AIFS0004

Written evidence submitted by Bonsai Smart Wealth Ltd

Introduction

I’m submitting this response on behalf of Bonsai, a UK-based fintech focused on making high-quality financial planning and investing accessible to everyone. We’re developing an AI-powered financial planning platform that combines intelligent automation with human oversight, allowing people to plan, save, and invest with confidence - regardless of their wealth, income, or background. Our mission is to finally remove the advice gap once and for all.

We are contributing to this inquiry because we strongly believe that AI represents the best chance the UK has to close the widening financial advice gap, helping millions of people access affordable, personalised financial advice at scale. However, as a fintech working at the forefront of AI innovation, we’ve seen first-hand how regulatory bottlenecks and outdated approval processes at the FCA are slowing this progress, particularly for companies trying to build innovative AI-first financial products.

How is AI currently used in financial services and how is this likely to change over the next ten years?

AI is already widely used in sectors like payments and fraud detection, but there is also significant transformative potential within financial advice and wealth management. In this space, AI can deliver personalised, responsive, and cost-effective guidance - something that has traditionally been limited to wealthier individuals due to the high cost of human advice, minimum investment sizes, and conflicting commercial models.

At Bonsai, we use AI to:

• Build goal-based financial plans, tailored to each user’s circumstances and preferences.
• Deliver on-demand financial guidance through a conversational co-pilot, answering questions that users might otherwise avoid asking a professional.
• Identify opportunities for smarter saving and investing, using transaction data and spending habits to offer personalised nudges.

Over the next decade, we expect this to evolve into integrated, and an intertwined financial ecosystem with seamless interaction between a user’s financial affairs. Continuous, embedded financial coaching, where AI provides real-time financial advice based on live changes to a customer’s situation - salary changes, unexpected expenses, or even broader economic shifts.

Fintech firms, particularly startups like ours, are better suited to adopting AI because we’re building products from scratch without legacy technology stacks, shareholds, competing budgets etc holding us back. However, the regulatory framework is still designed for the old model of human advice, making it far harder for AI-first firms to enter the market, and approval timelines are far too long for startups. For examples, our initial authorisation application took 2 months just to be picked up, and was then immediately rejected due to a few missing documents. In parallel we applied to the Reg Sandbox, which took over a month to be picked up.

To what extent can AI improve productivity in financial services?

The productivity potential of AI in financial advice is enormous. Traditional financial advice relies on lengthy, manual processes, and client relationships often unfold over weeks or months. AI compresses that into minutes, offering:

• Fully automated fact finds using open banking data and smart questionnaires.
• Instant risk profiling and asset allocation recommendations.
• On-demand financial education tailored to a user’s needs and understanding.
• Convenient time frames e.g. not having to meet your adviser Mon – Fri, 0900-1700
• Reduction in the error potential through intelligent automation and reasoning vs human error and continuous learning and development needs. E.g.  AI will not need to refresh its memory if it hasn’t worked on an IHT case for months.

This means that for every adviser or digital platform using AI effectively, they can serve exponentially more customers - at a fraction of the cost. This doesn’t just boost productivity for businesses; it also makes professional grade advice accessible to people who would never have engaged with traditional financial advisers.

The main barriers to this productivity boost are not technical. They’re regulatory and cultural. The FCA’s authorisation process is out of step with the pace of AI innovation, and the lack of clear AI-specific guidance leaves firms like ours uncertain about exactly what standards apply to AI-powered advice.

What are the risks to financial stability arising from AI and how can they be mitigated?

AI does introduce risks, especially when algorithms influence investment decisions across thousands or millions of customers simultaneously. Some of the key risks include:

• Herding behaviour, where multiple AI platforms respond to market signals in the same way, amplifying volatility.
• Model degradation, where AI models become less effective if not regularly retrained and revalidated.
• Over-reliance on a small group of third-party AI providers, creating potential concentration risk across the sector.
• Data dependency, where AI models rely heavily on the same historical data sources. If that data is incomplete, biased, or outdated, it creates systemic risk across the sector, with multiple firms building strategies on flawed assumptions.
• Algorithmic amplification during market stress, where AI models could automatically trigger the same defensive actions (e.g., mass sell-offs) in response to similar signals, accelerating downturns before human oversight can intervene. Although, this is not too dissimilar to algorithmic models already in existence.

These risks can be managed through sensible, proportionate regulation and internal audits/policies. Firms using AI in advice and investment should be required to:

• Conduct regular stress testing to understand how AI models would behave in adverse scenarios.
• Maintain clear, auditable records of how AI decisions are made, ensuring explainability.
• Diversify their AI technology stack, reducing over-reliance on a handful of major tech firms.
• Perform regular data quality and bias audits, ensuring models are trained on diverse, reliable, and up-to-date data sources.
• Introduce built-in circuit breakers within AI platforms, ensuring human intervention can override automated decisions when markets experience extreme volatility.

What are the benefits and risks to consumers arising from AI, particularly for vulnerable consumers?

The biggest benefit of AI in financial advice is that it breaks down barriers. It makes financial planning affordable and accessible to people who have traditionally been locked out - those with modest savings, younger investors, or people who simply didn’t know where to start.

AI can also improve financial literacy by offering simple, timely explanations alongside every recommendation. This is particularly valuable for vulnerable consumers who may lack confidence in their financial knowledge. AI also has much greater computing power than the human mind, subsequently allowing AI to identify a vulnerable customer much faster, without any conflicts of interest or biases.

That said, there are risks. Poorly designed AI could perpetuate or even amplify biases - for example, steering lower-income users toward riskier products simply because they need higher returns to meet their goals. There’s also a risk that some consumers will over-trust the technology, following AI-generated advice without fully understanding it; although, this is not too dissimilar to the current financial advice model.

To protect consumers, especially vulnerable ones, AI firms should be required to:

• Ensure all recommendations come with clear, simple explanations, so users know why a particular action is suggested.
• Conduct regular fairness and bias audits, ensuring the AI is treating all users fairly.
• Offer human support as a backstop, particularly for complex or sensitive financial decisions.

How can government and financial regulators strike the right balance between seizing the opportunities of AI and protecting consumers and financial stability?

The current regulatory approach is too slow and too vague to properly support AI innovation. A 12-month SLA for authorisation is completely incompatible with the speed of AI development. Worse still, the FCA’s practice of resetting the clock upon rejection creates misleading approval statistics and discourages firms from applying in the first place. It is a common experience and perception that the FCA is too slow for the UK to compete on innovation – whilst the UK arguably has the strongest financial regulations globally, it is hurting innovation, so I see the importance of this question.

The government and regulators need to modernise the authorisation process and make it a lot more inclusive for new businesses and existing businesses rolling our product advancements. This could include:

• Introducing a dedicated AI authorisation track, with faster timelines and AI-specific review criteria.
• Integrate AI into initial screening for almost immediate responses e.g. have all documents been added and signed, have the right permissions been applied for per the business plan etc. This would considerably reduce the cost and resource requirement for the FCA.
• Expanding the Regulatory Sandbox to offer ongoing guidance to AI-first firms, not just temporary testing windows. Temprary testing widows are great, but they pose huge threats and uncertainty for continuation i.e. a company acquires 1,000 customers and then the FCA decide not to proceed (although, I’m not sure how likely this is)
• Issuing clear, proportionate guidance on how Consumer Duty applies to algorithmic advice, whilst prioritising the development of AI frameworks and regulations, especially around consumer financial advice.
• Appointing an AI Innovation Lead within the FCA to bridge the gap between technology innovators and regulatory policy teams.

AI represents a once-in-a-generation opportunity to close the advice gap and improve financial resilience across the UK. If we embrace it, the UK can become a world leader in AI-powered financial services. If we allow outdated regulatory processes to hold it back, we risk falling behind more agile and pro-innovation/ pro-growth jurisdictions, losing investment, talent, and innovation to other countries.

February 2025