Investors are backing AI-native UAE banks, and Islamic finance could gain the most

Capital is flowing into UAE banks designed around AI rather than retrofitted with it — and Islamic finance, where compliance paperwork drives up costs, may have the most to gain. Abu Dhabi-based Mal raised a USD 230 mn seed round to build what it bills as the world’s first AI-native Islamic digital bank. Synapse Analytics raised USD 13 mn for decisioning infrastructure that automates credit, onboarding, fraud, and anti-money laundering (AML) checks.

The difference is architectural. Most UAE banks have spent years adding AI to customer service, fraud detection, and credit assessment without changing how the bank actually runs. Some 44% of UAE financial institutions are already using AI to cut operating costs, according to banking software provider Finastra. This is different: it’s about changing the engine itself, not retrofitting with new features.

“AI-native can be a marketing term until the architecture is examined,” Ahmet Faruk Aysan, associate dean for research at Hamad Bin Khalifa University’s College of Islamic Studies, tells EnterpriseAM. In a truly AI-native bank, Aysan says, data is designed to be machine-readable in real time, while models function as shared services across products and departments. Agentic AI systems — which can plan and execute sequences of actions — become part of the workflow rather than an add-on. People move from processing-heavy roles to a more supervisory capacity, he says.

Why it matters: The economics of banking shift. “Banking starts to look like software: high fixed cost and very low marginal cost,” Aysan says. It raises the upfront investment required to build and govern models but reduces the marginal cost of processing an additional customer, transaction, or document, Aysan adds.

An underwriter who spent 30 minutes gathering information could instead spend five minutes reviewing cases AI has already assembled and scored, says L. Guru Raghavendran, SVP of AI and innovation at core-banking software provider Azentio. The near-term gain is a lower cost per approved application and more capacity, he says, rather than an immediate cut in headcount costs.

Paperwork goes first, credit calls last

Rule-heavy work is first in line for automation, and the savings could be large. AI could cut operating costs across the UAE’s banking and financial system by 30-50%, according to early estimates. KYC checks, AML reviews, document processing, reconciliation, collections, and regulatory reporting are the prime candidates, Aysan says.

Credit decisions are a different story: Dubai-based SME lender Beehive uses its own AI tools to review borrower data and flag anomalies but keeps its credit model in-house and a human on every final decision, CEO Pete Tavener tells us. Unsecured loans approaching USD 300k won’t be left to a black box, he says, while smaller AED 50-100k tickets or data-rich cases can be automated further.

The time savings are real all the same: Beehive spent 11.5 hours assessing a case at the start of 2026 and expects to be at five hours by year-end. “I want to be able to see how the decision is made, and I want to be able to have a full audit trail,” Tavener says.

Islamic banking has the most to gain

Shariah compliance is expensive, and that’s the opening. UAE Islamic banking assets stood at AED 1.4 tn in June, with a target of AED 2.5 tn by 2031. “Shariah-compliant structures carry a premium for documentation and review over conventional products. If AI brings that compliance cost premium down, Islamic products become cheaper to produce relative to conventional ones,” Aysan says.

It could also revive risk-sharing finance. “The part of the model that could change most is the one that has frustrated Islamic economists for forty years,” Aysan says. The industry leans heavily on murabaha, a sale-based structure, rather than profit-and-loss sharing contracts like mudaraba and musharaka. That’s partly because monitoring many small businesses’ cashflows has been too costly. “If monitoring becomes cheap and continuous, risk-sharing finance becomes commercially viable at scale for the first time,” he says.

Scholars aren’t going anywhere: AI can’t decide whether a structure reflects the intent of Islamic law, Aysan says. “Scholars will spend less time on repetitive ex-post review” and more time translating rulings into machine-readable rules and overseeing how systems apply them, which means shariah boards will need members or advisers who understand how the models work.

The central bank wants a human on the hook

The Central Bank of the UAE’s AI guidance note, issued in February, keeps boards and senior management responsible for AI systems and their outcomes. That is a big reason banks are slow to move agents from pilot into production, Raghavendran says. Lenders are comfortable automating tasks like sorting customer complaints, but not final credit decisions. “Accountability cannot be delegated to AI,” he says.

Data is the other brake: Banks remain wary of sending sensitive data to external large language models, even with in-country hosting, Raghavendran says. Azentio’s workaround is to put engineers inside bank premises to train models on the bank’s own data.

Where it can go wrong

Attackers are running agents too. AI-powered cyberattacks have climbed sharply in recent months, with financial services among the most targeted sectors. Synthetic identities and deepfakes can undermine authentication while prompt injection can manipulate systems into taking unauthorized actions, warns Aysan. “When both the attacker and the bank run agents, an error propagates in seconds rather than days, and a bank that has not defined limits, identity, and a kill switch for each agent will discover the problem from its losses,” he says.

Plus: old data carries old biases. In the Gulf, nationality, employer, visa type, and residential location can all correlate with repayment, but they can also act as proxies for traits that shouldn’t decide access to credit, Aysan says. Used carefully, transaction data could open credit to thin-file borrowers, migrant workers, and small businesses.

GO DEEPER- Our AI + Innovation desk looked at whether AI can make lending fairer across Egypt and the Gulf, including how UAE credit reports now factor in buy-now-pay-later repayments and why Emirates NBD is running ChatGPT on in-country infrastructure. Read it here.

Plenty of projects won’t make it: Gartner expects over 40% of agentic AI projects to be canceled by the end of 2027 over costs, unclear value, or weak risk controls. The fix, Aysan argues, is to build governance into the bank’s architecture through “audit trails, model inventories, independent validation and human escalation mechanisms,” rather than adding it after deployment.

The UAE offers AI-native banks a favorable test market: a sizable Islamic banking sector, centralized shariah governance, an established financial services ecosystem, and specific regulatory guidance on AI, according to Aysan. Its diverse population also presents a demanding test for models used in onboarding and credit assessment. Incumbent UAE banks that adapt proactively could ultimately be better positioned than challengers, Aysan believes.

Success, however, will depend on more than AI launches or partnerships. Cost-to-income ratios, independently validated AI decisions, customer complaints, and human-review reversals will offer better measures of progress, alongside shariah non-compliance incidents and growth in risk-sharing products for Islamic banks, according to Aysan.

For some, the change is inevitable: Banking software is set to become AI-first, reshaping not just how employees interact with systems but how banks process information and make decisions. “The underlying workflow itself will change with AI, and it’s not [taking] five years. Trust me, two years, two and a half years,” says Raghavendran.