Scaling AI from Pilot to Operations (February 27, 2025)

On February 27th, the Asia Growth Exchange (AGX) virtually convened participants from China, Singapore, the Philippines, Germany, and France, who, who joined a virtual discussion to confront one pressing question: How do you successfully scale AI from isolated pilots into core operations? Facilitated by Erik Walenza and kicked off by Etienne Charlier from Asia Growth Partners, the conversation moved beyond familiar ideas of automation and efficiency, uncovering a few fresh insights worth remembering.

AI as Out-Tasking: A Shift in Mindset.

The group agreed on a shift in perspective, viewing AI not as simple automation but as “out-tasking”—strategically delegating specific tasks traditionally performed by people.

“For the first time, we communicate with machines using our own words. This changes everything about how knowledge is managed.”

“AI is about doing tasks, not taking over entire roles. This changes the narrative from job displacement to task redistribution.”

This subtle yet powerful distinction emphasizes active management rather than passive replacement, requiring clear global governance while allowing for local implementation flexibility.

You Can’t Save Yourself to Success

The dialogue emphasized a common oversight: organizations often limit AI to cost savings.

“It’s very hard to write AI cases that are not about cost savings. But you can't save yourself to success.”

Participants highlighted the necessity of defining clear metrics—such as customer satisfaction, revenue growth, or improved decision-making—from the outset. As one participant succinctly noted, "What we learned is that if you don’t establish a clear way to quantify ROI, then scaling AI becomes a challenge."

Structured Incubation: Experiment, Organize, Then Scale

Companies frequently jump from pilot to full-scale implementation, skipping essential steps. One participant pointed out their organization had broadly solicited AI ideas initially, leading to duplication and unnecessary administrative complexity.

“We asked everyone to dream up AI use cases, but this created redundancy and excessive paperwork.”

Organizing and consolidating AI use cases early by business function—such as sales, operations, or manufacturing—streamlines effort and aligns initiatives with core business priorities. Participants recommended a structured incubation step between initial experiments and full deployment to reduce risk.

“There’s an intermediate place between proof-of-concept and full-scale integration. Do something first, fine-tune it, then launch it in phases.”

This measured approach allows companies to refine and mature AI projects before integrating them fully.

Turning Potential Blockers into Strategic Allies

Proactively involving traditionally cautious groups such as Legal and IT from the start emerged as a best practice.

“We established a Digital Council, bringing Legal, Finance, IT, and business leaders together early. Suddenly, blockers became accelerators, enabling us to scale AI more effectively. Make sure you have legal teams working closely—they're not roadblocks; they are your safety valve. Without them, scaling safely isn't possible.”

Early collaboration transforms cautious stakeholders into proactive champions of AI-driven innovation.

Overcoming Resistance: People, Culture, and Leadership

One of the biggest barriers to AI adoption is employee resistance, often driven by fears of job displacement. Participants highlighted the critical importance of positioning AI explicitly as an enabler, not a replacement. “People are reluctant to engage with AI because they fear it will replace them,” one participant explained. To address this fear, clear communication is essential. “We told our people that AI is coming, and we need to work together to future-proof jobs.” By openly discussing AI’s role as an enhancer of human capabilities, organizations shift employee perspectives from anxiety to opportunity.

Upskilling emerged as another crucial tool to overcome resistance. Employees unfamiliar with AI often struggle to envision its practical relevance, leading to hesitation or resistance. A participant remarked, “When we first asked our finance colleagues for AI use cases, they didn’t know what we meant. But after introducing targeted training programs, we saw a significant shift in engagement.” Practical AI literacy programs empower employees to view AI as an asset rather than a threat.

Leadership’s role in supporting AI adoption was consistently emphasized. Without clear executive support, AI initiatives risk stalling.

“You need leadership to actively manage AI governance and clearly position AI as integral to the company’s future.”

China’s Unique AI Landscape

Balancing global AI standards with localized requirements is critical. Companies operating in China are adopting a pragmatic approach with three options: landing global solutions with local infrastructure, buying local solutions, or building in-house capabilities specifically for the Chinese market.

"China is a walled garden, and with all the regulations coming, it will remain a walled garden for the time being.” “We realized that using DeepSeek allowed us to conduct AI experiments locally while staying compliant with Chinese regulations.”

Local AI partnerships and on-premise models address data security and regulatory challenges, maintaining global consistency while respecting regional nuances.