Giving structure to unstructured data has become indispensable to private market investors, who must deal with what must feel, to the much of rest of the digitised financial world, like relics from antiquity – PDFs, spreadsheets, emails and even paper documents.
But the question that hangs over many solutions is what next? What happens to that data once it’s been extracted and put into text or tabular form?
It’s a question with which Bryan Dougherty, President, Product and Technology at Arcesium, grappled in creating a domain-focused addition to the American data management technology provider’s Aquata self-service data platform. Ostensibly created to enable clients to scale their artificial intelligence applications, the update was built with private markets front of mind. It offers general and limited partners, as well as the other emerging players in the sector, the capability to make something of their data, Dougherty told Data Management Insight.“Why are you collecting this unstructured data and what are you going to do with it next?” he asks. “How do you know if it’s fit for purpose? What’s fit for purpose for accounting might not be what you need to track records, or the like. We set about building something that could consume this unstructured data.”
Need for Integration
For Dougherty, it was critical that the Aquata update gave clients what comes next: integration into the company’s suite of data management, analytics and investment tools.
He argues that the welter of other data extraction products on the market are often not fit for purpose because, once pulled, they leave the data dangling – or as he puts it “rubbish left at the doorstep”. Even those built to integrate with separate downstream applications suffer.
“We see a lot of value in the integrated story,” he says. “Integration across workflows, whether it’s operations or analytics, is very critical. And it’s very expensive when you use point solutions, you absorb a lot of what I call impedance mismatch between the various tools.”
The update lets institutional asset managers, hedge funds, private markets funds and capital markets firms scale their AI strategies and deploy agentic workflows. Unstructured data extraction comes courtesy of a suite of large language models (LLMs) and Optical Character Recognition (OCR) technology.While OCR is a relatively old technology, Dougherty said Arcesium found through experimentation, supported by research, that they give better outputs when deployed in combination.
The update also features a Model Context Protocol (MCP) server that connects any enterprise AI tool to their governed data foundation in the Aquata platform.
Private Credit
While Aquata has been built to serve broader financial markets, it’s also been tooled to be domain aware for private credit, a corner of the alternatives sector that has been difficult to digitise because of its many moving parts and opacity.
It’s also one in which it’s seeing huge demand for automation, seeking to break away from the “spreadsheet-and-luck” method of tracking investments.
“We’re very focused on that sweet spot of complexity emerging in private markets,” Dougherty says. “That’s really where we’ve seen the most compelling work so far.”
Private credit is filling a vacuum left by traditional lenders following the regional banking crisis. The market has swelled to roughly $2 trillion, according to Bloomberg data, and a number of companies have begun providing data services to those investors, including Fitch Solutions and LemonEdge.
For those markets, investment accounting is a serious challenge, especially as funds become more complex and larger.
In recent applications, the agentic tool extracted loan lifecycle events—including drawdowns, interest repricing, and fees—from loan notices across more than 15 different counterparties. What previously required hours of manual validation by operations teams has been reduced to minutes of exception-based review.
Dougherty punctuates his explanation of the update’s capabilities with excited observations about the application of AI in his personal life – being able to use his phone to scan and identify individual plants is a particular favourite. The point being that the potential of AI is being experienced now. But that isn’t always apparent in financial servicing.
He sees AI as the “dessert” making the unappetising prospects of basic data management, centralising records and making them auditable more palatable.
Agentic AI
Unsurprisingly, then, Arcesium’s future plans are based in agentic AI, generally seen as the next step in the technology’s evolution. Equally unsurprising, these plans appear to be informed by another of Dougherty’s personal-world experience of AI.
While trying to digitise some decades-old documents he chanced upon a model that enabled him to turn them into searchable PDFs, a process that took half an hour but would probably have been prohibitively difficult to achieve manually.
Professing to have been “gobsmacked” by the experience, he likened it to the orchestrating nature of agentic AI in providing the glue between workflows and applications.
“We have an unstructured data story; you can pull that in together and then you can write a pipeline, but what would be even better is to have an agent where you could just describe what you want that pipeline to do, and it’ll help you author that,” he says. “That’s going to be incredibly powerful.”
Subscribe to our newsletter
