Ting Guides · For the COO

Where can AI help run a Saudi hospital or clinic group, outside clinical care?

The short answer

Start with the administrative work around care, not with diagnosis. In Saudi hospital and clinic groups, check first the appointment and call-centre load, the pre-authorisation and claims paperwork exchanged with insurers, pharmacy and supply stock across branches, and management reporting. Each needs a named process owner, read access to the source systems and strict health-data controls.

In this guide9 sections
  1. What it depends on
  2. Which operational use cases should a provider check first?
  3. What has to be in place before anything is built?
  4. How do operations AI projects in healthcare go wrong?
  5. Where should AI stay out of an operations project?
  6. What changes for healthcare providers in Saudi Arabia
  7. Rasikh Care: one pre-authorisation request, by hand and with AI preparing it
  8. Questions people ask
  9. What Ting does and doesn't do here
  10. Sources

What it depends on

  1. Whether the task is administrative or clinical

    Drafting a booking, assembling the documents an insurer asks for, or flagging a drug that is running low are operational tasks. Suggesting a diagnosis, triaging symptoms or recommending treatment is clinical work with its own regulatory path. Draw that line in writing before any design starts, and keep this kind of project on the operational side of it.

  2. How patients and insurers reach you today

    Count the channels: the call centre, each branch's front desk, the app, WhatsApp, referral letters and the insurers' replies. If agents cannot see every branch's schedule on one screen, an assistant answering those calls will not see it either.

  3. What your systems let you read

    List what the hospital information system, the scheduling module, the pharmacy and inventory system and the revenue-cycle tools expose: a documented interface, read-only views or only screens. The first projects should read through those interfaces, under a service account limited to the fields each task needs.

  4. How tightly health data is held

    The Personal Data Protection Law defines Health Data as personal data about an individual's physical, mental or psychological condition or the health services they received, and lists it among Sensitive Data (Article 1). An assistant that reads appointment notes or claim attachments is processing it, so its access has to be designed in from the first line.[4]

  5. Who owns each process

    Claims belong to a revenue-cycle lead, bookings to the call-centre or patient-access head, stock to the pharmacy and supply managers. Without that owner, no one decides what the assistant may do, reviews its drafts or answers when the numbers move.

Which operational use cases should a provider check first?

These five recur across hospital and clinic groups. Each is a task around care, done by administrative staff, where an assistant prepares the work and a person still decides.

Use caseWhat the AI doesSystems it readsWhat to check before building
Appointment and call-centre requestsSorts incoming requests into booking, rescheduling, cancellation, results and directions, and drafts the booking or the reply for an agentThe scheduling module of each branch, the clinic and doctor directoryHow many requests are routine, and whether the schedules in the system match what branches really run
Pre-authorisation preparationAssembles the request and its supporting documents from the patient's file for a coordinator to review and submitThe hospital information system and the insurance moduleThe reasons insurers asked for more information or declined over recent months, by insurer
Claim rejections and resubmissionsGroups returned claims by reason and drafts the correction for a coder or biller to checkThe revenue-cycle and billing toolsWhether rejection reasons are recorded the same way by every branch
Pharmacy and supply stock across branchesFlags items running low in one branch while another holds stock, and drafts the transfer or reorder requestThe pharmacy and inventory systems of each branchWhether the same item carries the same code in every branch
Operational reportingAssembles the weekly pack on waiting, no-shows, pending authorisations and stock, and drafts the commentaryThe scheduling, revenue-cycle and inventory systemsOne agreed definition for each indicator, and who owns it

What has to be in place before anything is built?

  1. Name the process owner

    One person per use case, with the authority to change the process and the time to review the assistant's output during the first weeks.

  2. Agree read access, field by field

    Ask IT for a separate service identity per connection, and list the exact fields each task reads. A pre-authorisation draft needs the diagnosis codes and the documents the insurer requires, not the whole medical file.

  3. Collect real cases to test against

    With your data protection officer's approval, gather a sample of past requests, returned claims or stock transfers, with what actually happened to each. Test the assistant against them before it sees live work.

  4. Write down what the assistant may never do

    It never sends anything to a patient or an insurer without a person, never answers a clinical question, and never writes to the medical record. Test each of these limits before launch.

  5. Build the health-data controls into the design

    The PDPL Implementing Regulation asks the controller to document every stage of Health Data processing and the person in charge of each stage, to bind processors to the same measures by contract, and to limit processing to the minimum needed (Article 26).[5]

  6. Take the baseline before go-live

    Pull the current figures from the systems themselves: requests handled, authorisations pending, claims returned, stock-outs. Without them, nobody can say later whether the assistant helped.

How do operations AI projects in healthcare go wrong?

FailureWhat it looks likeHow to prevent it
Clinical creepThe booking assistant starts answering questions about symptoms or medicationA written boundary, a hand-off to a clinician or nurse line, and test cases that try to cross it
Faster rejectionsClaims are resubmitted more quickly but fail again for the same missing documentReport the cause of each rejection to its owner and fix the documentation at source
Codes that don't matchStock suggestions move the wrong item because two branches code the same drug differentlyMap item codes across branches before the assistant reads stock
Access that is too wideThe assistant runs under one account that can open every patient filePer-task service identities, the minimum fields, and a log of every lookup
Drafts nobody ownsPrepared requests pile up in a queue no one is measured onA named owner, a daily review, and a count of drafts accepted, edited and discarded

Where should AI stay out of an operations project?

What changes for healthcare providers in Saudi Arabia

General information, not legal advice. Each regulatory point summarises the official source cited beside it, as checked on the date shown under Sources; confirm how it applies to your company with your legal adviser.

  • The national direction is digital

    The Health Sector Transformation Program, one of the Vision 2030 realization programs, says it is expanding e-health services and digital solutions, and sets the goal of extending the unified digital medical record to the whole population. Operational AI fits that direction only when it reads from the systems of record rather than from side spreadsheets.[1]

  • Insurance work runs through NPHIES

    NPHIES, led by the Council of Health Insurance with the National Health Information Center under the Ministry of Health's supervision, carries eligibility checks, prior authorizations, claims, payment notices and the supporting documents attached to them between providers and insurers. An assistant should prepare that work inside your own systems; the submission and the insurer's decision stay where they are.[2][3]

  • Access to health data is kept to a minimum

    The PDPL requires controls that restrict access to Health Data, including medical files, to the fewest employees needed to provide the service, and restrict its processing to the minimum extent needed to provide health services or health insurance programs (Article 23). The Implementing Regulation adds that providers adopt the requirements issued by the Ministry of Health, the Saudi Health Council, the Council of Health Insurance and other regulators (Article 26).[4][5]

  • Names in two scripts

    A patient's name can sit in Arabic in the hospital system and in a transliteration in an insurer's reply or a referral letter. Match records on the medical record number or the identity number, never on the name alone.

Rasikh Care: one pre-authorisation request, by hand and with AI preparing it

ExhibitWorked example

Ting

Rasikh Care, the fictional clinic group of Rasikh Holding, prepares pre-authorisation requests by hand at each branch. This is how the work splits when an assistant prepares the request and a coordinator still submits it.

StepToday, by handWith AI preparing the work
Spot that a service needs authorisationThe front desk checks the insurer's rules from memory or a printed listThe assistant flags the booked service against the insurer's rules held in the system
Gather the documentsA coordinator searches the patient's file for reports, results and the doctor's notesIt lists the documents this insurer usually asks for and attaches the ones already in the file
Spot what is missingGaps show up when the insurer asks for more informationIt marks a missing report before submission and sends a task to the doctor's assistant
SubmitThe coordinator enters the request and submits itUnchanged: the coordinator reviews the prepared request and submits it
Learn from repliesReasons for requests for more information sit in individual emailsEach reply's reason is recorded by insurer and service for the revenue-cycle lead to review

The decisionNothing in this version decides anything clinical or financial: the doctor documents, the coordinator submits and the insurer decides. The assistant only removes the searching and catches gaps earlier, which is why it is a sensible first project.

Source: Illustrative example. Rasikh Care and Rasikh Holding are fictional: not a Ting client and not client data.

Questions people ask

Can AI approve pre-authorisations or claims?
Not on the provider's side. The insurer decides. On the provider's side, AI prepares the request, checks it for gaps and tracks the reply, and a coordinator still reviews and submits it.
Do we need to replace our hospital information system first?
No. The first projects read from the systems you already run, through their interfaces or read-only views. Replacing a hospital information system is a separate decision with its own reasons.
Is a patient-facing chatbot the right first step?
Only when most requests are routine and the schedules in your system are reliable. Otherwise start on the agents' side, where a person checks every draft before a patient sees it.

What Ting does and doesn't do here

What we do

  • In the AI Transformational Audit we trace how appointment requests, pre-authorisations, claims and stock requests actually move between your branches, people and systems, and where they wait or are typed twice.
  • You receive the Executive Report, maps of where the work waits, a ranked list of what to build first by impact and ease, and the expected return with every assumption written down.

Where Ting helps: AI Transformational Audit

What we don't do

  • The Audit doesn't cover diagnosis, triage or treatment decisions; it studies how the administrative work around care runs.
  • We don't promise savings figures, and the Audit is not a roadmap or a phased plan; the optional next step is a Blueprint for the opportunities you select.

Sources

Primary sources only, each read in full on the date shown.

  1. Health Sector Transformation ProgramSaudi Vision 2030 · checked
  2. NPHIES services: eligibility, prior authorization, claims, payment notices and supporting information (Arabic)NPHIES · checked
  3. About NPHIES: led by the Council of Health Insurance with the National Health Information Center (Arabic)NPHIES · checked
  4. Personal Data Protection Law (Royal Decree M/19 of 9/2/1443H, amended by Royal Decree M/148 of 5/9/1444H)SDAIA · checked
  5. Implementing Regulation of the Personal Data Protection Law (published 1445/2/22H; English translation, July 2025)SDAIA · checked

Ready to try this
on your own workflow?

Contact us