Parth Desai and the trouble with the wrong address

A bank can send money across an ocean and then spend its afternoon arguing with an address. Someone has put a town in the wrong field. Someone else has abbreviated a name. The payment is perfectly understandable to a person, which is an unfortunate qualification when the next reader is a machine. Parth Desai has spent much of his working life in that gap between what people mean and what banking software can accept.

His career began with artificial intelligence research and developed into a business making payment instructions intelligible to computers. He founded the company that became Pelican, led it as chief executive, and built products around the awkward business of understanding financial language. Now, as founder and chairman of ioNova AI, he is working on the address problem again. The technology has changed considerably. Human beings continue to put things in the wrong box.

There is a pleasing stubbornness to this professional trajectory. While the public conversation about AI has expanded to almost everything, Desai keeps returning to an exacting question: can a system understand enough about a transaction to handle it correctly? A useful answer has to survive a bank’s operations department. Eloquence earns very little credit there.

A research scientist meets the payment message

Desai studied at the Indian Institute of Technology Bombay and earned a master’s degree in artificial intelligence at Georgia Tech. Before starting Pelican, he worked as a research scientist at Cognitive Systems in New Haven under Roger Schank, the AI researcher associated with Yale. Natural language processing became a central part of his expertise: teaching computers to work with the language people actually use.

Banking offered a wonderfully unforgiving place to apply it. The early business involved reading and converting payment telexes into Swift messages. A telex instruction and a standardised financial message might concern the same money, but getting from one to the other required interpretation. Software needed to recognise meaning amid the variations, then put the information into a format another system could process.

That founding problem gives the later career its shape. Desai was dealing with AI as an industrial tool, attached to a particular task and judged by what happened downstream. A payment message is a small object with substantial consequences. If its contents are misread, a tidy-looking output is little consolation to the people waiting for the money.

His subsequent education included senior management programmes at Harvard Business School and INSEAD. The combination is apt: research alone would not build a bank technology supplier. He had to organise a company around the work, retain expertise and keep translating a technical proposition into something customers could buy and operate.

When ACE acquired a bird’s name

In September 2016, ACE Software Solutions announced its rebrand as Pelican, adopting the name of its existing platform. The change also brought an advisory board of payments and financial technology executives. Desai presented the moment as a reaffirmation of the company’s long investment in AI, with a product approach joining natural language processing, knowledge-based systems and machine learning.

The three ingredients matter. Language processing helps interpret an instruction. A knowledge-based system supplies explicit understanding of a domain. Machine learning can detect patterns in examples. Desai’s business rested on putting these methods to work together inside payment processing and compliance. The result had to fit the procedures, formats and obligations of institutions already moving money.

In a 2018 discussion of banking’s digital transformation, he argued that AI would sit at the centre of banks’ response to the real-time economy. He also talked about “customer cognizance”: understanding customers well enough to respond to their needs. His interest extended from repairing instructions behind the scenes to how an institution could use its information to serve people.

The scale eventually became substantial. During his leadership, Pelican reported handling more than a billion transactions worth over $5 trillion, with customers in more than 55 countries. Their relevance is operational: the ideas had been put to work across different institutions and jurisdictions, repeatedly, outside a demonstration.

Five hundred million possible customers, one connection

Open banking gave him another translation problem. Banks could expose services through application programming interfaces, or APIs, but an open door is only so convenient if every doorway requires a different key. PelicanPay’s proposition was an interoperability layer: a way for businesses and financial technology providers to connect across differing banking interfaces.

The launch described a potential European market of up to 500 million customers through a single API. Potential is the operative word; it was an addressable market, rather than a claim that PelicanPay had acquired that many customers. Desai saw regulatory change as an opportunity to make access simpler for organisations building new payment services.

By 2020 he was discussing practical applications including invoices, salaries and collections, with PelicanPay apps already in use in the Netherlands. That detail brings the ambition down to a recognisable size. An employer needs to pay people. A business needs to collect an invoice. The grand future of finance must eventually do something useful before Friday.

He described the internal attitude in equally direct terms: “Change is constant”. His account of leadership centred on developing the team’s AI expertise and asking how it could improve customers’ efficiency. It is a working philosophy expressed through client problems, rather than a collection of personal commandments. In April 2020, he joined Pelican sales director Abhinav Desai for a discussion of open banking across the UK and Europe.

“Change is constant”

Parth Desai, discussing Pelican’s approach to innovation, 2020

The expensive nuisance of a false alarm

Financial crime screening provided another test of understanding. A bank must investigate potentially prohibited activity, but an alert can also be a false positive. Every unnecessary investigation consumes attention that could be used elsewhere. Desai’s work addressed the difficult requirement of improving efficiency while respecting the controls that make the process necessary.

In November 2021, BNY Mellon selected Pelican technology for sanctions screening and financial compliance. The product was designed to complement existing screening systems and reduce the work associated with false positives. This was a practical route into a complex institution: improve an existing process, with its obligations intact, rather than demand that everything be replaced.

Recognition followed the same professional territory. Desai was included in The Financial Technology Report’s Top 25 RegTech Executives of 2021. The list placed him among people building technology for regulatory and compliance work. In his public response, he called recognition for a field so important to him “very special”. After years spent making machines more discriminating, he could allow himself a small celebration.

A sale, a separation, another company

In July 2024, KOAT Capital announced the acquisition of Pelican and its subsidiaries. The announcement outlined a strengthened management team and said Desai would continue driving the company’s technology. It was a transition from the founder-led business he had run for decades into a different ownership structure.

The later relationship became contentious. In a September 2026 corporate update, Pelican said litigation involving Desai and related parties remained ongoing, and that a subsidiary had filed a counterclaim in New Jersey. The company disputed the claims against it. At that announcement, no court had determined the merits of the parties’ claims or defences. The dispute forms part of the company’s recent history; it establishes no finding against either side.

Desai’s current public role is founder and chairman of ioNova AI. The business concentrates on making entity and address data usable in regulated financial processes. He also offers board advice and fractional chief AI officer services. His preferred description of the approach is knowledge-first AI: establish the relevant domain knowledge and governance as part of building a system that people can understand and examine.

The deadline moved. The address remained.

During 2026, Desai wrote about the need to improve address data for ISO 20022, the financial messaging standard. His argument was that changing the format of a message while postponing work on the data inside it repeats an old mistake. A better envelope does little for a letter addressed to the wrong place.

The calendar needs care. Swift had planned to remove unstructured postal addresses from relevant payment messages in November 2026. On 27 August, it agreed to extend the migration timeline after requests from the industry. Its September update left structured-address timing under consultation, with an update promised by December. The separate June 2027 release date did not establish a new address deadline.

That change makes Desai’s underlying subject more interesting than a countdown. The question is whether institutions will improve their information because it helps them operate, even when the immediate scheduling pressure eases. ioNova’s products address that work by separating address elements, verifying them and keeping reasons for corrections. There is a direct line back to the payment telex: human language arrives, and a financial system needs to know precisely what it means.

Across the research job, the company he built and the business he chairs now, Desai has kept choosing variations of the same demanding task. He wants AI to understand the particulars of banking well enough to be useful inside it. An address is an unglamorous place to pursue that ambition. It is also where the money sometimes gets stuck.