NEWS
NEWS
In AI We Trust? Rethinking the Rules of Confidence in Finance
In AI We Trust? Rethinking the Rules of Confidence in Finance
As 2026 draws to a close, more and more people are turning to AI with questions about managing their money. TD Bank’s 2026 survey found that 55% of respondents use AI to help manage their finances, up from 10% a year earlier. Banks are wary, as in October 2026, several large US banks moved to restrict third-party AI agents from accessing customer accounts.
So, if an AI assistant starts handling your money, can you trust it, and who is responsible if it makes a mistake?
As part of the Women Leading the Way partnership with CryptoCon Sydney 2026, the Drofa Comms team interviewed five female experts, each of whom brings a fresh perspective on the matter. Let's break the topic down and find the answer.
Who Takes the Blame for an AI’s Wrong Answer?
When customers use a bank’s own app, they generally know which institution to contact if a problem arises. Introducing an AI service into that interaction can make responsibility less clear.
Imagine authorising an AI assistant to initiate a bank transfer on your behalf. That could involve you, the AI service provider and the bank, raising questions about permission, oversight and accountability in case of a problem.

Yvonne Nehme, Senior Associate at HWLE Lawyers, calls this a shared relationship. Banks may keep the product, customers’ funds and regulatory obligations, while AI platforms increasingly control the daily conversation with the customer, and with it the data and influence over purchasing decisions.
Yet handing the conversation over to an AI does not mean handing over the responsibility. Gabby Lewis, Head of Fraud & Financial Crime at Swyftx, stresses: "you can't outsource responsibility, risk or relationship ownership to a model or a vendor. If you build it or deploy it, you own how it's governed and controlled."

The caveat is that there is still no common agreement on who is liable when an agent sends money to the wrong place. Justine Bone, Executive Director of Crypto ISAC, puts it this way: "Ownership often follows liability, and liability hasn't been assigned yet." She adds that liability cannot be assigned until it is clear who acted: a person or an AI agent, and with whose authority.
Australia illustrates the implementation questions involved. Legislation passed in 2024 created a framework for Consumer Data Right action initiation, allowing authorised instructions to be sent to service providers on a customer’s behalf. Legal analysis identified the allocation of liability as a priority for implementation. Reporting in May 2026 indicated that the system still supported data sharing rather than initiating actions.
Some experts believe the focus should remain on the person. For instance, Olivia Long, Founder and CEO of SMSFAI, emphasises: "it's your money, and you should remain in control."

At Drofa Comms, we see that there is still no definite consensus on these questions, even though AI agents are making their way into more and more of our daily lives.
As AI takes on a greater role in financial services, it is important to establish who is acting, what authority they have and who is accountable if an error occurs.
Why Human Judgement Still Carries Weight
If AI gives financial guidance for free, do people still need human experts? On the first part the speakers agree: the basics are getting cheap.
Yvonne expects budgeting tips, product comparisons and simple portfolio suggestions to become available instantly to anyone. Jenny Wei, General Manager of Binance Australia and New Zealand, welcomes it, since more people can now get information once reserved for those who could pay for it. But she is careful to add: "Information may become a commodity, but judgement won't." In other words, facts get cheap, but knowing what to do with them does not.

Olivia explains why. Behind a question like "Can I afford to retire?" is a whole life, with family commitments and an uncertain future. AI can run the analysis, she says, but an experienced adviser notices what the client has not asked.
There is also the matter of how AI sounds. Gabby notes that AI models "sound just as confident when they're wrong, and tend to agree with whoever is asking." For someone already leaning towards a risky investment, an agreeable assistant can simply confirm the plan.
The public seems to draw the same line. A 2025 CFP Board survey of US adults aged 25 to 64 with household incomes above $50,000 found that three in four would act on a financial adviser's guidance without verifying it elsewhere, while only 38% felt comfortable acting on advice from generative AI tools.
We read it this way: the scarce skill is no longer producing advice, but being able to tell where that advice is useful and where it isn't. That is the part that stays with humans. And when a mistake cannot be undone, it becomes an indispensable skill.
Risks That Differ in TradFi and DeFi
So far, we have focused mainly on traditional finance. AI is also entering crypto, as according to American Banker, Coinbase and Robinhood are opening their platforms to outside AI agents. These are centralised platforms, distinct from decentralised finance (DeFi), which uses blockchain-based protocols to provide financial services. But how do the risks of AI differ between traditional finance and DeFi?
In TradFi, Yvonne says, the problem is opacity. For example, AI already works inside regulated processes, making decisions such as who gets a loan, and it can produce biased outcomes and models that nobody can explain.
In DeFi, smart contracts, which are programs that run on their own, and AI agents can act automatically and irreversibly, while the technology and its safeguards are still being built. TradFi, at the very least, has decades of regulation behind it.
Justine highlights whether a payment can be reversed. Traditional finance can offer recovery routes that may not be available for on-chain transactions. "Reversibility is the difference that matters, and in crypto irreversibility isn't an oversight, it's a feature by design." If an on-chain transaction cannot be reversed, recovering funds after an AI error or scam can be difficult.

Jenny sees the flip side: blockchain activity is transparent, so it can be monitored in real time, with transaction limits and a human ready to step in. These differences decide which safeguards each side needs, but is there anything we can already rely on?
Readiness remains uneven. As Yvonne sums it up: “Neither sector is fully prepared, but the nature of the gap differs.”
In April 2026, APRA found that AI governance and risk management in the financial institutions it regulates had not kept pace with adoption. The regulator did not propose new requirements at the time, but called on firms to improve their ability to monitor and control the technology. This suggests that financial institutions already have work to do, regardless of whether new rules are introduced.
AI and Money in 2027: What Firms Need to Settle
Summing up, can we trust AI with our money? Conditionally, yes. Confidence in AI depends less on how clever the model is than on the rules around it: it is clear who is responsible, who is acting, and who can stop the operation when the stakes grow higher.
TradFi firms and DeFi platforms can already take steps to address these questions:
Name the owner. For every AI-driven operation, it is clear who answers if something goes wrong.
Verify who is acting. Check whether a person or an AI program is acting, and on whose behalf.
Keep a human where the stakes are high. Where a mistake cannot be reversed, the final decision stays with a person.
Build the brakes before the speed. Set transfer limits and watch for unusual activity, with a way to stop an operation, before raising speed.
Let's not forget that trust is a matter of communication. Companies that explain clearly where AI is used, when a person steps in and who is responsible are better poised to earn customers’ confidence.
Acknowledgements: Drofa Comms would like to thank Jenny Wei, Gabby Lewis, Justine Bone, Yvonne Nehme and Olivia Long for sharing their insights and contributing to this partner article.
About Women Leading the Way
Women Leading the Way is an independent initiative by Drofa Comms that promotes the visibility of women in finance, fintech, Web3 and technology. Launched in London in 2024, the project gives female professionals a platform to share their expertise and experiences through interviews, podcasts, an annual magazine, a professional community and industry events.
About CryptoCon Sydney
CryptoCon Sydney is Australia’s flagship crypto, blockchain and Web3 event. Held at ICC Sydney on 28–29 November 2026, the event connects founders, investors, builders, institutions, industry professionals and the wider crypto community through insight-led programming, startup discovery and high-value networking.
As 2026 draws to a close, more and more people are turning to AI with questions about managing their money. TD Bank’s 2026 survey found that 55% of respondents use AI to help manage their finances, up from 10% a year earlier. Banks are wary, as in October 2026, several large US banks moved to restrict third-party AI agents from accessing customer accounts.
So, if an AI assistant starts handling your money, can you trust it, and who is responsible if it makes a mistake?
As part of the Women Leading the Way partnership with CryptoCon Sydney 2026, the Drofa Comms team interviewed five female experts, each of whom brings a fresh perspective on the matter. Let's break the topic down and find the answer.
Who Takes the Blame for an AI’s Wrong Answer?
When customers use a bank’s own app, they generally know which institution to contact if a problem arises. Introducing an AI service into that interaction can make responsibility less clear.
Imagine authorising an AI assistant to initiate a bank transfer on your behalf. That could involve you, the AI service provider and the bank, raising questions about permission, oversight and accountability in case of a problem.

Yvonne Nehme, Senior Associate at HWLE Lawyers, calls this a shared relationship. Banks may keep the product, customers’ funds and regulatory obligations, while AI platforms increasingly control the daily conversation with the customer, and with it the data and influence over purchasing decisions.
Yet handing the conversation over to an AI does not mean handing over the responsibility. Gabby Lewis, Head of Fraud & Financial Crime at Swyftx, stresses: "you can't outsource responsibility, risk or relationship ownership to a model or a vendor. If you build it or deploy it, you own how it's governed and controlled."

The caveat is that there is still no common agreement on who is liable when an agent sends money to the wrong place. Justine Bone, Executive Director of Crypto ISAC, puts it this way: "Ownership often follows liability, and liability hasn't been assigned yet." She adds that liability cannot be assigned until it is clear who acted: a person or an AI agent, and with whose authority.
Australia illustrates the implementation questions involved. Legislation passed in 2024 created a framework for Consumer Data Right action initiation, allowing authorised instructions to be sent to service providers on a customer’s behalf. Legal analysis identified the allocation of liability as a priority for implementation. Reporting in May 2026 indicated that the system still supported data sharing rather than initiating actions.
Some experts believe the focus should remain on the person. For instance, Olivia Long, Founder and CEO of SMSFAI, emphasises: "it's your money, and you should remain in control."

At Drofa Comms, we see that there is still no definite consensus on these questions, even though AI agents are making their way into more and more of our daily lives.
As AI takes on a greater role in financial services, it is important to establish who is acting, what authority they have and who is accountable if an error occurs.
Why Human Judgement Still Carries Weight
If AI gives financial guidance for free, do people still need human experts? On the first part the speakers agree: the basics are getting cheap.
Yvonne expects budgeting tips, product comparisons and simple portfolio suggestions to become available instantly to anyone. Jenny Wei, General Manager of Binance Australia and New Zealand, welcomes it, since more people can now get information once reserved for those who could pay for it. But she is careful to add: "Information may become a commodity, but judgement won't." In other words, facts get cheap, but knowing what to do with them does not.

Olivia explains why. Behind a question like "Can I afford to retire?" is a whole life, with family commitments and an uncertain future. AI can run the analysis, she says, but an experienced adviser notices what the client has not asked.
There is also the matter of how AI sounds. Gabby notes that AI models "sound just as confident when they're wrong, and tend to agree with whoever is asking." For someone already leaning towards a risky investment, an agreeable assistant can simply confirm the plan.
The public seems to draw the same line. A 2025 CFP Board survey of US adults aged 25 to 64 with household incomes above $50,000 found that three in four would act on a financial adviser's guidance without verifying it elsewhere, while only 38% felt comfortable acting on advice from generative AI tools.
We read it this way: the scarce skill is no longer producing advice, but being able to tell where that advice is useful and where it isn't. That is the part that stays with humans. And when a mistake cannot be undone, it becomes an indispensable skill.
Risks That Differ in TradFi and DeFi
So far, we have focused mainly on traditional finance. AI is also entering crypto, as according to American Banker, Coinbase and Robinhood are opening their platforms to outside AI agents. These are centralised platforms, distinct from decentralised finance (DeFi), which uses blockchain-based protocols to provide financial services. But how do the risks of AI differ between traditional finance and DeFi?
In TradFi, Yvonne says, the problem is opacity. For example, AI already works inside regulated processes, making decisions such as who gets a loan, and it can produce biased outcomes and models that nobody can explain.
In DeFi, smart contracts, which are programs that run on their own, and AI agents can act automatically and irreversibly, while the technology and its safeguards are still being built. TradFi, at the very least, has decades of regulation behind it.
Justine highlights whether a payment can be reversed. Traditional finance can offer recovery routes that may not be available for on-chain transactions. "Reversibility is the difference that matters, and in crypto irreversibility isn't an oversight, it's a feature by design." If an on-chain transaction cannot be reversed, recovering funds after an AI error or scam can be difficult.

Jenny sees the flip side: blockchain activity is transparent, so it can be monitored in real time, with transaction limits and a human ready to step in. These differences decide which safeguards each side needs, but is there anything we can already rely on?
Readiness remains uneven. As Yvonne sums it up: “Neither sector is fully prepared, but the nature of the gap differs.”
In April 2026, APRA found that AI governance and risk management in the financial institutions it regulates had not kept pace with adoption. The regulator did not propose new requirements at the time, but called on firms to improve their ability to monitor and control the technology. This suggests that financial institutions already have work to do, regardless of whether new rules are introduced.
AI and Money in 2027: What Firms Need to Settle
Summing up, can we trust AI with our money? Conditionally, yes. Confidence in AI depends less on how clever the model is than on the rules around it: it is clear who is responsible, who is acting, and who can stop the operation when the stakes grow higher.
TradFi firms and DeFi platforms can already take steps to address these questions:
Name the owner. For every AI-driven operation, it is clear who answers if something goes wrong.
Verify who is acting. Check whether a person or an AI program is acting, and on whose behalf.
Keep a human where the stakes are high. Where a mistake cannot be reversed, the final decision stays with a person.
Build the brakes before the speed. Set transfer limits and watch for unusual activity, with a way to stop an operation, before raising speed.
Let's not forget that trust is a matter of communication. Companies that explain clearly where AI is used, when a person steps in and who is responsible are better poised to earn customers’ confidence.
Acknowledgements: Drofa Comms would like to thank Jenny Wei, Gabby Lewis, Justine Bone, Yvonne Nehme and Olivia Long for sharing their insights and contributing to this partner article.
About Women Leading the Way
Women Leading the Way is an independent initiative by Drofa Comms that promotes the visibility of women in finance, fintech, Web3 and technology. Launched in London in 2024, the project gives female professionals a platform to share their expertise and experiences through interviews, podcasts, an annual magazine, a professional community and industry events.
About CryptoCon Sydney
CryptoCon Sydney is Australia’s flagship crypto, blockchain and Web3 event. Held at ICC Sydney on 28–29 November 2026, the event connects founders, investors, builders, institutions, industry professionals and the wider crypto community through insight-led programming, startup discovery and high-value networking.
London office
Rise, created by Barclays, 41 Luke St, London EC2A 4DP
Nicosia office
2043, Nikokreontos 29, office 202
DP FINANCE COMM LTD (#13523955) Registered Address: N1 7GU, 20-22 Wenlock Road, London, United Kingdom For Operations In The UK
AGAFIYA CONSULTING LTD (#HE 380737) Registered Address: 2043, Nikokreontos 29, Flat 202, Strovolos, Cyprus For Operations In The EU, LATAM, United Stated Of America And Provision Of Services Worldwide
Drofa © 2024
London office
Rise, created by Barclays, 41 Luke St, London EC2A 4DP
Nicosia office
2043, Nikokreontos 29, office 202
DP FINANCE COMM LTD (#13523955) Registered Address: N1 7GU, 20-22 Wenlock Road, London, United Kingdom For Operations In The UK
AGAFIYA CONSULTING LTD (#HE 380737) Registered Address: 2043, Nikokreontos 29, Flat 202, Strovolos, Cyprus For Operations In The EU, LATAM, United Stated Of America And Provision Of Services Worldwide
Drofa © 2024
