Exploring the Upside, and Risks, of Artificial Intelligence in Alts

The rise of artificial intelligence (AI), and its potential use in the global alternatives’ ecosystem, is reshaping views about processes ranging from onboarding to investment management and cybersecurity.

The extent to which AI has shifted industry discussions about technology in the year since ChatGPT launched was evident during the discussion, “Rewiring the World: How Technology is Reshaping FinTech,” at our recent Global Alternatives Summit in Miami. Adil Rehman, our Global Head of Payments & Liquidity, was joined by panelists Mark O’Donnell, Financial Services Industry Lead for AWS, Steven Shwartz, Owner of AI Perspectives, and Evangelos Skianis, our Chief Technology Officer, to exchange views on AI’s strengths and weaknesses, as well as potential risks.

The panelists touched on the distinction between traditional software, generative AI (e.g., large language models) and predictive AI (e.g., machine learning), and their uses in processing large volumes data for functions including asset management, reporting, and onboarding new clients and employees.

“What you have to do as an organization is understand what each of the three types of automation can and can’t do,” Steven said. “Now you start making use cases of what might be possible, and only then do you go and look at the cost of building each capability.”

Mark noted that AI can automate data stored in individual documents and reuse that data across processes, including employee onboarding, Anti-Money Laundering and Know Your Customer applications. In addition, funds are exploring ways to use AI as an investment management process tool.

“You have text extraction, summary, search, and the ability to generate as well as to cull tools,” he said. “These are the building blocks and then you ask the question, ‘What can I build in my business using this?’” Typically, client questions on AI focus on “where do I start? Do I partner up with a vendor? And what about my data,” Evangelos said. He suggested an incremental approach—beginning with mature process where results can be measured and compared to established metrics, and then finding partners to assist with more complex processes to protect data and manage costs.

“Don’t invest in building anything that’s generalized,” he said. “Find ways to partner with your cloud providers, use their managed models, don’t invest in building your own model. Understand your use cases, then you can start fine tuning your models.”

While panelists noted the significant potential for AI technology, they added that existing tools posed challenges—generating inaccurate outputs and providing possible entry points for cybercriminals—that require vigilance to protect proprietary data and digital channels.

“You’ve got to test for bias, you’ve got to put guardrails around the tools to prevent them from generating toxic output,” Steven said.

The potential for inaccurate or inappropriate use of AI means that “we should never forget we still need people in the loop to look at outputs,” Evangelos said. “And you also need people on top of the loop to monitor, adjust and look at metrics, and if necessary, turn systems off on the public-facing side.”

With real cybersecurity risks to consider, funds and their service providers must “keep in the arms race with the bad guys,” Mark said. “This is an accelerator—more digital, more data strategy, better cybersecurity posture.”