Why this is a business problem, not only a health one

A company operating across the three countries has to act on three different rulebooks. Singapore uses the PSI and a separate 1-hour PM2.5 band, Malaysia acts on API thresholds of 100, 150 and 200, and Indonesia's picture is told in hotspot counts.

The sources also disagree. On 8 October, Malaysia's Department of Environment explained that readings on third-party platforms can differ from its own because of different scales, calculation methods and data sources, and pointed the public to its official system. Any enterprise that has put AI on top of messy internal data will recognise that problem.

Then come the knock-on effects. When Johor closes schools for two days, or Sarawak closes 591 schools as it did on 20 August (Wikipedia), the working day changes for parents on your payroll too. That is our inference, not a figure from the sources, but it is the point where an air-quality story becomes an operations story.

The haze is not an IT problem. Responding to it well is: conditions change by the hour, several sources must be reconciled, rules trigger actions, and people need clear answers fast.

Where Cxrus services meet the haze

Cxrus Solutions has offices in Singapore, Malaysia and Indonesia, the same three countries the smoke is crossing. Here is how our AI services map onto the problem.

1. AI Workflow turns thresholds into action.
   AI Workflow covers smart approval flows across departments, real-time notifications and task routing, and AI assistants for operations, HR, finance and IT. A haze playbook is exactly that: when the forecast crosses a band, alert site managers, trigger approval to move outdoor work or shift to remote, and keep a record of who confirmed.
2. AI Report gives you one reading and one narrative.
   AI Report turns your data into narrative summaries and recommendations, and works with existing tools. Picture a morning brief that sets the NEA forecast, Malaysian API readings and Indonesian hotspot counts side by side, flags where they disagree, and names the official source for each.
3. Private LLM keeps sensitive staff data in-house.
   NEA's advisories single out the elderly, pregnant women, children and people with chronic lung or heart disease. A haze response inevitably touches health-related and location data. A Private LLM keeps that inside your environment, with guardrails and governance, aligned to PDPA and SS 714:2025 as described on our service page.
4. AI Infrastructure and DMS with AI keep operations steady.
   When teams work remotely for days at a time, the backbone carries more load. AI Infrastructure provides secure data pipelines and stable, controlled deployment, while DMS with AI builds automation and insight into managed services.
5. Answer Engine Optimization makes sure AI gives the right answer about you.
   Customers now ask AI assistants whether a venue is open, a delivery is running or an event is on. AEO restructures your content into answer-ready, citable form so that when an AI engine answers, it reaches for your accurate status update rather than a stale one.

The lesson: governed AI, not autopilot

Our earlier piece on a missing flag argued that the problem is never generative AI itself but unmanaged AI output. The haze raises the stakes. A banner with the wrong flag is embarrassing; a staff notice built on a stale air-quality reading is a safety risk.

The same three guardrails apply:

1. Human in the loop: AI drafts the alert and proposes the action; a named person approves before anything goes to staff or customers.
2. Private and secure models: Health-adjacent and location data stays inside your environment.
3. Compliance and quality checks: Every message cites its official source (NEA, DOE or the relevant agency) and is checked against it before release.

What to do this week

Haze season will not end on a single date. Rain over the fire areas or a change in wind will decide it, so plan for good days and bad days to alternate.

• Write down your haze triggers (PSI, API, 1-hour PM2.5) and who acts on each.
• Pick one official source per country and make it the only one your teams and your AI cite.
• Decide where a human must approve before AI-drafted messages go out.
• Check which staff and location data your AI tools can see, and where it is processed.

If you want a second pair of eyes, Cxrus can run a structured AI Audit to uncover misconfigurations, governance gaps and operational blind spots before they become incidents. Book a demo to see how people and AI can work together.

Precision. Discipline. Made simple.