Banks around the world are letting AI agents open accounts, approve loans, and flag fraud without a human touch in the process. One bank cut its onboarding time from three days to eleven minutes. ZAND is one of the banks building that future right here in the UAE. Joining me is Michael Chan, Chief Executive Officer of ZAND Bank. Michael, welcome to Wall Street to Mena.
Great to be here. Thank you for having me.
ZAND is rolling out generative AI across banking operations including onboarding. In plain terms, what does an AI agent do that takes a human days to finish?
The generative AI we are using actually is across our bank — not only onboarding. But for today's topic, when a client is onboarded by the bank, the process involves collecting documentation, capturing data, and verifying information against different registers. If you are registered in the DIFC, we need to check that. We also need to check whether you are a sanctioned person. In the past, every single one of these steps was done by a human being, one at a time. Now everything is done by different agents in parallel. You can imagine a troop of agents all working on the same process at the same time.
Some banks have cut onboarding from three days down to eleven minutes using AI agents. Has ZAND hit that number?
I think we are doing way better. For simple structures, we are hitting under three minutes. It depends very much on the client — a UAE-based company will be faster than one with a complex ownership structure. But for straightforward cases, under three minutes.
If an AI agent approves something it should not — a loan, a transaction — who is actually accountable? The bank, the AI vendor, or the AI itself?
Definitely the bank. The agent is doing the job for you, but that does not mean the bank can remove its liability and responsibility. At the end of the day it is the bank. It is me.
80% of financial leaders say AI agents are now a board-level topic. Is the board actually comfortable with how much authority AI has inside the bank?
We started the AI journey two and a half years ago and there was some uncertainty at that time. Right now we have board-level discussions about how to better use AI. We have confidence at the board level — not only because they have more knowledge about AI, but also because we took the governance seriously. We built everything ourselves rather than relying on vendors. We contain all the data within the UAE on our own GPU. Your data does not leave the UAE.
AI agents are increasingly catching fraud and money laundering patterns humans would miss. Does that make you trust AI more or worry about it?
Both. You should trust it more because it can catch what a human cannot. But you should also worry because bad actors are using AI agents to create more sophisticated fraud. It is a competition between a good agent and a bad agent. The bank can probably automate 90% of detection. The remaining 10% — the high-stakes decisions — stays with the human.
Does heavy AI automation risk making banking feel cold and impersonal?
The human being should ultimately deal with unhappy customers. That is what bankers should focus on. Many processes can be automated by AI agents, including customer support for routine issues. But for clients who have a real problem or dispute — that is where the human touch is critical. AI agents will get smarter and smarter. But the human being will always be needed for the unhappy customer.
Two years from now, how much of ZAND's operation do you expect will run without a human ever touching the process?
From a staffing perspective, probably 40% fewer people. But from a process perspective, the people who remain will be dealing with more complicated, more structured decisions. I do not think the human disappears from banking. Most banks will be using AI by then — but people will be more sophisticated, more focused on the decisions that genuinely require human judgement.
Thank you so much for being with us today, Michael.
