Labs

Labs are rethinking banking, as AI remains king, but human insight directs banking improvements.

So, the robot banker remains a long way off. But S&P Global estimates that up to 59% of financial institutions worldwide were actively using artificial intelligence in 2025. Beyond simply relying on technology for research (“Summarize new anti-money laundering mandates for me”), financial institutions have begun operationalizing AI processes.

This is a significant advancement. Instead of using AI like an intelligent chatbot, banks now direct systems to perform complex, multi-step tasks—saving untold human hours while both speeding up and improving operations.

How does this newer type of AI (called “agentic AI”) work?

Consider loan processing as an example. Someone applies for a loan. AI agents retrieve credit reports, verify income, calculate debt-to-income ratios, apply underwriting rules, approve or reject applications (or forward them to a human underwriter for approval), and generate documentation. In compliance monitoring, agents can read regulatory texts, map new mandates to internal policies and processes, identify where the financial institution (FI) falls short, generate remediation tasks, and track progress.

For proof of the increased operationalization of AI, look at some of the innovations germinated in the world’s best fintech labs, incubators, and accelerators.

At inovabra, a lab hosted by Banco Bradesco, innovators have developed an AI product that can generate initial drafts of legal pleadings. The Bank of Georgia’s AI Research Lab has launched Software Developer: Powered by Code2Doc.

This software can write other software. And Garanti BBVA Partners has nurtured Skymod, an AI-orchestration platform that enables financial institutions to securely delegate operational workflows to intelligent AI agents.

Knowing When AI Isn’t The Answer

Then there’s TD Lab. TD Lab is now experimenting with Physical AI, or AI-embedded machines (think robots, drones, and smart devices) capable of interacting with the physical world.

“Physical AI is about convergence,” said Chris Halabecki, senior manager and lab leader. “It’s about combining AI with objects that can sense or maneuver through the real world. As a lab team, we’re exploring how we can use physical AI to integrate more intelligence into everyday scenarios to better serve our colleagues and clients today and in the future.”

The lab has already developed proprietary software for a quadruped (robotic dog) device. Using LiDAR (Light Detection and Ranging), which is a sensing technology that uses pulsed laser light to measure distances, and AI together, the quadruped can detect and learn about objects in the surrounding area, then follow commands linked to those objects. Halabecki provided examples such as “Walk to the white couch” and “Go find Evan.”

Future use cases for these technologies may include robots that can count, sort, and verify cash.

One day, robotic relationship managers may recognize when a customer walks into a branch and guide them in making investment decisions. In the field, physical AI may be able to conduct home appraisals and complete other tasks.

With the much-ballyhooed capabilities of artificial intelligence, it’s a little surprising to hear Kadry Boutaina, chief of innovation for the digital transformation lab of Morocco’s Attijariwafa bank, say, “Sometimes, the answer is not AI.”

That doesn’t mean the lab, called Wenov, isn’t driving technological advancements. It works with external startups to offer “more and more digital services for our customers — both retail and business.” Boutaina notes that Attijariwafa faces significant competition in this field, from both established banks and newcomers — notably neobanks entering the Moroccan and broader West African markets.

But providing digital services does not always entail a wholesale AI revolution.

When Banks Let Employees Innovate

The Moroccan Ministry of Economy and Finance recently formalized laws governing crowdfunding in the country. The first regulated platform of this kind is being provided by Kiwi Collecte, a fintech company. Under Moroccan law, Kiwi Collecte may not directly hold or move funds. It must partner with a licensed Moroccan bank for those tasks. A partnership with Attijariwafa empowers the bank to hold and safeguard funds, process payments, and disburse money to beneficiaries.

Fraud prevention is ever important. Sandbox CAIXA has found an old-school way to fight it. Sandbox CAIXA is the innovation lab of Caixa Econômica Federal, a major state-owned bank in Brazil. Lucas Zaccaro, Sandbox CAIXA manager, said that in his country, technologically unsophisticated people are often victimized by scammers. When the bank is closed, thieves stand near ATMs. They then offer to help patrons who are unsure how to use the machines. These criminals “help” by tricking users into revealing their PINs, then stealing their cards.

A “really great idea” from a rank-and-file Caixa employee led the bank to broadcast recorded messages at 10 of these ATMs, warning patrons about the scam. Theft at those banks has stopped.

Zaccaro says that this fraud-prevention idea was submitted through an established process designed to encourage rank-and-file employees to submit innovative ideas. Employees use Microsoft Copilot to refine their concepts and submit them to Sandbox CAIXA for review and potential testing. The lab will now assess the feasibility of rolling out its scam warning across the ATM network—potentially using cameras to detect when people are at the ATM and triggering automated messages.

At the Banking and Financial Institutions Association of Colombia (Asobancaria), the focus is less on rapidly advancing technologies and more on meeting existing societal needs. One development from the Asobancaria Social Innovation Lab is a reference framework for identifying, classifying, and reporting on the banking sector’s social portfolios. This proposal—the second of its kind in Latin America after Guatemala’s Social Taxonomy—was developed through multiple sessions of analysis, technical feedback, and sector-wide validation with member institutions. To create this framework, Asobancaria worked closely with the Global Green Growth Institute (GGGI), which develops social-portfolio standards aligned with the United Nations Sustainable Development Goals.

Andrea Guzmán, GGGI’s sustainable finance officer, said the problem with sustainability reporting among Asobancaria member banks was a lack of alignment on standards, with each bank setting its own measures of success for its social portfolios.

The framework addresses issues such as financial inclusion, social infrastructure, affordable housing, and services for small and medium-size businesses. A single framework to which all member banks agree “improves transparency,” Guzmán says. “It supports better decision-making by investors and helps mobilize more resources for our social sector. It’s a framework we can base bonds on. It’s a framework that helps banks avoid accusations of greenwashing.”

Consider the framework for sustainable and affordable housing. Guzmán notes that in rural Columbia, many houses lack access to water and may have only dirt floors. Therefore, a bank could claim success in affordable housing if it funded units with wood floors, fully plumbed and connected to the electrical grid. But what if those apartments are so far from public transportation that no one can get to work or school? Guzmán said that under the framework, banks agree that any affordable housing projects they fund would have access to the nation’s external infrastructure and social services.

Here’s a closer look at some of the world’s best fintech labs and the innovations they’re nurturing.