The traditional data technology build-vs-buy decision facing asset managers has been blurred by artificial intelligence. Not only has the technology made the “build” option faster to achieve and more affordable, it has also reframed the notion of where exactly the building happens.
At the same time, the risks associated with AI’s inclusion in the equation have made “buy” a security as well as an operational choice.
AI and associated automating agents have been successfully deployed in a range of asset management use cases that might have been included in earlier third-party provisions, including data reconciliation as well as document processing and automation. While these are narrow applications, the potential is there for AI to be incorporated across firms, breaking down the technology disconnect between front-, middle-, and back-offices, said Boden Anderson, a director at consultancy Alpha FMC’s North American asset and wealth management practice.
“New tools are giving people in the front office or middle office, functions that didn’t traditionally build without support from technology teams, the opportunity to experiment,” Anderson told Data Management Insight. “Previously there was this translation issue between ‘here’s what I want and I pass it over to my tech team and I get something back and it isn’t quite right’. But with vibe coding and prototyping that is shifting and shrinking the gap. Now there is this blurring of the lines.”
Slow Adoption
Adoption of AI in asset management and the broader buy-side has been slower than in other parts of the financial industry. That’s down to a combination of factors including research licensing restrictions, cybersecurity worries and regulatory burden, which has given pause to wide-scale upheaval of asset managers’ tech stacks.
The drag caused by the compliance risks posed by AI transformation may be alleviated with a recent proposal by the UK regulator – the Financial Conduct Authority (FCA) – to streamline asset managers’ reporting obligations. The shake-up could cut their reporting burden by 75%, according to technology consultancy Capco, and could ease asset managers’ concerns over AI investments.
AI is presenting opportunities that make overcoming the change hurdles worthwhile, Anderson said. For instance, it can reduce the manual preparation burden around reporting, though regulatory filings should retain a human in the loop, he added.
Advanced operational use cases are already being considered. Among them is the use of AI to surface trade and investment ideas from a firm’s own history, including prior decisions, internal research, meeting notes and stated portfolio objectives as suggestions for a portfolio manager to weigh.
Rather than committing large sums of capital and multi-year timelines to a full tech-stack refit, firms can move from pilot to production faster and pull more non-technical staff into development.
“That’s partially why we see pretty narrow use cases or experiments right now, because most firms don’t have a full unified view of their data,” Anderson said. “There’s a hesitancy to fully tie up to [platform providers] because then they have full pricing power over firms in this rapidly changing world.”
New Definitions
While the bar to entry for in-house development has been substantially lowered by AI, the technology has also redefined what buy-in means.
With the technology reliant on access to expensive large language models (LLMs) and agents, these third-party technologies have to be incorporated into even the most extensive own-build programmes. That has rendered the option “a disguise for a different sort of buy” because even proprietary AI-powered builds piggyback the larger providers, Anderson said.
Many of the traditional factors within the build-vs-buy debate remain in the age of AI. Larger companies, those with competent IT teams and firms whose operations are tech-reliant, such as those with quant teams, are still more able to afford or develop in-house capabilities. They are also more experienced and adept at managing the culture shift that such transformations entail.
“Those firms are more comfortable with uncertainty, the development lifecycle in general and always have already worked in an iterative way,” Anderson said.
Companies that have already invested in front-to-back solutions will still be more comfortable with bought capabilities, as will smaller companies that can’t afford an extensive IT team.
Even so, the emerging overlap between the two strategies means that vendors and platforms must have a better understanding of clients’ AI needs and know the trajectory of the technology well enough to tailor a plan to them. That also goes for asset managers, who must assess the potential evolution of the models on which they are likely to rely.
Pace of Change
Another factor that is playing on the decision-making processes of asset managers’ data chiefs is the opportunity cost of buying now. Buy-in decisions have always needed to balance satisfying the immediate needs of the firm with the likely additional benefits that could be accrued by waiting to see how third-party technology evolves.
Because AI has accelerated that rate of development, asset managers considering third-party provisions now face shorter decision deadlines.
This consideration is making it difficult to assess how the future of AI investment among asset managers will develop. At the moment, even those leveraging third parties are using them mainly as a data workflow backbone to operations that also leverage “mini-builds” to plug gaps, Anderson said.
Guiding the direction of travel will be data quality, he said. New solutions can only be scaled if the user has the data to do so.
“It’s not just, is the data correct, but it’s also, is it tagged appropriately? Do I know who owns it? And do I know how it relates to other data domains that we have?” he said. “Without doing some of that foundational work, you’re hamstrung in what you can actually accomplish.”
Subscribe to our newsletter
