THE fastest way to future-proof Malaysian business isn’t about more artificial intelligence (AI) or less AI, but different AI. Business leaders rarely disagree that AI will reshape competitive advantage. But what exactly are we optimising for?

We have a bad habit of falling in love with numbers. Call times, click-through rates, engagement scores. They look neat, they feel objective, and they fit nicely into a spreadsheet. But anyone who has worked inside a big organisation knows the secret. Once a number becomes the target, it stops telling the truth.

Take call centres. Shorter calls are supposed to mean efficiency. But they often mean staff rushing people off the line, problems half-solved, and customers calling back even angrier. The metric improves. The service does not. We used to say “measure what you care about”. Now we care about what is easiest to measure. If it cannot be counted, it gets ignored. Judgement, trust, quality, these are messy, so they lose out to whatever fits on a dashboard.

Now add AI to the mix. AI systems are built to maximise a single score. Give them a number and they will chase it with obsessive focus. They do not understand intent, values, or common sense. They understand winning the game you set up. That is where the danger lies. AI can make your engagement graph soar while making your product worse. The real risk of AI is not that it will fail. It is that it will succeed too well, but at optimising the wrong things, faster than we can notice the damage.

If we let numbers replace judgement, AI will not save us from bad decisions. It will automate them. Malaysian business needs a more operational framework: ProSocial AI — systems designed and deployed to deliver measurable benefits for people and planet while staying bankable and competitive.

In boardrooms, “responsible AI” often sounds like legal cover. ProSocial AI flips the script. It treats responsibility as smart risk management and a way to build better, more valuable systems. Why? Because AI doesn't just automate tasks; it automates choices. Which customers get approved for credit, insurance, or hiring. Which patients get prioritised in triage systems. Which communities receive public services. Which stories spread through recommendation engines.

When we deploy AI without a clear prosocial goal, we quietly outsource moral decisions to whatever crude metric is easiest to code. Ethical dilemmas do not disappear; they just get simplified and ultimately replaced by a number. That can produce "success" on paper and failure in the real world: reputational blowback, regulatory exposure, talent attrition, and fragile systems that break under stress.

ProSocial AI starts with a different set of questions: Success for whom? Over what time horizon? With what wider consequences beyond the balance sheet?

Malaysia is actively pushing AI adoption via the National AI Roadmap 2021–2025. At the same time, corporate accountability is tightening through sustainability reporting momentum, with Malaysia's National Sustainability Reporting Framework (NSRF) aligned to International Sustainability Standards Board (ISSB)-style disclosure. The strategic connection many firms are missing is that IF your AI increases short-term efficiency but damages trust, fairness, or planetary health, the bill will arrive later. It shows up as heavier regulation, contract loss, higher cost of capital, supply-chain headaches, and a weaker brand.

ProSocial AI does the opposite. It turns AI investment into a defensible narrative: We are building systems that grow profitability while strengthening social trust and the long-term resilience of people, planet and profitability.

This matters, especially in Asean supply chains, where buyers increasingly demand verifiable data on emissions, labour practices, and governance. A 2024 McKinsey survey found that 68% of global procurement teams now require ESG (environmental, social, and governance) metrics from suppliers. ProSocial AI isn't a feel-good theme; it's a market access strategy.

One helpful distinction emerging in ProSocial discussions is the shift from principles to proof. Ethical AI often asks: "Is this fair?" and then gets stuck arguing about definitions. ProSocial AI asks a harder question: "What concrete prosocial outcomes will this system deliver, and how will we measure and audit them?" 

The ProSocial AI Index uses a 4T framing, Tailored, Trained, Tested, Targeted, to move teams from slogans to evaluation. Whether or not you adopt that exact framework, the core business move is the same: treat social and planetary outcomes as first-class performance metrics, not public relations tools.

For CEOs, ProSocial AI means AI systems that increase human agency, reduce systemic harm, and strengthen long-term resilience while remaining economically viable. In practice, that means four design decisions:

1. Write the "intent spec" before the tech spec: Not "build a chatbot". Instead: "reduce complaint resolution time without increasing vulnerable-customer churn or misinformation". When RHB Bank Malaysia implemented their AI customer service system in 2023, they embedded metrics for customer satisfaction alongside resolution speed, recognising that rapid deflection could harm vulnerable customers who needed human assistance.

2. Measure second-order effects: If your AI cuts costs but increases the burden on existing staff, or employee burnout, it's not efficiency, it's cost-shifting. A manufacturing firm in Penang discovered their AI scheduling system reduced overtime expenses by 22% but increased worker injury rates by 14% due to compressed work patterns. The real cost was higher than the apparent savings.

3. Build for auditability: Assume you'll need to explain outcomes to regulators, partners, and your own people. If you can't trace decisions, you can't defend them. Singapore's Model AI Governance Framework provides useful guidance that Malaysian firms can adapt.

4. Treat planetary health as operational reality, and internal priority: AI expands compute, energy demand, water use and procurement footprints. ProSocial design forces you to ask: "What's the net impact?" Training a single large language model can emit as much carbon as five cars over their lifetimes. This aligns naturally with sustainability disclosure trajectories.

Behavioural scientists have warned us for decades about WYSIATI — what you see is all there is. We make decisions based on the information right in front of us, and ignore what sits outside the frame. In AI projects, what we usually see is model accuracy, speed, and return on investment. These numbers are seductive. They look precise, objective and reassuring. What we do not see, until later, is drift, exclusion, hallucination, manipulation, and wider knock-on effects. By the time those surface, the system is already embedded and expensive to unwind.

ProSocial AI is a structured antidote to this blind spot. It forces slower, more deliberate thinking at exactly the moment organisations are most tempted by quick wins. It pushes leaders to look beyond prediction and performance, and towards governed impact with a clear strategic direction. And it turns a vague aspiration like "AI for good" into a hard-edged business capability: competitive advantage with social licence rather than reputational risk baked in. 

Awareness: List your top four AI use-cases and write the real objective for each, not the proxy metric. Identify what could be unintentionally optimised. "Lower call time" might become "rush customers off the phone".

Appreciation: Map who benefits and who pays. Include employees, suppliers, vulnerable customers, regulators, and communities. If you can't name the "payers", you're probably not seeing them.

Acceptance: Accept that AI is not neutral and that governance is part of performance. If you don't set the intent, the objective function will.

Accountability: Pick two auditable outcome metrics per AI system: one business metric (margin, retention) and one prosocial metric (complaint fairness parity, accessibility, verified emissions impact). Review both quarterly at exco level.  

Dr Cornelia C Walther is a humanitarian practitioner with more than two decades of experience at the United Nations, advancing social change across global contexts. She is a senior fellow at the Sunway Centre for Planetary Health, the University of Pennsylvania’s Wharton School, and the Harvard Learning and Innovation Lab where her work focuses on hybrid intelligence and ProSocial AI. She is an advisor to UNFPA on hybrid intelligence and the European Policy Center on ProSocial AI; and is collaborating with Sunway’s Global Strategy and Competitiveness Centre for ProSocial AI in business in Malaysia. Through the global POZE alliance, she pioneers research and practice to strengthen human agency in the age of AI.