Why enterprise AI stalls after adoption in APAC

Across Asia-Pacific, the conversation around enterprise AI has shifted from whether to adopt it to why so many deployments stall after the pilot phase. The pattern is familiar: organizations invest in tools, training and infrastructure, run successful proofs of concept, and then watch the value dissipate once the technology moves into day-to-day operations. For Mimrah Mahmood, Vice President of Enterprise for APAC at Meltwater, the explanation lies not in the algorithms themselves but in the information that feeds them.

The disconnect, Mahmood argues, is structural. News, social media, consumer behavior, customer feedback, CRM records, support tickets, surveys and internal reporting tools all exist within most large enterprises, yet they rarely connect in a way that gives leaders a full and usable picture. Each team may see a different slice of reality, but AI systems need connected context to produce insights that executives can trust. When that context is missing, the technology may still work, but the intelligence behind it remains incomplete.

The Intelligence Delivery Problem

Mahmood frames the issue as an intelligence delivery problem rather than an AI access problem. The tools have become easier to deploy, but useful intelligence still needs to reach the right people, in the right format, at the right point of decision.

AI tools have become easier to access, but useful intelligence still needs to reach the right people, in the right format, at the right point of decision.

He uses an air traffic control analogy to make the point. Adding more aircraft to the sky does not make an airport more effective; what matters is whether the control tower can see all movements clearly, understand the conditions around each flight, and guide decisions safely in real time. Enterprise AI works the same way. The value does not come from having more systems producing more outputs, but from giving decision-makers a clear, connected view that helps them act with confidence.

In Mahmood's experience, the output of enterprise AI often stays trapped in dashboards, reports or isolated systems. The insight may exist, but it does not always reach the business workflow where a leader is deciding on a campaign, reviewing a market risk, responding to a customer issue or planning the next move. For businesses in Asia, where markets can shift quickly and customer sentiment can change before it appears in internal reports, that gap carries real operational risk.

Why Generic Inputs Fall Short

A second layer of the problem, Mahmood says, is that generic AI-ready inputs rarely reflect how a business actually makes decisions. A clean, structured dataset may be easy for a system to process, but that does not mean it captures the way an organization understands its customers, categories, competitors and risks.

AI becomes far more useful when it is shaped around the business, rather than forcing the business to act on generic inputs. The stronger foundation is intelligence that reflects the organization's language, markets and decisions.

One company may define customer risk through service complaints and churn signals. Another may define it through public sentiment, media coverage and competitor activity. A regional business may also track the same issue differently across Singapore, Indonesia, Malaysia or the Philippines because each market carries different customer behaviors, regulations, languages and cultural signals. If the inputs are too broad, the output may sound useful but still fail to match the questions leaders need to answer.

Workflows matter just as much as inputs. A communications team may need early warning on reputational risk. A marketing team may need to understand a shift in customer interest. A leadership team may need a clearer view of how market signals affect growth plans. Each group needs intelligence that fits its role and decision point, and insights lose value when they sit outside the systems and teams that need them.

The Operational Reality in APAC Markets

Mahmood is candid about the practical challenges of closing these gaps across the region. Markets shift quickly, customer sentiment can change fast, and competitive activity can reshape decisions before it appears in internal reports. A retailer using AI to guide pricing based on past sales and margins, for example, may miss a competitor launch that is changing how customers think about value. A bank using AI to personalize customer engagement may rely on transactional data alone and miss a shift in public trust or customer expectations.

Enterprise AI often struggles after adoption because the information behind it does not reflect the full reality of the business environment.

In both cases, AI may produce a reasonable answer based on the data it can see, while missing the wider market signals that explain what is changing. The technology may work, but the intelligence behind it remains incomplete, and gaps in business understanding can move faster across the organization as a result.

Mahmood points to emerging approaches such as model context protocols as one practical response. These approaches address a recurring enterprise challenge: AI needs a consistent way to access trusted business and market intelligence, rather than relying on disconnected inputs or one-off integrations. They reflect a broader shift in how organizations think about AI deployment, moving from a focus on models to a focus on the context those models can draw from.

The Next Phase of AI Maturity

Looking ahead, Mahmood expects the next phase of enterprise AI maturity to be defined by delivery rather than deployment. Companies that connect intelligence to decision-making workflows will be better placed to turn AI from a tool for analysis into a source of trusted action.

The next phase of AI maturity will be shaped by delivery. Companies that connect intelligence to decision-making workflows will be better placed to turn AI from a tool for analysis into a source of trusted action.

For APAC organizations, that means asking harder questions about where intelligence is created, how it is interpreted, and how it reaches the AI systems and business workflows that support decisions. It also means accepting that regional complexity is high and that a single issue can move through markets in different ways, across different channels and at different speeds. The companies that build intelligence around their own language, markets and decisions, rather than forcing their teams to act on generic inputs, will be the ones that turn scattered information into trusted intelligence that supports decisions at the right moment.