Many organisations still approach AI adoption as they would traditional enterprise software procurement. They select a vendor, standardise on one model and scale it across the business. Yet AI performance depends on context. A model built for code generation may boost developer productivity, while another may be better suited to security analysis, governance or compliance requirements.
In Asia Pacific, where organisations operate across diverse regulatory environments and infrastructure needs, flexibility matters. Enterprise AI delivers the most value when teams can choose the right model for the task. Some workloads require advanced reasoning and large-scale models, while others benefit from smaller, specialised models tuned to specific domains or local requirements. This makes model choice a strategic capability. Organisations that can mix and match models based on use case are better positioned to scale AI securely, efficiently, and with greater business impact.
AI delivers more value when it supports the full software lifecycle
AI adoption today focuses almost entirely on accelerating code generation. But coding represents a fraction of what developers actually do. According to GitLab's DevSecOps research report, developers in Singapore spend only about 13% of their time writing code. The rest goes to planning, reviewing code, testing, debugging, managing dependencies, coordinating with teammates, and navigating compliance requirements.
This creates an AI paradox, where AI is accelerating coding, but disconnected toolchains and manual coordination slow overall productivity enough to cost nearly a full workday per developer each week.
To break out of that paradox, AI needs to work across the entire development lifecycle, not just code generation. Different activities across the software lifecycle carry fundamentally different performance requirements:
- Speed-critical tasks like auto-completing code or suggesting fixes during active development need sub-second response times, which might favor smaller, locally hosted models.
- Quality-critical tasks like architectural planning or security analysis justify the cost of frontier models with superior reasoning.
- Cost-sensitive tasks at high volume, such as running tests or updating dependencies across hundreds of repositories, require cost-effective options.
This is why multi-model customisation is important. Not all tasks across the software lifecycle carry the same value. Standardising on one model can result in overpaying for some functions or underserving others.
The organisations that get this right build systems flexible enough to route each task to the model that best fits its performance, quality, and cost profile.
Scale AI with a multi-model approach
The practical move is matching model cost to task value.
For high-volume, routine work such as writing commit messages, summarising log files, or writing test cases, teams lean toward cheaper, faster options, including open-source models where feasible. For tasks that demand complex reasoning, teams pay for greater capability. For specialised models that are more deterministic, teams might be willing to pay a premium for infrastructure-as-code generation or high-accuracy data transformation.
Being able to choose between different models based on the task provides a hedge against performance differences, pricing swings, and the reality that providers may sunset products or exit the market entirely.
That flexibility comes from three sources, each with tradeoffs.
- Commercial frontier models from Anthropic, OpenAI, and Google deliver strong performance and improve continuously, but create dependence on vendor roadmaps and pricing.
- Self-hosted commercial or open-source models give you control over data residency, costs, and availability, but require infrastructure management and, in the case of open source, still can't handle agentic workflows.
Domain-specific models you've trained can outperform general models on narrow, high-stakes tasks where you have unique data and clear success criteria, but they require specialist expertise and can be operationally expensive.
Each approach involves trade-offs. The key is building systems that let you use all three strategically.
Apply FinOps discipline to AI operations
Model flexibility only creates value if you can manage the economics behind it. The price gap between models is substantial. Complex reasoning models can cost 500% more per request than general-purpose models that work fine for routine tasks.
Model routing, the ability to define which models get used for which tasks, becomes critical here. A code review might route to a frontier model, while commit message generation uses a faster, cheaper option.
But routing alone isn't enough. Enterprises need the same financial controls they apply to cloud infrastructure, including quotas to prevent runaway spending, limits to enforce budget discipline, and chargeback models that allocate costs to the departments consuming AI resources. Without these guardrails, AI adoption becomes difficult to justify at scale.
This is why FinOps practices are extending to AI. IDC estimates that organisations will underestimate their AI infrastructure costs by 30% through 2027, and that combining GenAI with FinOps processes will be essential for managing this complexity. Organisations that treat AI spend like cloud spend, with visibility, accountability, and governance, position themselves to scale AI successfully.
AI returns depend on shared context
Model flexibility also depends on context. AI needs information spread across systems that weren't designed to work together by default. A developer debugging an issue might need to reference the work backlog, pull recent Slack discussions, and review app performance metrics in Grafana. If every system has its own AI experience and none of them connect cleanly, AI creates friction instead of removing it.
Fortunately, recent open-source developments, such as Model Context Protocol (MCP), address this by enabling tools to share relevant context and actions within a single workspace.
This shared foundation enables meaningful customisation, and the most effective customisation works in layers, each one encoding how your organisation performs work.
Most developers rely on pre-built agents and workflows that make AI available for common tasks without requiring expertise. Power users shape how a model operates through detailed prompting, essentially teaching it to follow their organisation's playbook. Experts connect multiple agents into governed flows that mirror how humans deliver work, with strict review protocols in place.
Organisations see the strongest ROI when they design systems where AI operates within defined context and accountability structures, and in which teams can connect different models based on their requirements, whether those are frontier commercial models, self-hosted instances for data residency, or specialised models trained for domain-specific work.
Focus on orchestration at enterprise scale
Enterprise AI succeeds when it fits naturally into the way teams already work. Leading organisations in Asia Pacific are building environments that support multiple models while maintaining consistent governance, security and operational standards. They evaluate models across workloads for quality, speed and cost, and provide teams with transparency into how those decisions are made.
This enables flexibility to use frontier models for advanced reasoning, self-hosted models for regulated workloads and specialised models for domain expertise, all within a unified control framework. The next phase of enterprise AI will be shaped by orchestration, as organisations connect the right models, workflows, and governance structures to support software delivery at scale.