Airwallex has recently named its in-app AI assistant Kai, following months of its expanded functionality inside the Airwallex WebApp. Yet, renaming does not eliminate structural inefficiencies this support method has, according to public reviews.
The company’s Chief Product Officer Shannon Scott announced the new AI assistant’s name in a company blog post on June 17. The assistant launched without a public name in February 2026 and has since added onboarding automation, recurring-payment detection, spend approval workflows, transfer initiation, and card management.
Airwallex says Kai now handles 88% of user conversations without escalating to a human agent. The company reports that Kai has helped more than 14,000 customers complete onboarding with a setup process customized to their business information.
What Kai Does
Kai operates inside the Airwallex WebApp and draws on a customer’s account data, transaction history, and the content of the page they are viewing. According to Airwallex, current capabilities include:
- Generating customized onboarding steps based on company information
- Detecting recurring payment recipients and prefilling transfer details
- Building multi-step spend approval workflows from natural-language instructions
- Initiating and tracking transfers
- Creating new cards and adjusting spending limits on existing ones
- Answering questions about payment status, deposits, fees, and invoice line items
Airwallex says future versions of Kai will add broader business analytics, guided payment acceptance setup with method recommendations, and employee onboarding support, including role assignment and permission configuration.
The Escalation Gap
The tricky point is that Kai does not connect users to a live human agent. Airwallex’s Help Centre states that when a case requires human review, Kai collects details and raises a request, and a member of the Customer Support team follows up by email and other existing channels.
This design has drawn criticism in customer reviews predating the Kai rebrand. A summary of Trustpilot reviews reveals that the delayed response, AI assistant creating multiple tickets, alongside difficulty reaching human support, were main points of dissatisfaction for some users. Individual reviews describe related problems: one user reported using the AI escalation path for an urgent request and receiving no reply for over a week. Another described a 15-day wait on a held-funds complaint, with follow-up limited to repeated “checking with our partner bank” messages and no live channel to press for an update.
One more complaint on the platform describes a structural side effect of the ticket-based handoff: every time the AI assistant is used on the web portal, it automatically creates a separate ticket for each inquiry. The customer said this is confusing both end users and the support team, and is resulting in redundant tickets being sent to them.
Airwallex’s own documentation frames the email handoff as a deliberate guardrail rather than a gap: the company states the AI Assistant is designed to complement, not replace, existing channels, and that for complex or sensitive issues, it creates a ticket so the Customer Support team can continue support by email and other existing channels.
How This Experience Compares Across The Industry
The friction Airwallex users describe is not unique to one platform. A May 2026 study from payment infrastructure company DECTA, based on a survey of 1,506 UK adults and a two-year analysis of nearly 160,000 app store reviews across ten European banking apps, found that chatbots resolve customer issues at a low rate industry-wide.
Among survey respondents who had a payment blocked and sought resolution, 76.3% interacted with a chatbot at some point in the process. Only 11.4% said the chatbot resolved their issue. Half had to escalate to a human agent, and 14.9% said they never received a real resolution. The same study found that 65.2% of all respondents said they trust a human agent most to resolve money problems, against 5.4% who said the same of a chatbot or AI assistant.
The DECTA study also tracked app store sentiment over time: negative reviews citing a chatbot or AI assistant as the cause of an unresolved problem rose 55.49% year over year, comparing April 2024–April 2025 against April 2025–April 2026. Identity verification, declined transactions, login and access issues, and account suspension accounted for 57% of the underlying problems in those reviews.
Airwallex has not published a comparable resolution-time figure for the 12% of Kai conversations that are escalated, leaving open how the email-based follow-up performs against the urgency of the cases it handles.
