Breaking: Zafin launches AIOS for governed agentic work500M+ accounts processed dailyNine of the world's top 15 banks use ZafinFounded in Canada in 2002

Company profile / Fintech / Enterprise AI

Zafin Spent 24 Years Escaping the Banking Core. Now It Wants to Govern the AI Agents Moving In

The Canadian fintech made its name by putting a clean control layer over stubborn bank infrastructure. Its new bet is that the same trick - externalize the messy logic, preserve the evidence, keep humans accountable - can tame enterprise AI.

The most revealing object in enterprise banking is not the vault. It is the spreadsheet. Somewhere between a product manager's idea, a relationship manager's negotiated discount and a customer's monthly statement, rates and fees pass through an archaeological dig of workbooks, approval emails, custom code and core systems old enough to rent cars. Zafin built a business by standing in that gap. Its software pulls the rules for products, pricing, offers, deals and billing out of the transaction engine, then gives banks a governed place to change them.

That sounds like plumbing because it is plumbing. It is also where banks decide who qualifies for a fee waiver, how a household earns a better savings rate, whether a commercial client's negotiated terms reach the invoice and how quickly a new product appears on a phone. Zafin's pitch is not that the old core must vanish. The core can keep doing what it does well - accounts, balances, posting - while a separate cloud layer handles the business logic that changes every week.

Founded in 2002 by Al Karim Somji, Anugopal Venugopalan and Dinesh Krishnan, the Canadian company released its product-and-pricing platform in 2005. HDFC Bank went live five months later. Europe followed, then CIBC became the first North American client in 2011. Today Zafin names ING, HSBC, Wells Fargo, PNC, ANZ, Navy Federal Credit Union and Kiwibank among its customers. The company says its software processes more than 500 million accounts each day and is used by nine of the world's top 15 banks.

500M+accounts processed every day
9/15top global banks using Zafin
16+countries with deployments

The move is to make the core boring

A traditional bank core often owns two very different jobs. It keeps the ledger correct, and it stores the commercial rules wrapped around that ledger. The first job rewards stability. The second demands constant change. Combining them makes a simple pricing update feel like open-heart surgery. Zafin separates the jobs. Business users configure a catalog, rates, fees, eligibility rules and offers in one layer. APIs, batch files and event streams send approved decisions to channels and downstream systems.

The difference from a full core replacement is the size of the bet. A bank can modernize a product family or workflow without moving every account at once. The difference from a point pricing tool is breadth: Zafin connects the product definition to the rate, the personalized offer, the negotiated deal, the bill and the audit trail. Zafin IO handles the untidy border between old and new systems. IO Canvas adds a visual builder for data pipelines while generating code that customers can inspect and modify.

Zafin product interface showing banking products and an AI-assisted proposal workflow
The banker's cockpit: less Top Gun, more “please make the fee on the invoice match the fee we promised.”

What it did, what it cost, what resisted

The clearest public implementation numbers come from Zafin's own customer case studies, so they belong in the useful-but-not-independent bucket. A Gulf bank replaced decentralized pricing and manual, Excel-based deal negotiation with centralized pricing for more than 100 products, connected to over 10 applications. Zafin says the bank automated quote-to-cash, supported 35 currencies and produced more than $3 million in annual revenue uplift, with payback inside 12 months. A Canadian bank launched 52 personalized offers and projected $44 million in revenue uplift over five years. Another North American institution cut a product-variant launch from nine months to one day.

The glamorous competitor is another fintech. The daily competitor is a workbook nobody dares to delete.

There is no public price card. Zafin sells negotiated enterprise SaaS and implementation programs, often involving multiple cores, channels and control functions. The honest buying equation is therefore wider than a software subscription: implementation and integration cost on one side; avoided infrastructure, manual work, revenue leakage, errors and delayed launches on the other. At company level, disclosed financing totals about $47.2 million before Nordic Capital bought a majority stake in 2024 on undisclosed terms. Nordic reports that Zafin produced EUR 100 million in revenue in 2024 and employed 680 people.

What failed first was certainty that banks would welcome the cloud. In 2016 Zafin began moving from perpetual on-premise licenses toward SaaS. Somji later recalled that banks were not fully on board with cloud deployments. Customers holding perpetual licenses had to be shown why the new model would cost less overall, even if the invoice looked different. Zafin responded with candid conversations, hired digital-banking operator Chris DeBruin as president and used partners including Microsoft and Accenture to make the transition credible.

The change of mind was economic, not cosmetic. Cloud customers no longer needed to own the infrastructure or keep recruiting people to maintain Zafin's application. Zafin gained recurring revenue and one product base it could improve continuously. By 2019, the company had proof that tier-one banks would adopt SaaS and pay market rates. A $17.2 million round from Vistara Growth, Accenture and Beedie Capital helped build analytics, operations and the global growth engine.

Building, not merely selling

Zafin says roughly a quarter of annual revenue goes back into research and development.

0%~25% R&D100%

The second act: govern the agents

In June 2026, Zafin stepped beyond its familiar category with AIOS, an orchestration platform and control plane for agentic work. The product registers an institution's own agents and approved third-party agents, governs the models and tools they can use, routes work through human decision points and records the result. For each action, AIOS is designed to retain what was requested, what context was supplied, which permissions applied, where a person reviewed the work, what changed and what it cost.

This is a broader market than bank pricing, but the architectural rhyme is obvious. Once again, volatile activity is spreading across systems that were not designed to coordinate it. Once again, regulated buyers need speed without surrendering authority or evidence. And once again, Zafin proposes an external control layer rather than a single all-knowing replacement. The company tested AIOS on its own software delivery, modernization and internal operations before launch, calling itself “Customer Zero.”

The wager is sensible, though not automatic. Microsoft and OpenAI already sit inside Zafin's operating story: the banking platform runs on Azure, Azure AI has been folded into the partnership, and ChatGPT Enterprise was deployed across Zafin's product development and delivery teams in 2025. AIOS is the attempt to turn those experiments into an accountable operating system for work that crosses models, tools and people.

The playbook worth stealing

  1. Find the fast-changing rules trapped inside the customer's slowest system.
  2. Externalize those rules without demanding a big-bang replacement.
  3. Put approvals, audit evidence and exception handling into the workflow.
  4. Prove value on a visible metric: launch time, leakage, deposits or operating cost.
  5. Use partners to cross trust gaps your product cannot cross alone.

Where it works - and where it breaks

Zafin is best suited to a regulated institution with enough product complexity to justify a shared control layer, multiple systems that must coexist and leaders willing to modernize progressively. The economic case gets stronger when pricing changes are frequent, negotiated deals are common, billing leakage is measurable or a core migration risks duplicating product logic across platforms. Banks can use it to launch deposit campaigns, personalize offers, manage fee schedules, negotiate commercial deals, generate bills and preserve a defensible trail of decisions.

It is not magic middleware

The model weakens when source data is unreliable, nobody owns the product rules or teams refuse to redesign approval paths. A separate layer adds another system to govern. If a small institution has simple products and one modern core, the integration may outweigh the benefit. If leaders want a total core replacement in one move, progressive externalization may feel like an extra stop. AIOS faces an additional test: governance has to be embedded in real work, not added as a dashboard after agents have already scattered across the company.

The competitive field includes product modules inside Temenos, Finastra and Oracle suites; pricing specialists such as SunTec and Earnix; and the ever-present in-house stack of rules engines, spreadsheets and custom billing code. Zafin's position between them is unusually specific. It is neither the ledger nor merely the front end. It is the place where a regulated institution decides what should happen, sends that decision into old and new machinery, and keeps the receipt.

Nordic Capital's majority acquisition in 2024 and the handoff from Somji to CEO Charbel Safadi gave Zafin resources and permission to widen the idea. Banking execution remains the proven business. Zafin IO makes the integrations reusable. AIOS asks whether the same discipline can travel beyond products and prices into the work itself. If the answer is yes, Zafin will have escaped the core only to discover that every new technology cycle eventually needs one thing banks understand very well: control.

Keep exploring