Womenleading the way

Our project sought to change that, and as we kept expanding, I’ve had the pleasure of meeting many remarkable people through it. Founders, executives, researchers, policymakers—women who came from different backgrounds and each brought some unique perspective shaped by their personal journeys.

Over the past year, Women Leading the Way has grown into a global community. We’ve launched a podcast, partnered with leading international events and, most importantly, brought together hundreds of remarkable women who continue to inspire one another through their stories and achievements. When we first set out on this road, I honestly did not imagine that things would get so busy.

This edition is somewhat different from what came before. We wanted to make it bigger, more accessible and, frankly, give ourselves more freedom to be creative with it. And one of the biggest changes is that we’ve decided to turn it into a digital publication.

I have to admit that this is not an entirely natural choice for me, because I love printed books. I love holding them, turning the pages with my own fingers, even the smell of a new book.

I have spent decades building up my own library, and in more recent years I’ve enjoyed building one for my daughter as well.

Our team can take copies to industry events and mail them to our contributors, but that’s just putting them into the hands of people we interact with directly. We cannot physically bring the magazine to every person who might find something valuable in it.

By going digital, Women Leading the Way removes that restriction and can reach readers across borders and communities in ways that simply weren’t possible before.

And for a project whose entire purpose is to give more people access to great stories, that feels like an opportunity we should take.

It’s not just about changing up the format, either. With this edition, we wanted to bring in more voices to explore the subjects that are shaping our industry right now.

The goal of this project has always been to promote female expertise, but we never wanted to put that expertise in a box and tell women which subjects they were supposed to talk about. Quite the opposite, in fact—we wanted to open that particular door wider.

This has nothing to do with “women’s topics”; these are the topics that matter broadly today, and women have every right to be at the centre of those conversations.

They can contribute much more than some people still tend to assume.

Many of the women we’ve had the privilege of meeting and interviewing repeat the same idea: they don’t want to be recognised simply because they are women. They want to be recognised because they are professionals who are great at what they do, and I honestly couldn’t agree more.

Women Leading the Way has always been about creating a more complete conversation around finance, and while women remain at the heart of that mission, there is more to it than gender alone. Meaningful conversations happen when different experiences come together, challenge one another, and ultimately contribute to a shared understanding for everyone. That’s why, in this edition, you will also find voices from men whose ideas we believed important to include.

In the past, our project has already received a lot of support from male industry representatives who came to us recommending their female colleagues and thus validating our mission. Men who believed in the importance of creating more opportunities for women and who actively helped this initiative grow. Including their voices to stand side by side in this edition feels like a natural progression for us and a reflection of that shared commitment.

The financial industry changes very quickly, especially today as technology keeps running forward faster than many can keep up. I believe that sharing expertise in such an environment has become more valuable than ever before, and Women Leading the Way intends to continue playing its part in making that happen.

ORIGINAL RESEARCH
BY NIGAR (NIYA) IBRAHIMOVA

Head of Content at Drofa Comms & Editorial Director at Women Leading the Way

Ask an AI assistant who you should follow to understand monetary policy and it will give you an answer in seconds. Central bankers, economists, academics. Each name arrives with an affiliation and a neat explanation of why this person matters.

It feels remarkably definitive for a question that has no definitive answer.

There is no official register of the world’s best financial experts. Nobody has agreed how much weight should be given to original research, investment performance, institutional seniority, media visibility or decades‑old influence. Yet an AI system has to turn those signals into a recommendation—and increasingly, people use that recommendation as a starting point for their own research.

That made us curious about what happens one step earlier. Before AI can decide who counts as an expert, who had the opportunity to become visible as one?

For much of modern financial history, that opportunity was distributed unevenly. Women had fewer routes into the institutions producing financial knowledge and fewer opportunities to publish, be cited, manage capital, speak publicly and occupy the positions that decide whose expertise reaches everyone else.

The imbalance begins surprisingly early. A Federal Reserve Bank of New York study of NBER Summer Institute programmes found that women represented 20.6% of authors on scheduled papers in 2013–2016. In finance, their share fell to 14.4%; in macroeconomics, to 16.3%.¹

Getting the work published does not necessarily close the gap. Economist Marlène Koffi found that female‑authored papers in leading economics journals received fewer citations from top‑tier journals and male scholars.² And by the time specialist knowledge reaches the wider public, women remain less visible: the 2025 Global Media Monitoring Project found that they accounted for only 23% of experts quoted or cited in legacy news worldwide.³

Even the people deciding whose voices reach the public come from an uneven leadership structure. Reuters Institute research across 240 major news brands in 12 markets found that women held 27% of the top‑editor positions identified in 2025, despite representing an average of 40% of journalists in those markets.⁴

These figures describe different stages of the same journey. Knowledge has to be produced, attributed, published, cited, noticed, and preserved before an AI system can ever retrieve it.

So we wanted to see what happens at the other end of that process.

The closer our questions came to the direct allocation of capital, the fewer women AI recommended.

Monetary policy produced the highest female representation at 23.8%. Fintech fell to 16.7%, digital assets to 13.6%. Investment management reached just 7.9%.

Across all three regional investment‑management outputs, only three women appeared: Kristina Hooper, Wei Li and Johanna Kyrklund.

The category also revealed something about what AI understood an “expert” to be.

Fintech produced founders, consultants, authors, and industry commentators. Monetary policy brought central bankers, academics, and former government officials. Digital assets leaned towards researchers and executives connected to established financial or investment firms. Investment management reached for a much more traditional canon: famous investors, fund founders, authors, and people whose authority had accumulated over long careers.

John Bogle and David Swensen both appeared despite being deceased. Their inclusion is perfectly understandable if the question is about historical influence. It is more complicated if someone is looking for people to follow today.

Yet the system presented historical importance, current responsibility, investment performance, publication, and public visibility inside the same apparently coherent hierarchy.

The list looked precise. The definition of expertise underneath it was much less so.

Brazil changed the prompt. It barely changed the expert class.

We also wanted to know whether asking from different regional contexts would change whose expertise the system considered relevant.

It did, but very unevenly.

Only two members of this twelve‑person cross‑regional core were women: Theodora Lau and Claudia Sahm. This may be one of the most consequential patterns in the experiment.

Ask once, and a familiar name looks like a recommendation. Ask again from another country and see the same person, and they begin to look unavoidable. Visibility can compound.

Someone who is already widely published, quoted, invited to conferences and included in expert lists leaves an enormous number of digital signals connecting their name to a field. Those signals make them easier for a system to identify. Recommendation then brings new readers, followers, citations, invitations, and mentions, creating still more evidence linking that person with expertise.

Recommendations were dominated by people associated with institutions including Fidelity, Goldman Sachs, Standard Chartered, Coinbase, Galaxy, CoinShares, VanEck, Bitwise, and 21Shares. Independent crypto‑native researchers were much less prominent.

There is a logic to this.

Digital assets are still a relatively young and contested field. Expertise can be difficult to compare directly. A senior position at a globally recognised bank, asset manager or investment firm provides a much cleaner signal. Somebody else has already selected the person, given them a title and published a biography establishing their credentials.

Institutional prestige can therefore become a shortcut for a much harder question:

How good is this person's actual work?

Fintech produced a different version of the same mechanism. Here, authority clustered around authors, newsletter publishers, consultants, conference speakers and people repeatedly included in “thought‑leader” rankings.

These people generate signals machines can read exceptionally well. Their names appear beside the same subject again and again. They have stable biographies, searchable publications and repeated descriptions of their expertise.

Meanwhile, some of the most consequential forms of financial judgement are much harder to see.

Throughout our Women Leading the Way interviews and podcasts, we noticed that professional expertise often came well before public recognition. Long before these women were regularly invited to speak, quoted for their opinions or trusted with greater responsibility, almost all of them had to put in extra effort to get noticed, recognised, and promoted.

For this article, we went back through six of our previous interviews and podcast conversations and looked at their stories through one specific question:

What did these women do before the market started treating them as authorities?

The answers were different. Some highlighted that recognition grew from going unusually deep into the technical side of their work, while others built an extensive track record, made themselves more visible, spoke publicly, or found people willing to open the right doors.

Taken together, their stories suggest that doing the job well was only one part of the equation. Recognition usually requires another layer of effort.

Claurelle Rakipovic, CEO of Pipe, was the one who had to dive deep into the technical side of what she was doing. Throughout finance, lending, and product, she says it mattered to understand not only her immediate area of responsibility, but how the whole system worked and where different parts connected.

In simple terms, before people could trust her decisions, she first had to scrutinise the subject to defend them if someone questioned their viability. Without that scrutiny, it could have been much harder for others to recognise her authority.

For Catherine Jenkin, Co-Founder and Head of Editorial at Kernel Media and a former Cointelegraph editor, constantly proving her value was never something she aimed to do.

Instead of putting in extra hours to prove herself, she focused on producing work strong enough to do that job for her.

Mahsa Doorfard, Sales, and Marketing Manager at coinIX, told us that alongside her work in marketing and investments, she built a public presence through educational content, community work, events, webinars, conferences, and panel appearances. She also says that external representation is a key part of her role at coinIX.

Saori Honorato, Editor‑in‑Chief at Portal do Bitcoin, says visibility also meant writing and publishing as much as she could. She says she often had to work harder to receive the same recognition as male colleagues, while the results of her work eventually brought new opportunities and opened doors for her career. Her advice today reflects that route directly:

We also believe visibility works best as a multiplier of expertise. It gives strong work a wider audience, creates more chances to be invited into industry discussions and helps people associate a name with a particular field. Ignoring visibility altogether, on the other hand, can leave even substantial expertise confined to one company or professional circle.

But if we ask another question, for example, “Would you like the algorithm to find the bank with the best deposit rate for you in five seconds?”, the same people who said no would most likely nod.

The difference between these two questions is the so‑called AI trust gap, which is, in fact, the topic of the last few years for regulators, banks, and wealth managers around the world.

For instance, in January 2026, the British FCA published its Mills Review, a report on how advanced AI, including agent systems capable of operating autonomously, will change retail financial markets.

In this work, the regulator discusses what happens when there is no chatbot with prompts between the consumer and the bank, but an agent that compares products, makes decisions on its own, and executes them on its own.

At the same time, the Bank of England is conducting rounds of consultations on the risks of such systems, and large English retail banks such as NatWest, Lloyds, and Starling are already testing customer pilots under FCA supervision.

Given how fast AI develops, it seems clear that it is ready to run even faster. The question is how much faster we, people, will let the technology into our financial lives and where it will hit the trust ceiling.

So the first one is research, and through it, AI has made a grand entrance. By research, we mean, for example, asking an assistant how an index fund differs from an actively managed one.

Before AI, it took hours and days to understand those concepts on your own using the Internet. However, AI, beyond simply answering the customer’s question, can also parse a complex insurance contract, and it’s almost like making a Web search, but way smarter.

Kristy Kim

The CEO and Founder of TomoCredit

“For example, if you’re using AI to ask basic (and fairly generic) financial questions, then a clear explanation with no personal details should be enough and also be pretty low risk.”

That is why recent surveys show 78% of adults already use AI tools for research, and 55% have asked AI theoretical questions to help make financial decisions. Not much, but it is still up from 10% just a year earlier.²

Among Generation Z, this share reaches 77%, and among millennials it’s up to 72%. That said, asking the algorithm for advice has become no scarier than googling your symptoms before visiting a doctor.³

If the last door opens easily, the recommendation door is where hesitation starts. A global survey recorded that not even half, namely, 49% of consumers in the world have used AI to support savings and investment decisions over the past six months.⁴

More importantly, they were “getting a financial product recommendation from AI” noticeably less often—28% of full‑time employees, 23% of students and only 7% of pensioners.⁵

However, such a decrease in trust should not surprise us. People treat recommendations this way because they are no longer facts but judgments. So consumers feel the need to ask why it should work out in their favour.

When it comes to portfolio construction, the gap between “try” and “trust” becomes a chasm. According to estimates, only 44% of respondents say they are comfortable using self‑service AI tools for investment management.⁶

The other half of respondents said they would choose the joint AI‑plus‑human format, and that makes sense. Making a portfolio is a series of interrelated decisions, the error in any of which accumulates over the years.

Mostly, people in these conditions prefer to share responsibility for the final result with someone else. If AI makes a mistake, will it be accountable?

Of course, no. So it is psychologically inconvenient to share it with a system that has no advisory license and no personal stake.

Aelin Golsarry

Founder of AAG Consulting and a former CIO and CFO

“Having AI in financial planning means understanding where AI is being used, what decisions it is allowed to make, what controls are in place, when a human needs to intervene, and what happens when something goes wrong.”

When we use such a door metaphor, it’s good to ask ourselves whether these doors are equally open to everyone. Unfortunately, they are not, and the difference also lies in gender, not just risk tolerance.

Let’s start with some estimates. The next two decades will bring the largest transfer of capital between generations in history—83 trillion dollars will go from the baby boomers to the heirs. And a significant part of this amount will end up in the hands of women.⁷

Although this clearly shows that gender equality is leveling off, this influx of capital also opens a gap in the financial confidence women experience. Many women admit they are not confident in their ability to manage an inheritance or an unexpected financial windfall. Only 49% of female investors opened their own investment account, compared to two‑thirds of men.⁸

But as practice shows, the confidence gap does not reflect the results at all. In fact, on average, women’s portfolios often outperform men’s because of a more disciplined, less reactive strategy.⁹ For example, during market turbulence, many women remain calm and move along the chosen course, while men make decisions in a hurry.

In addition, women turned out to be more open to technology, and they demonstrated a higher intention to use financial AI tools than men. But women tend to be more careful about the technology, viewing it through a rational and utility lens.

“I'd want to know what the system does when it's uncertain. Good systems say so. Weak ones smooth it over,” adds Ruchi Bhatia.

We have witnessed the same case with robo‑advisors a decade ago. For women, the crucial factors when choosing the technology were trust and the tool’s expected usefulness. Although they had anxiety, it was not the main deterrent. This, of course, opens a new opportunity for the wealth management industry and accelerates the adoption of AI.

But there is also a downside. Algorithms are trained on historical data, and they systematically reflect the under‑servicing of female clients.¹⁰ So it risks transferring the old bias of human advisors into a new, much less noticeable form of algorithmic recommendation. In other words, AI can either close the confidence gap or cement it forever.

“Currently, most AI financial assistants are geared toward people who have already amassed some wealth or a portfolio. We are interested in the rest of everyone else—the average consumer who is struggling to build up an emergency savings or a retirement fund,” finishes Ms. Kim

Even though AI is not yet at the level to open all the locked doors, this may change very soon. So the question “will you trust AI to manage your money” has not one answer, but four—one for each door. And the most interesting thing in the coming years will happen on the third and fourth. Today it can be locked, but tomorrow it may open, certainly not without a struggle for trust, which the industry has yet to win.

It is almost impossible to find an advisor who has not commented upon “The Great Transfer” coming, and while it is indeed an impressive figure, it almost tells nothing about what exactly is going to happen.

Wealth management has spent decades building personal relationships with a particular client, but soon they will need to work with their clients’ spouses or children. These heirs may have very different expectations and a different approach to managing their money. And in a surprisingly high number of cases, may even already have an adviser of their own, leaving their benefactors’ advisor without a job.

This is why advisers should not wait until the transfer is about to happen to think about retention. They need to start preparing now, understand who their future clients are and make sure their strategy works for them too.

Managing wealth isn't easy, even when you have an army of managers to advise. The main challenge is that, in many cases, people who, in the best case, have an opinion on where to invest and what exactly to buy still have to make the final decision, and this sudden responsibility can be stunning.

This problem is structural and sits at the heart of modern wealth management firms. 97% of them still segment clients primarily by wealth band², and at first sight it is difficult to find something more appropriate. But this division only represents the properties of a portfolio itself instead of a person or even household behind it. It cannot distinguish between a couple who share financial authority and one who delegates it. It cannot register that the individual who will control the assets in eight years currently holds a modest balance, and when this person becomes the one who signs documents, they are rarely prepared for this.

No wonder that 83% of spouses reported difficulty taking sole control of household wealth, and one in four did not know where all of their partner’s assets were held.³ They feel paralysed and, in many cases, leave everything untouched, and while this may sound like good news for an initial wealth manager who may suggest their services, those clients are very likely to leave soon.

To keep them, advisors should not bombard them with allocation ideas, and they certainly should not stay quiet like nothing happened. Their first goal should be to prepare a full map of assets and explain to clients what they own and what it is needed for. Very few firms offer such services, which may be the most important one, actually.

Wealth managers are rarely prepared for situations where a client freezes and becomes unable to make a decision; likewise, they are often unprepared for the heirs’ impending desire to change everything. The push for change frequently comes from the capital owners’ children, among whom only 20% intend to keep their parents’ advisors.⁴ The easiest way to explain this trend would be to point to the new generation’s affinity for digital technology: they prefer mobile apps to human advisors and digital assets to traditional ones, and simply lack the patience to sit through a multi‑hour review of an investment portfolio.

However, this explanation is very far from being true and, worse, it leads advisers in the wrong direction. The main challenge is not that heirs may want to invest all of their parents’ money into some niche project they found online but that, according to Cerulli Associates, more than half of them already have an advisor they’ve worked with for years. Only 14% of heirs said they didn’t want to work with an advisor at all.⁵

In most cases, discussions around this cultural break between the current generation and the next ones miss the point of who exactly will inherit the money. Heirs can typically be between 40 and 60, suggesting that they have been earning for more than two decades already.

Their investment culture rarely differs from their benefactors, and they definitely won’t leave their family’s current advisor just because they prefer digital technologies. In many cases, the heirs’ decision to change managers would come from the fact that they did not get proper attention from them earlier and have already found another wealth management specialist.

This fact completely changed how the Great Transfer problem should be approached. It moves the whole problem earlier, and if wealth will pass in the next two decades, advisers should start building relationships with future heirs right now.

The Great Wealth Transfer usually comes with another generational shift happening on the adviser side as well. Advisers are ageing alongside their clients, and many will soon need to pass those relationships to younger colleagues.

Think of an adviser who started working with a 45-year-old business owner in the 1990s. Decades later, the client may be preparing to pass their wealth to their children, while the adviser is preparing for retirement. In fact, almost 40% of US advisers are expected to retire in the coming years, and many are approaching that point without a clear succession plan.⁷

For some families, these two changes may happen within just a few years of each other. Wealth passes to a spouse or adult children, and then the adviser who has known the family for decades retires. Suddenly, the new owner of the money is working with a new adviser. Neither has the history or understanding that existed before.

This is why advisers need to think about succession long before they retire. Passing a client to a younger colleague takes more than an introduction and a few meetings. The retiring adviser knows who makes the decisions in the family, how they approach risk, what has happened before and often who needs to be involved in a conversation. That knowledge needs to be passed on too.

While it looks like a challenge, it also offers advantages. A 38-year-old heir may ultimately have more in common with an adviser of a similar age than with someone who has played golf with their father since 1998. If the transition starts early enough, younger advisers have time to build their own relationships with the next generation rather than inheriting a list of unfamiliar clients when a colleague retires.

The fact that wealth changes owners has very little to do with how exactly all those assets will be used. Ownership can shift completely while the underlying system will reproduce itself with different names on the accounts.

That sets a considerably higher bar than the sector has set for itself, because a firm that keeps every dollar and leaves every portfolio exactly as it was won’t win anything. In fact, so many advisory companies are not willing to change anything that the largest ownership change in history may have zero impact on finance.

New owners will likely have broader investment preferences and longer horizons, and wealth managers must adapt. Building relationships with future heirs is only half of the work, as the other half is to find a way to make this transfer an inexhaustible revenue source for the years to come.

Within seconds, it can return a list of names with short explanations of what each person is famous for and why they matter for the industry. The user does not need to search through newspaper archives, check conference programmes or ask colleagues for recommendations, as one question can replace several stages of research.

But where does that shortlist of names come from?

It originates in decades of repeated public recognition. Newspapers decide whom they quote, conference organisers choose who appears on stage, while universities and research institutes attach institutional credibility to experts through titles, research, and affiliation.

Those choices accumulate, so even one executive who is regularly quoted by major publications, media platforms and industry venues can leave behind dozens of articles connecting their name with a particular subject. Someone with equally strong expertise but a much thinner public footprint gives a discovery system far less material to work with.

Women are one particularly visible example of this imbalance. The 2025 Global Media Monitoring Project found that women accounted for only 26% of news subjects and sources globally,¹ while their share among experts quoted in legacy media was just 23%.² Reuters Institute also found that women held only 27% of top editor roles across its 2025 sample of major news brands.³

So, the public record from which professional authority is built does not represent every expert equally. And now, as the route to discovery changes and traditional media pass part of that role to AI systems, those existing differences in visibility can matter in a new way.

Maurício Magaldi

Founder/Host of the BlockDrops Podcast and Venture Builder at IOG

“Any AI solution is only as good as the data it’s been trained on. If the discovery was already biased before AI, there’s little chance that the bias went away with AI.”

Yet, AI does not necessarily inherit the old gatekeeping model in exactly the same form. Some experts see the technology itself as potentially broadening the pool of information from which experts are discovered.

Stephen Sargeant

CEO at Airdropd

“I’m not as concerned, because I feel it evens out the playing field more than if they were just pulling from the most popular sources. I am concerned that those that will be creating the AI systems and putting guardrails on them will predominantly be male.”

When AI Is Asked Who Matters

In 2026, we already have evidence that women can be far less visible when AI systems are asked to identify prominent figures in a particular field.

A 2026 study published in AI and Ethics found a sharp difference between how GPT-4, Claude, and Llama-3 handled factual and subjective questions about notable people. Gender disparities were relatively limited when there was a verifiable answer, but once the models were asked to name the most accomplished, notable or important person in a field, women appeared as the primary answer in only 5% of GPT-4 responses, 4% of Llama responses and 12% of Claude responses. The wording mattered too: asking for the “important” person produced exclusively male primary answers across the subjects and models tested.⁴

A similar problem is explored in our “The Missing Half of the Record” article, where we observe how AI can reproduce existing visibility gaps and influence whose expertise gets recognised in the first place.

The important part here is the mechanism. Once a system is asked to make a subjective judgement about prominence or authority, the existing public record can start influencing who gets surfaced first.

For experts who have received less public attention in the past, that creates a compounding problem. Fewer recommendations may mean fewer interviews, fewer speaking opportunities, fewer citations and fewer new references connecting a person’s name with a particular field.

Women are especially exposed to this effect because their existing share of the public professional record is already smaller. Over time, that can leave future systems with even less material connecting their names to particular fields of expertise.

To say for sure what will happen to professional authority, say, ten years from now, is difficult, as it is still too early to predict which side of AI‑led discovery will prove stronger.

One outcome is fairly easy to imagine, where AI systems keep returning the people with the strongest existing footprint, those names receive more attention, and over time it gets harder for others to replace them. So, in this case, professional authority concentrates around people who were already visible when the technology arrived.

The opposite outcome sounds compelling as well. AI gets better at identifying expertise through the work itself, rather than relying mainly on fame or repeated media exposure, which allows someone with strong research or technical contributions to reach an audience without first spending years entering the traditional circles that once controlled professional recognition.

We think both futures are equally plausible.

Paul Brody

Founder & CEO, Nightfall Networks

“I am very optimistic about the ability for AI to recognize talent and authoritative opinions. While the pace of improvement has left me optimistic, I have not seen rigorous proof that the problem is solved.”

What separates them may have less to do with whether AI is “good” or “bad” for representation and more to do with what these systems are built to recognise. If relevance, attribution, and actual subject expertise carry enough weight, AI can widen the field. Imagine the existing prominence continues to dominate, it may simply exacerbate an old hierarchy.

The Record We Leave Behind

So, the question we started with—who are the leading experts in digital assets?—may look trivial when typed into a chatbot.

It is not.

The answer reflects years of choices about whose work was published, whose name was attached to it, who received recognition and who remained harder to find. And for female experts, that makes visibility today part of a much longer process.

An article, a conference appearance or a piece of research may seem temporary when it first comes out, even so, every named contribution adds another connection between a person and their expertise. After a while, those connections form the material that future systems may use to decide who deserves to be found.

That’s why, striving to make female expertise visible today, which is Women Leading the Way’s goal, means to ensure that ten years from now, when someone asks a machine who matters in finance, fintech or Web3, there is enough evidence for women’s names to appear in the answer.

Today, though, that assumption no longer holds true, and that is on account of artificial intelligence and the growing role it plays. A joint survey by the Bank of England and the Financial Conduct Authority previously discovered that only 34% of financial firms have a “complete understanding” of the AI technologies they use. Meanwhile, nearly half of them admit to having only “partial understanding.”¹

Yet despite that, AI is actively moving into decisions that affect highly important elements of financial operations, such as fraud detection, insurance, credit assessment and investment. This is a dangerous contradiction, since it means financial institutions are growing comfortable with relying on systems the workings of which they openly acknowledge they do not fully understand.

That said, modern financial markets have spent ages relying on technologies that few people could really explain in detail—quantitative risk models and high‑frequency trading algorithms aren’t exactly the most easily understood mechanisms, either.

So it doesn’t automatically mean that AI is unsafe and shouldn’t be relied on. But we do have to acknowledge that it introduces uncertainty about why particular decisions are being made, and that’s a very different conversation. It means that for this technology to be deemed “safe,” explainability has to be at the core of operations.

The Bank for International Settlements previously reported that many explainability techniques can themselves be inaccurate, unspecific, and outright misleading. An explanation produced after the fact may look convincing and still fail to reflect how the model actually arrived at its decision.²

This is a problem, because the financial industry often assumes that having an explanation is in itself evidence that a system is understood. And here we can clearly see that it isn’t.

Many explainable AI methods (SHAP, LIME, etc.) don’t actually open up the model and reveal its internal reasoning; they just try to approximate why the model produced a specific output by analysing its behaviour. It’s essentially generating a plausible story about what happened.

And yes, those stories can be used as guidelines for debugging and governance decisions, or even customer communication, but they are not, by any means, the same as achieving complete transparency. Even more so when you realise that different tools can explain the same AI output differently. The obvious question becomes: “Which explanation should be trusted?”

In other words, having an explanation does not equal having an understanding of what actually goes on inside the model.

Admittedly, in some ways, it’s not that different from how human beings operate. Imagine, for example, asking a veteran portfolio manager why they bought a particular stock. They could give you a general rundown and explain the main factors behind the decision, but it’s unlikely they’d cover every market signal they looked at and every subconscious judgment call that was involved. Instinct and experience are hard to quantify that way.

And then we have to consider that AI models operate on a much larger scale. Their internal reasoning involves millions of parameters interacting simultaneously—compressing all of that into a neat explanation inevitably means losing information.

So the industry is faced with a subtle but important distinction: if AI cannot be made entirely explainable, then the goal should be making it understandable enough for organisations to deploy it responsibly.

At the beginning of 2026, the Bank of England led discussions that covered how traditional model validation could become increasingly impractical as generative and agentic AI grow more sophisticated. Instead, greater focus is being placed on continuous testing, monitoring, governance, and outcome‑based controls.³

To put it in simpler terms, the underlying philosophy behind explainability itself is changing. The question of “Can we explain every decision?” is being replaced by “Can we demonstrate that the system behaves safely, and that it does so consistently, within acceptable boundaries?”

In the long run, this could prove to be a more realistic standard. Trust often comes from governance, rather than perfect comprehension.

Alina Sysoeva

Head of PR & Crisis Communications Lead at Drofa Comms

That instinct is understandable. It is also where companies can lose valuable time.

A crisis creates two problems at once. The company needs to establish what actually happened, while everyone outside the company is already trying to explain it. Those processes move at very different speeds. An investigation may take days or months. A narrative can form within minutes.

That gap is where crisis communications really happens.

Customers do not expect a company to have every answer immediately. They do expect signs that somebody is in control: acknowledgement that something has happened, clear instructions where necessary and a reliable indication of when they will hear more. The first hours are therefore less about producing the perfect explanation and more about establishing the company as the primary source of information about its own crisis.

The first 30 minutes should therefore be used to establish an internal source of truth.

At this stage, I would want one live document containing three things: what we know, what we do not know and what we cannot yet comment on. Alongside it, the team needs to establish who has authority to publish, who needs to approve information and which channels will carry the response.

This sounds basic until a real crisis happens. Then someone discovers that the person with access to the corporate X account is asleep in another time zone, legal is reviewing a paragraph that has already become outdated, the CEO is answering investors separately and customer support is working from information that communications has not seen.

That final commitment matters more than it may seem. During a crisis, uncertainty is often more damaging than a lack of detail. “We will update you again at 14:00 UTC” gives customers something concrete to work with. Once that promise has been made, however, missing it creates another problem. The company has turned a crisis it may not have caused into a commitment it has personally broken.

The other discipline is resisting the pressure to name a cause too early. A preliminary explanation can travel around the world in minutes; correcting it can take considerably longer. And while the company is establishing the facts, other people will speak.

Fraud has always existed, but the straightforward evidence used to distinguish legitimate customers from criminals was generally considered reliable. And then artificial intelligence came along and threw a massive spanner into that established system.

Today, a convincing voice can be generated from a few seconds of audio, just like a video call can be manipulated in real time. All it takes to assemble a realistic fake identity is some fragments of stolen information and AI‑generated documents, and soon it appears entirely legitimate.

The financial impact from this transformation has already proven massive. According to the Global Fraud Intelligence Report, fraud‑related losses reached roughly $442 billion in 2025, with AI being primarily responsible for reducing both the costs and the complexity of launching sophisticated attacks. The report also identified nearly 14,000 fraud rings in the first half of 2026 and warned that synthetic identities and deepfakes are fundamentally changing how financial crime works.¹

In other words, financial institutions face a greater issue than simply identifying forged documents or customer impersonation cases—increasingly, their real challenge is determining whether the person behind the identity exists in the first place.

This shift forces banks, exchanges, and payment providers to rethink one of the industry’s most fundamental concepts: trust. As AI continues to make digital deception much easier, trust itself is becoming a premium asset.

A convincing fake video of a CEO authorising a transfer is a very impressive technical accomplishment, which doesn’t stop it from also being deeply unsettling at the same time.

Yet many fraud investigators believe another threat could ultimately prove even more disruptive.

Kristiina Kikajoon

Director and Forensic Investigator at Crypto Legal

“From a forensic perspective, synthetic identities and AI‑generated social engineering are particularly concerning because they allow fraudsters to construct an apparently credible person or organisation rather than simply impersonating an existing one. The problem is therefore moving beyond identifying whether a document, photograph, voice or message is genuine; investigators increasingly need to establish whether the underlying identity and transaction itself are legitimate.”

Traditional fraud typically involves stealing an existing identity, but synthetic identity fraud creates an entirely new one by combining genuine information with fabricated details until it passes verification checks. The use of AI boosts the speed of that process, allowing criminals to generate convincing documentation, realistic profile pictures, and even seemingly authentic video interactions in a matter of hours.

After analysing over 1 billion identity verification events across 195 countries, Entrust found that deepfakes today account for 1 in every 5 biometric fraud attempts. Deepfaked selfie attacks increased 58% year over year, and injection attacks (where attackers bypass a device’s camera entirely) rose by 40%.²

In parallel with that, generative AI is making social engineering significantly more effective. Fraudsters can now generate highly personalised messages using publicly available information and tailoring them to individual victims.

As a result, we see before us a world where neither visual evidence, nor written communication, and not even a live conversation can be accepted at face value anymore.

I started my career in Australia’s real estate fund sector, advising high‑net‑worth investors on property fund allocations. One thing you learn very quickly is that capital moves when people understand the risk and know who is accountable when something goes wrong.

Traditional finance has spent decades building around that. There are established rules around custody, governance, conflicts of interest, segregation of assets and oversight. Technology can make transactions faster or improve the way people access an investment, but it does not make counterparty, legal or operational risk disappear.

Real estate makes this particularly clear because there is always a physical and legal asset underneath the financial product. You can improve how an ownership interest is recorded, transferred or serviced, but you still need to establish what somebody actually owns and whether that right can be enforced.

That is why I see issuance as the beginning of the work, not the end of it.

For me, AiFi starts one step before AI. It starts with financial assets that machines can actually understand.

A lot of financial information still relies on human interpretation. Someone reads the lease, checks the valuation, understands the legal structure, looks at the cash flow and puts all of that together to judge the risk. If software is going to take on more of that work, the information underneath the asset has to become much more structured and verifiable.

And price is only a small part of what a machine needs to know. What does the instrument legally represent? Who can hold it? How was the underlying property valued? What income does it generate? What restrictions apply if it is transferred? Can the transaction actually settle?

If we can make those rights, rules, and data clear enough for software to interpret reliably, then AI becomes genuinely useful. If we cannot, we are essentially asking it to process ambiguity faster.

Yes, and I think this becomes more important as AI moves deeper into professional financial workflows.

OpenAI’s recent launch of ChatGPT for Financial Services is a good example. We are starting to see financial datasets, research tools and firm‑specific ways of working brought directly into AI systems. That can be incredibly useful, but it also makes the origin of the information much more important.

AI does not encounter financial knowledge in a vacuum. It works with the sources, methodologies, and assumptions available to it. If some institutions or schools of thought appear far more often than others, the system can keep returning to them simply because they are easier to find.

Finance already knows how important this is. We ask where the data came from, what methodology was used and whether there is an audit trail. We should be asking the same questions of AI. A convincing answer is not enough if we cannot understand what sits behind it.

On Saturday, 28 February 2026, crypto markets suddenly became the only place where prices could react in real time, as traditional markets were closed. The first attack on Iran took place over the weekend, when every venue that normally prices geopolitical risk was closed. Oil‑linked perpetual contracts on Hyperliquid moved more than 5% almost immediately and delivered the first real‑time explanation of what this conflict could mean for the global.

At the same time, traders on traditional stock exchanges couldn’t react at all. It revived the question: can the industry built around trading hours govern markets that increasingly operate continuously?

Annabelle Huang

Co-Founder and CEO of Altius Labs

“No, and they are already being routed around. When the conflict with Iran escalated, traders did not wait for the opening bell. They moved onto blockchain rails to trade oil and gold around the clock, because that was the only place they could act on news that broke at three in the morning in someone else’s time zone.”

Of course, traditional institutions are seeing changes and trying to adapt—through banking partnerships with DeFi companies and the launch of their own stablecoins and blockchain‑based payments. Or they can buy this infrastructure instead of spending months on infrastructure development, as there is already a developed market for it.

However, even if those institutions can purchase required technical solutions, they still can’t buy their own decision cadence since it depends on regulators speed. This is best illustrated by the example of perpetual futures. Contracts with no expiration date on traditional assets became extremely popular during the decade, reaching $1.32 trillion in turnover, and financial companies also saw opportunities in these assets. However, the CFTC approved the first US perpetual contract only in May 2026.¹

Any large exchange could have listed those contracts years ago, but they still had to go through multiple legal barriers and board approvals.

Estelle Roiena

Institutional Finance & Assets Integrations Lead at Cardano Foundation

“The answer is not to automate every decision, but to distinguish between what can be governed through predefined, programmable rules and what still requires human judgement and accountability. At the Cardano Foundation, we call this “trust by architecture”: building verifiable records, clear rules and accountability into the infrastructure itself, so institutions can operate continuously without sacrificing governance.”

The SEC approved Nasdaq's move to 23 hours, launching on 6 December 2026, and approved 22-hour trading for the NYSE on the same date. It is not technically 24-hour trading because there is a one-hour break between sessions to process transactions for the next trading day. Also, weekends remain closed.

Even if the exchange could technically run trading 24 hours a day, settlements still move more slowly. The Fed wanted to run 22-hour-a-day, seven-day-a-week operations as early as 2024, but after reviewing industry comments, it announced six-day trading in 2028 or 2029, with any further expansion considered no sooner than two years after that. Seven-day operation for major currencies is expected to start no sooner than 2031, compared with 24-hour trading that began on DeFi markets in 2009.

Having been one of only two women studying physics in my year, then moving into technical and banking roles in London in my twenties, I’m accustomed to working in male‑dominated environments. That’s why I don’t give it much thought in my day‑to‑day work. However, that isn’t to say I’ve had an easy ride. Sexism and biases were a consistent feature of my career in the early 2000s.

I’d say the hardest part of my journey was pivoting from a successful commercial career into tech. I chose to start again from scratch in my thirties, kicking off with a full‑stack immersive software engineering bootcamp, and building my knowledge from there.

Of course, the programming experience I already had from university provided a solid foundation for my future career, but there have been plenty of moments where I’ve felt like an imposter in Web3. So I decided to focus on my contributions, instead of evaluation, as it’s difficult to hold in mind “Am I good enough?” at the same time as “How can I improve things for others?”

For now, my focus is mostly on bringing a pragmatic approach to decision‑making at 1inch, bridging strategy with architecture and technical execution so we can move fast with laser focus. The combined role suits my prior experience, but also makes a lot more sense in this new era of AI. In it, I think consolidation matters.

I’ve worked in startups and corporates, also within various management structures, and I personally prefer flatter organizations with more of a consensus‑driven approach. My leadership style is firm but fair, and the more visibility I give my teams, the more autonomous they can be. I genuinely care about my people—their successes are my successes.

My special flavour of autism means that I stay calm and collected under pressure. I believe it has also driven me to seek out bigger and more complex challenges throughout my career. Coupled with a strong sense of justice, it makes perfect sense that I’d be drawn to DeFi as a mechanism for promoting economic equity across digital and physical systems.

My job is to define what “good enough” looks like, and that depends on the product. As with other DeFi projects, we take security extremely seriously, and there are some aspects, such as audits, that we must always factor into our timelines.

As for Aqua’s technical case, it is no harder to explain than AMM pools or RFQ methodologies.
With Aqua, users’ wallets simply become the pools.

But according to the International Labour Organisation, that may well have been the wrong thing to focus on. Earlier in 2026, it came out with a research that showed female‑dominated occupations are almost twice as likely to be exposed to generative AI as male‑dominated ones (29% vs. 16%). Women are heavily concentrated in administrative, clerical, and business support roles, where many daily tasks are susceptible to automation—and yet they remain underrepresented in a lot of AI‑related and STEM occupations that are expected to benefit most from this technology.¹

To be clear: this doesn’t mean artificial intelligence will eliminate women’s jobs. History shows that new technology rarely replaces occupations outright; it changes them. Online banking, for example, changed the role of physical bank branches, but it did not erase them altogether. Likewise, AI today is expected to automate various tasks within many occupations, but that’s not the same as making those occupations disappear.

The more serious question is what happens when that same AI reshapes the roles that have traditionally served as the starting point for a career. Because before someone reaches top management positions, they usually spend years learning the business from the ground up.

And if that ground starts disappearing? What then?

If AI takes over performing a lion’s share of that “junior” work, firms may become more efficient, but they also risk losing a crucial leadership training ground. In fact, this concern is already spreading across the labour market. PwC’s 2026 AI Jobs Barometer highlights that artificial intelligence is compressing the traditional career ladder, and entry‑level roles exposed to it increasingly demand skills that once belonged to more senior employees.²

That sounds impressive, but then you begin to wonder where people are supposed to develop those skills in the first place. If AI increasingly handles the work that once gave junior professionals their first exposure to the realities of financial services, how are future executives to build up experience and know‑how?

Tracey Knowles

Chief People Officer at Wirex

“Entry‑level roles have never simply been collections of administrative tasks. They are where people learn how an organisation really works; how decisions are made, how customers are affected, how risk is managed and how to respond when things do not go according to plan.”

The issue here is not preserving existing jobs: few would argue that companies should continue asking people to manually perform tasks that modern tech can complete more efficiently. But it’s highly important to make sure that, as those routine tasks disappear, there is something to replace them. More specifically, that there are still ways and opportunities for people to learn and grow.

Otherwise, we risk creating a dangerous paradox. Companies may become more productive in the short term but gradually lose the very building blocks that have prepared people for leadership and allowed those organisations to prosper in the long term.

Nadia Edwards-Dashti

Co-Founder and Chief Customer Officer of Harrington Starr

“The more firms understand what AI can actually do well, the more I am seeing them redesign roles rather than simply remove them… I think the firms who get this right now will have a real advantage in five years' time when everyone else is scrambling for experienced talent that was never trained and simply isn’t there.”

“How many repetitive tasks can we get rid of and how many hours can be saved as a result?”

These are reasonable questions to ask, but efficiency is not the whole story; we also need to ask what happens after that efficiency has been achieved.