number match an existing patient? Does the invoice total match the line items? Is a required claim field empty? Anything uncertain is sent to a person to review rather than passed on silently. In a clinic setting, a human should always confirm anything that touches clinical decisions.
Step 4: Post into your systems
Approved data is posted into the clinic management system, accounting software or claims tracker the clinic already uses, through an API where one exists or a structured export where it does not. Our integrations page lists the kinds of systems this can connect to. The original document is filed and linked, so staff can open it when they need to.
Where to start
Not every document is a good first candidate. A practical first project tends to share three features: high volume, a predictable set of fields, and a clear place where the data should end up.
Supplier invoices
Invoices are usually the easiest place to start. They are not clinical, the fields are well understood, and the result goes into accounting software. This lets the team see how the system behaves before it touches patient information.
Panel and insurance claims
Claim forms are repetitive and errors are costly because returned claims delay payment. Checking forms for missing fields before submission can be a useful second step.
Lab reports and referral letters
These carry clinical information, so they deserve more care. Start with routing and filing (matching the document to the right patient and alerting the right doctor) before extracting individual results.
For more on how we approach clinics and group practices, see our healthcare page.
Handling patient data under the PDPA
Health information is treated as sensitive personal data under Malaysia's Personal Data Protection Act 2010 (PDPA), so a Document AI project in a clinic has to be designed around data protection from the start, not added at the end. The points below are practical design choices, not legal advice; a clinic should confirm its own obligations with its adviser.
- Collect only what you need. Extract the fields the next step requires and leave the rest in the original document.
- Control access. Limit who can see each document type. Front-desk staff do not need lab results, and the accounts clerk does not need referral letters.
- Keep a record. Log who viewed, approved or changed each document, so questions can be answered later.
- Decide where data is stored and processed. Ask any provider where documents are hosted, whether data leaves Malaysia, and how it is encrypted in storage and in transit.
- Set retention rules. Agree how long each document type is kept and how it is destroyed when that period ends.
These questions are worth asking whoever builds the system, including us.
Document AI and EMR are not competing choices
A clinic that adopts Document AI now is not betting against EMR. The work of getting clean, structured data out of paper is useful whichever records system a clinic ends up with. If a clinic later moves to a full EMR, the extraction rules and integrations can be pointed at the new system, and the archive of filed documents is already organised.

In other words, the choice is not between waiting for a complete system and doing nothing. It is about removing the retyping that happens today, one document type at a time.
A sensible first step
Before choosing any software, it helps to count. For two weeks, note which documents arrive, how many of each, and how long staff spend on them. That list usually makes the first project obvious.
If you would like a second view, our AI audit looks at where documents enter your clinic, which tasks are worth automating first, and what the data protection requirements are. You can also read how we work with SMEs from first call to handover.
Codech Solutions builds AI agents, document automation and business insights for SMEs in Malaysia and Hong Kong.
Frequently asked questions
Do we need to replace our clinic management system to use Document AI?
No. Document AI reads incoming documents and posts the data into the systems you already use, through an API where one exists or a structured export where it does not.
Which documents should a clinic automate first?
Supplier invoices are usually the simplest start because they are not clinical and the fields are predictable. Claim forms are a common second step. Lab reports and referral letters need more care, so start with routing and filing before extracting results.
Will the system make clinical decisions?
It should not. Document AI extracts and checks data, and anything uncertain or clinically relevant is sent to a person to confirm.
How is patient data protected?
Health information is sensitive personal data under the PDPA. A sound design limits the fields extracted, restricts access by role, logs who viewed or changed each document, agrees where data is stored and sets retention rules. Confirm your own obligations with your adviser.
Is this wasted effort if we move to a full EMR later?
No. The extraction rules and integrations can be pointed at a new system, and documents already filed and linked remain organised.
