China’s leading digital bank has folded application and review for small-business credit into a single AI conversation, with humans making the final call.
For millions of small-business owners in China, borrowing has long meant paperwork, branch visits and days of waiting for an answer. MYbank, the leading digital bank in China and an associate of Ant Group, is compressing that process into a conversation.
In one case the bank is highlighting at the Singapore FinTech Festival (SFF) 2026, a restaurant owner surnamed Wang, preparing to open a second location, asked for a higher credit line. Over a few rounds of conversation with Lark 2.0, the bank’s AI agent for inclusive finance serving small and micro-sized enterprises (SMEs), his business plan and funding gap were clarified, his supporting materials completed and the request submitted for review. After human approval, Wang received a financing plan offering a credit line of RMB 400,000, available in tranches as his new store progresses. From request to finalized plan: about 10 minutes.
For a closer look at how the model behaves in practice — what it asks, how it decides and where the conversation ends — a walkthrough of the private banker in the chat window is a useful guide.
The scale behind the service matters. MYbank opened Lark 2.0 at the Inclusion Conference on the Bund in September 2026, extending credit, bill discounting, accounting and tax services to 42 million small and micro business owners — a customer base that manual review alone would struggle to keep pace with.
Fewer Wrong Turns, Faster Builds
The changes run deeper than the application step. MYbank’s AI quantitative risk-control workbench has cut the mishandling of legitimate business owners — applicants wrongly delayed or flagged during checks — by around 40%.
The AI human-review risk-control workbench has lifted capacity for complex loan applications nearly 20-fold while keeping asset quality stable. And the AI product R&D workbench has shortened the development of a product’s first verifiable version from 33 days to 2 days.
For owners, the difference is practical: fewer legitimate businesses stuck in queues for decisions they should not have been waiting on.
Agent First
Ma Xiaohang, Chief Information Officer of MYbank, frames the shift around an “Agent First” principle: information gathering, analysis and execution go to AI first, while employees set standards, review results, make judgment calls at key nodes and step in promptly for exceptional cases. Inside the bank, the principle shows up as eight AI workbenches.
“Whether a bank serves its customers well depends not on willingness, but on capability,” Ma said. In his view, the development of AI agents offers a new answer to problems that have long limited banks: customer information that is scattered and volatile, and professional capabilities spread across departments. The point, as he puts it, is moving inclusive finance from ‘standard answers’ calculated on existing information to answers ‘generated in real time’ based on needs. He believes a bank’s future competitiveness will move further from “talent density” toward “intelligence density” — the amount of professional intelligence an organization has accumulated and can invoke.
The timing is not accidental. The AI Playbook for Financial Services, a white paper published in June 2026 by the World Economic Forum in collaboration with Accenture, pointed out that financial institutions’ AI adoption is moving from local experiments to organization-wide, scaled deployment.
MYbank’s approach puts it on that path — with the agent, not the branch, as the front door.
The Small-Business Test
Whether conversation-first banking becomes the default for small-business finance is a question the next year will answer. The direction of travel, though, is already visible.
A loan that once started with paperwork now starts with a question typed into a chat window — and, in the case the bank put on show, ended with a plan in about 10 minutes.