Why AI pilots struggle to scale in banks

This article summarizes an episode of Fintech Fireside Asia’s video series featuring Cynthia Siantar, general manager at Dyna.Ai.

Cynthia Siantar, general manager at Dyna.Ai/ Photo credit: Dyna.Ai

Integrating automated systems into major banks often fails due to broken internal reward systems and heavy bureaucracy rather than technological limits. Cynthia Siantar, general manager at Singapore AI-as-a-Service firm Dyna.Ai, says corporate hesitation can kill working software long before it reaches deployment.

To survive this restrictive environment, success requires executives to enforce rapid testing timelines and prioritize immediate financial returns over technical perfection.

The approval delay

Seeking total agreement across a large organization actively destroys technological progress. Adding extra layers of management to an evaluation process creates confusion and permanently stalls deployment.

Siantar believes the enterprise sales cycle in banks is slowed by multiple layers of decision-makers and repeated sign-offs.

She argues that organizational bureaucracy presents a much harder barrier than the software itself, noting, “Even if the tech works, more often than not, the human systems do not.”

Surviving organizational changes
Software quality rarely determines whether a project actually launches. Executives frequently mistake endless committee reviews and a lack of direct ownership for safe strategic planning. This lack of visibility kills forgotten projects the moment a company reorganizes its staff.

Siantar says, “If an AI project is not visible enough to management, it may never happen.”

Teams deliberately choose small and safe tests to avoid internal complaints, which fails to show actual business value and causes leaders to demand even more reviews. Successful projects force scary operational changes, and Siantar notes that resistance occurs because “a lot of people are not willing to change.”

Moving past the need for agreement

Breaking through this bureaucracy requires finding internal champions willing to force new tools into daily operations. Once an initial test concludes, the challenge immediately shifts from proving the technology to spreading that success across departments.

Siantar says that progress relies entirely on finding the right internal team, explaining, “It is driven top-down, and other people will also take the lead.”

Treating vendors as internal partners
Calling external technology providers partners rather than vendors also provides a psychological advantage during this process.

“I do not consider my client as just a pure client,” Siantar explains. “They are my partners, because we are partnering with them to refine our solutions along the way with their key inputs.”

Planning tests that survive

The myth of internal control

Wrong goals and blockers