One of the biggest problems with the current AI investment supercycle is that the breathless pace of change obscures the core reasons why companies adopt new technology – improving efficiency and boosting return on investment.
“It's hard to remember where we were two years ago. The world has changed so much,” said Siobhán Noble, chief data and AI officer at Carne Group.
So, while not that long ago, GenAI’s ability to quickly summarize a meeting or construct a presentation seemed revolutionary, today Noble is demanding agentic AI deliver far more tangible results.
Noble joined Carne Group, the world's largest independent third-party fund management company, in 2022 after a succession of business and tech roles in the finance sector. She is also a member of the AI Office of Ireland, the agency responsible for implementing the EU AI Act in Ireland.
That means that she has always looked at AI, like any other technology, through a business transformation lens. And looked at through that lens, she said in an interview with The Stack at the International AI Summit in Dublin, the tangible benefits of gen AI and agentic AI for enterprises are not always clear.
Show your work
A few years ago, people could boast that they’d become more personally efficient through the use of copilots and similar tools by summarizing information or helping produce a presentation, Noble said.
Carne Group has rolled out Microsoft Copilot licenses across the company, backed with training and workshops to help employees work out how to use the technology to solve their own problems. But it can be hard to position these “benefits” in the context of enterprise transformation.
“You're telling me you're saving two hours a week,” she said. “What are you doing with the two hours? Are you spending more time with the customer? Are you dealing with the regulator? Are you onboarding new business?”
The question becomes even more pressing when it comes to agentic AI. Agents have the potential to be a “gamechanger,” she said, but “if you look out to the market, there's not huge compelling evidence that many companies have figured out how to leverage AI, Gen AI, agentic AI in a way that actually leads to a positive ROI.”
At the same time the risks of hallucination or bias with gen AI or agentic AI are a very real concern for companies in the global asset industry, which are already highly regulated and can expect “increased scrutiny all the time,” Noble said. Deterministic technologies were often a better starting point, she said.
In Noble's view, it’s critical to look at business problems and outcomes — centering how to transform the business as a whole — to avoid embarking on experiments that end up in the “proof of concept graveyard…it’s a real thing and it’s led by technology.”
“If agents can actually do the work, certain parts of the job that an individual is doing today, it changes how the jobs are done, it changes people's roles. There's an organization design and a people piece.”
So, when it comes to using technology to solve important business problems, “there's a decision tree around when you use traditional tech and traditional engineering, where you use process redesign. There's an order of priority. AI can really, really help and turbocharge some of that journey, but it doesn't start with AI,” she said.
Zero to fund
Carne Group has previously applied agent technology to reasoning and unstructured data ingestion tasks, in the shape of AI-Ristotle, a family of agents built on its core data platform.
Now, according to Noble, “the absolute ambition and vision is that we’re going to speed up fund launches in our industry.”
Launching a fund is a complex fragmented workflow spanning multiple teams,including internal and external legal teams, due diligence and compliance teams, and inevitably external regulators. This is all happening against the background of a massive shift in the investment market away from public markets.
Traditionally, much of the launch preparation relies on email, spreadsheets, document repositories, workflow tools, multiple systems and manual processes, Noble said. “The challenge is not necessarily a lack of technology, but the fragmentation of information and the number of hand-offs across the lifecycle.”
That’s why it typically takes 10 to 12 months to get a fund to market. The program is targeting a 30 percent reduction in time to market for funds. It will be delivered in phases, with the first launch scheduled for December.
Automating that workflow gives the opportunity to rethink the jobs that make up the process, she said. The new process brings together data, automation, document intelligence, AI, agents and workflow orchestration, Noble explained, as well as existing systems such as its Curator platform.
“The important distinction is that we're not simply adding AI and agents to existing tasks," she said. "We're redesigning the underlying processes based on jobs to be done and outcomes, with appropriate human oversight and controls built in.”
And people have to be at the center of the process, “What we've learned is you can't design those new processes without the subject matter experts, who actually understand how this actually works,” she said.
“Every day, one of our customers, our asset managers, can't deploy a fund is a day that they can't deploy their capital and they can't generate revenue.”
But while the firm is automating large parts of its work, the goal is not to drive down headcount.
While Ireland’s graduates are taking a hit when it comes to job opportunities overall, key tech sectors are still recruiting. Noble said the firm had experienced 20 percent compound growth for the last few years and it was heavily investing “at the associate level."
“Our view is very much for the foreseeable future, we expect that we will be onboarding the brightest and smartest graduates, and we will be heavily investing in them.”