AI may be the fourth major technology revolution Richard Forss, CTO at Exante, has witnessed during his three decades in financial technology, but unlike the PC, internet and smartphone, this revolution isn’t giving financial institutions years to adapt.
Forss is no stranger to technology revolutions, with AI being the fourth major industry shakeup. During his three decades developing technology solutions across the financial sector, Forss has seen first-hand how financial institutions have transformed operations with the advent of PCs, the internet and smartphones. At first glance, AI might seem like a similar transition. But unlike the previous three revolutions, it follows different rules, ones that could leave financial institutions behind if they fail to adapt quickly. This major difference is time.
Forss explained, “The first three you could see coming and you had time to adjust. The PC arrived, and you had years to work out what a spreadsheet meant for your business. The internet, the same. Smartphones, the same. Each one was enormous, but the clock ran at a human pace. You could be a fast follower and still be fine. AI is different in one specific and important way. It is the speed, not the intelligence.”
Workflows that would normally take big chunks of a day can suddenly be automated down to just minutes. This can send shockwaves across the business, giving staff more time back in their day to handle other tasks or spend more time focusing on important jobs that cannot be automated. However, Forss noted that this automation can work both ways. Illicit players can use AI to find weaknesses quicker than ever. A recent high-profile example of this was Anthropic’s Claude AI, which escaped a test to hack three systems. Forss said, “The previous revolutions changed what we could make. This one changes the tempo of everything, including the threats.”
He added, “I’ve watched a lot of “this changes everything” moments over thirty years and learned to discount most of them. This one I don’t discount. Not because the technology is magic, but because it removes the breathing room the last three gave us.”
While AI might differ from the previous three major revolutions, there are still plenty of lessons firms can leverage to ensure they are able to capitalise on the technology as efficiently as possible.
The big one, according to Forss, is ensuring they measure the cost of production. Enterprises are judging AI by their output volume, such as the tickets processed, reports generated and lines of code written, but they are not measuring the downstream cost of acting on this output.
For instance, Meta reportedly has an internal leaderboard to rank user token usage. This is a similar method used by other companies to track and encourage employees to use AI. However, tokens come at a cost. As AI adoption grows and token costs rise, a better measure would be employees making the most out of the fewest number of tokens.
One stack that is often missed by budgets, Forss noted, is the engineering hours spent fixing hallucinated code, the rework when AI-led analysis sends a strategy the wrong way and the compliance exposure when someone signs off on output that was never proofread. “AI can add real financial cost while creating the illusion of savings.”
Ensuring the correct ROI metrics is not the only lesson firms are forgetting when racing to adopt AI. He said, “The unglamorous fundamentals still matter, and they still go undone. Get rid of legacy that’s beyond support. Fix the bugs you already know about. Understand what you actually depend on, because you cannot defend what you have not mapped. Every revolution tempts people to skip the foundations and rush to the shiny part on top. It never ends well, and AI hasn’t repealed that rule. If anything, it punishes it faster.”
Reshaping wealth management
When firms talk about the implementation of AI within workflows, it is often described as something that will reshape the industry. While that might be the case, there is often a connotation with that suggesting the role of the human will be replaced. That is not the case, the best implementation of AI does not involve removing human workers.
Forss said, “The most valuable thing AI does right now is remove the manual overhead that slows good judgment down, not the judgment itself.” For instance, if a portfolio manager who previously spent hours replicating the same trade across dozens of accounts can hand that task to an AI tool, they can instead focus on work that requires a human’s judgement. That is where AI becomes genuinely transformative. He added, “That’s genuinely transformative, but it’s transformative because it frees up the human, not because it removes them.”
Unfortunately, not all firms see the value of AI being tied to the human. He explained, “We’ve deliberately gone the opposite way to a lot of the industry. Plenty of firms have used technology to depersonalise: automate the client away, push everyone to a chatbot. We’ve kept a named person on the other end of the line.
“So, my honest view is that AI reshapes the industry most where it takes friction out of the plumbing, and reshapes it least, and should reshape it least, where trust and accountability live. The firms that survive this will be the ones that know the difference.”
Regardless of how it is implemented, AI has the ability to transform operations across most departments within wealth managers. Whether it is within their investment research, operations or client servicing, the technology can be implemented to bring transformation. The question is, where will its impact be the greatest?
While many might assume it’s within investment research, Forss is not convinced. While that is the “visible, glamorous bit”, middle and back-office operations stand to gain the most. “The unglamorous machinery behind every investment decision. That’s where the gains are largest and least discussed.”
The reason for this is simply because research has always had smart people and good tools. “The bottleneck was never thinking it was the manual execution sitting underneath the thinking.” Each new client has its own allocation rules, he explained. Every market move triggers a rebalance that is required across multiple portfolios, with most firms still trading with spreadsheets and manual trade entry through a combination of tools that were not designed to work together.
“The margin for error widens with every account you add. That’s the layer AI and automation transform most directly over five years.”
Supporting firms through the revolution
Exante, which provides access to global stocks, ETFs, bonds, futures, options, commodities, and currencies through its platform, is helping the wealth management sector in its journey through the next revolution.
As part of this effort, the WealthTech recently launched the Gecko Fund, a €1m grant programme that will support critical open-source software projects across trading and financial data systems. Direct funding will be offered to developers of APIs, single-maintainer libraries and core components of global trading infrastructure.
Speaking on its launch, Forss noted there is an imbalance in the support of critical open-source software, with nearly 70% of a modern software stack relying on open-source components but most of the infrastructure is maintained by small, underfunded teams. This creates fragility in technology the sector relies on daily. As AI adoption increases, the illicit players are using the technology to accelerate the discovery and exploitation of vulnerabilities.
Exante understands the problem first-hand. Forss noted, “Our job is to concentrate on what clients need, not to rewrite the libraries that calculate options from scratch when QuantLib exists, is free, and is the global standard. We use open source so we don’t have to rebuild the world every morning. Everyone does.” The problem is identifying what is at the bottom of the stack.
After an audit of its own infrastructure, it identified a serialisation framework called Kryo, which was holding up high-performance data processing maintained by two people on different continents as a passion project and with no outside funding. “That’s not unusual. That’s entirely typical. The systemic risk isn’t in the big projects with foundations and legal teams behind them. It’s in the small, load-bearing libraries patched late at night by one person after their actual job has finished.”
“This used to be a slow problem you could nod gravely about at a conference and forget. AI changed the clock.
“That’s why this moved from a theoretical conversation to a board-level one, and why we treat it as a business investment in resilience rather than charity. “
Wealth management firms can invest in their own cybersecurity, governance and operational resilience, but that work can be unravelled by a weak link in the underlying infrastructure.
Forss offered an example where a vendor, which operated as a blood-test provider, was several steps up the supply chain but still attacked from a vulnerability down the line and led to a patient’s death despite the vendor never being directly attacked.
“You are only as resilient as the components you did not write and may not even know you depend on. If a widely used library stops being maintained, or a serious vulnerability goes unfixed because there’s no one funded to fix it, the damage doesn’t stop at one firm. Understanding those dependencies, and helping keep them healthy, is becoming part of responsible technology management, not a nice-to-have.”
In addition to the fund, Exante also recently launched its Allocator solution, a portfolio management infrastructure that helps asset managers scale operations by deploying and rebalancing investment strategies across multiple accounts. Demand for separately managed accounts has exploded over the past year, with that trend expected to continue over the coming years.
Portfolio management continues to become more personalised, and every client wants their own mandate with their own restrictions and preferences. This is not sustainable without support. Forss stated, “You can have the best strategy in the world, but if deploying it across 40 accounts means 40 manual executions, you’ve built a bottleneck and a source of error into your own business.” Allocator was built to solve this gap by allowing a manager to build a master framework and push allocation changes and rebalancing across every linked account.
“I want to be precise about the intent. We are not automating judgment out of the investment process. We are removing the manual overhead that slows judgment down, so the human attention goes where it’s actually needed. That distinction is the whole point.”
Final thoughts
Coming back to the AI technology revolution, having experienced the previous three major technology shakeups Forss has seen where businesses and investors typically go wrong when trying to maximise the impact of new technology. While we are only in the early stages of AI’s adoption, there are already signs of firms making mistakes.
Forss noted that during these revolutions people often focus solely on the technology and not what it depends on. “Every revolution, people stare at the shiny layer on top, the new device, the new model, and make confident predictions about it, while completely missing the second-order effects and the foundations underneath.
“Technology in finance is a bit like architecture. A good façade turns heads, but it’s the engineering beneath that decides whether the thing stands up. People consistently bet on the façade.” In the world of AI, this mentality sees firms rush to implement the solution before they have their data infrastructure in shape. Fragmented and siloed systems limit the potential of AI, becoming a costly mistake.
This also relates to the second error firms make, not giving it enough time to reshape operations. “People reliably overestimate what a technology does in two years and underestimate what it does in ten. They pile in at the peak of the noise, get disappointed when it doesn’t reshape everything by Christmas, and then aren’t paying attention when it quietly does reshape everything five years later.”
He added, “If I had to compress it to one sentence: the mistake is thinking the advantage comes from buying more of the new thing, when it actually comes from knowing what you depend on. The resilience problems we’re all going to face over the next five years (concentration in a handful of cloud providers, the software supply chain, AI agents acting inside live workflows) none of them are solved by buying more of anything. They’re solved by understanding your own dependencies.
“That’s the unglamorous discipline that separates the firms who cope from the ones who get caught out, and it’s the one nobody finds exciting enough to do until it’s too late.”
