There is a question quietly working its way to the top of the agenda at board meetings, risk committees, and technology investment panels across the financial services sector.
It is not particularly new, but it has grown considerably more urgent: can artificial intelligence genuinely be trusted in an investment context and if so, under what conditions?
For financial institutions, private banks, wealth managers, and asset managers, the conversation around AI investment advice has moved well beyond the theoretical.
AI is already embedded in how data is processed, how anomalies are flagged, and how internal teams are supported in making decisions. The question is how to deploy it with confidence, clarity, and appropriate oversight.
This article is written specifically for institutional stakeholders, who are exploring digital advisory infrastructure, evaluating AI-powered tools, or simply trying to understand where the technology sits within a regulated financial environment.
Let us start with what is actually happening on the ground. Across the financial services industry, institutions are integrating AI into internal workflows in genuinely meaningful ways, though perhaps less dramatically than the headlines might suggest.
According to a November 2024 report by the Financial Stability Board (FSB), financial institutions are primarily using AI to enhance internal operations and improve regulatory compliance, with revenue-generating use cases still relatively limited. (1)
In practical terms, this means AI is being applied to data aggregation, pattern recognition, transaction monitoring, fraud detection, and reporting support.
The European Securities and Markets Authority (ESMA), in its May 2024 Public Statement on AI in Investment Services, identified a range of potential applications for AI at investment firms. The key word there is “support.” (2)
AI, as currently deployed by most financial institutions, is a support layer for professional judgement and not a replacement for it.
This distinction matters enormously when evaluating AI for investment management. Under MiFID II, investment advice and portfolio management remain regulated services, subject to specific obligations around suitability, transparency, and professional accountability. (3)
AI tools operating in this space must therefore be understood as instruments that assist licensed professionals, not as autonomous decision-makers.
Trust is a word that gets used rather loosely in conversations about technology. In the context of AI financial advisor tools and AI-assisted investment infrastructure, trust is a set of verifiable properties.
Transparency: Can the institution understand, at a meaningful level, what the AI system is doing and why? A system whose logic is entirely opaque to the people using it is difficult to trust and difficult to defend to regulators.
Explainability: Can outputs be traced back to identifiable inputs and reasoning? This is particularly relevant in investment contexts, where decisions may need to be reviewed, justified, or challenged.
Auditability: Does the system produce records that document its operations, data sources, and decision support processes? ESMA has explicitly stated that it expects firms to maintain documentation covering AI deployment, including decision-making processes, data sources, and algorithms implemented. (4)
Human Oversight: Does the governance framework ensure that licensed professionals remain accountable for outcomes? Regulatory responsibility cannot be delegated to an algorithm.
ESMA has also been clear that management bodies within investment firms need to have an appropriate understanding of how AI technologies are being used, and to ensure adequate oversight.
Trustworthy AI, in other words, is not magic. It is infrastructure with guardrails, documentation, and human accountability baked in.
Whether you are evaluating an AI financial advisor platform, building out AI for investment management workflows, or reviewing the governance frameworks around existing tools, the following questions are a useful starting point.
The quality and relevance of training data directly affect the reliability of outputs.
Systems trained on narrow, outdated, or biased datasets will reflect those limitations in their outputs.
Understanding the provenance and composition of training data is a baseline expectation for any responsible deployment.
ESMA has specifically warned of risks, including algorithmic biases, data quality issues, and opaque decision-making by staff relying too heavily on AI outputs.
Any evaluation of an AI tool should include a clear account of how the system has been tested for bias, what happens when data quality is poor, and how the system behaves under unusual or edge-case conditions.
Firms integrating AI into regulated workflows need to be able to demonstrate, if asked, how their systems operate and what records they maintain.
The expectation from ESMA is explicit: documentation of AI deployment, including decision-making processes, data sources used, and algorithms implemented.
This is perhaps the most critical question of all. AI tools do not carry regulatory authorisation.
They support institutions that do. The burden of compliance with MiFID II, with data protection law, with whatever sector-specific obligations apply, remains with the licensed firm.
Understanding the limitations of AI in a financial context is just as important as understanding its capabilities. Here are the things that no AI system, however sophisticated, can do on its own.
It is a realistic accounting of where AI sits in the financial services ecosystem today. The technology is genuinely powerful, but its power is best realised when it is deployed within a framework of clear accountability and strong governance.
Gambit Finance is an IT solutions provider to financial institutions. Its role in the ecosystem is to provide the technology infrastructure that helps institutions build and run their own regulated workflows, including those that incorporate AI investment advice support tools, more effectively.
Financial institutions that work with Gambit use its tools to support their own internal processes. Gambit’s products are not offered directly to consumers or retail investors, and nothing in Gambit’s technology changes the regulatory obligations of the institutions that use it.
This is, frankly, the right model for how technology providers should operate in a regulated space. The goal of a good IT partner is to make the people providing those services more capable, more efficient, and better supported.
Whether you are exploring AI investment advice infrastructure, reviewing how AI for investment management might fit into your operating model, or simply keeping pace with a rapidly changing regulatory and technology landscape, the conversation starts not with the AI itself but with the governance and accountability framework that surrounds it.
The question of whether AI can be trusted with investment decisions does not have a simple yes or no answer. AI can be a powerful, valuable support layer for financial institutions when it is governed responsibly, deployed transparently, and kept firmly within a framework of human accountability.
The same applies whether you are deploying a standalone AI financial advisor support tool or a more comprehensive AI-assisted workflow.
The institutions best placed to benefit from AI are those that approach it not as a solution in search of a problem, but as a technology that requires the same rigour, documentation, and oversight as any other material system they operate.
Before evaluating or implementing any AI-driven tool in a regulated context, we would encourage all institutional stakeholders to consult their own legal and compliance teams.
The regulatory landscape around AI in financial services is evolving quickly, and the right framework for your institution will depend on your specific circumstances, jurisdiction, and obligations.
Gambit’s role is to be a capable, responsible technology partner to the institutions navigating this landscape. The judgment and the accountability always belong to you.
1. What is the “black box” problem in AI, and why does it matter for financial institutions?
The “black box” problem refers to the difficulty in understanding how some AI models arrive at their decisions. In financial contexts, this lack of transparency can create challenges for auditability, compliance, and client trust.
2. Can AI in financial services produce biased or discriminatory outcomes?
Yes. AI systems can inherit biases from their training data or design. This is why robust data governance, testing, and oversight are essential when implementing AI solutions.
3. How does data privacy law interact with the use of AI in financial services?
AI systems often process large volumes of personal and financial data. Institutions must ensure compliance with applicable data protection regulations, including how data is collected, stored, and used.
4. What systemic risks could arise if many financial institutions rely on the same AI models?
Widespread reliance on similar AI systems could lead to correlated behaviours across institutions, potentially amplifying market volatility or systemic risks. Diversification in models and strong governance frameworks can help mitigate this.
Disclaimer: This publication's content is solely intended for informational and promotional purposes. This publication is not intended to provide contractual offers or professional financial advice.
References:
1. Financial stability implications of artificial intelligence - Executive Summary. (2025, June 26). https://www.bis.org/fsi/fsisummaries/exsum_23904.htm
2. ESMA provides guidance to firms using artificial intelligence in investment services. (n.d.). https://www.esma.europa.eu/press-news/esma-news/esma-provides-guidance-firms-using-artificial-intelligence-investment-servicesESMA provides guidance to firms using artificial intelligence in investment services. (n.d.). https://www.esma.europa.eu/press-news/esma-news/esma-provides-guidance-firms-using-artificial-intelligence-investment-services
3. ESMA provides guidance to firms using artificial intelligence in investment services. (n.d.). https://www.esma.europa.eu/press-news/esma-news/esma-provides-guidance-firms-using-artificial-intelligence-investment-services
4. ESMA issues guidance on AI in retail financial services as EU AI Act takes effect. (2024, August 19). https://www.morganlewis.com/pubs/2024/08/esma-issues-guidance-on-ai-in-retail-financial-services-as-eu-ai-act-takes-effect
5. ESMA issues guidance on AI in retail financial services as EU AI Act takes effect. (2024, August 19). https://www.morganlewis.com/pubs/2024/08/esma-issues-guidance-on-ai-in-retail-financial-services-as-eu-ai-act-takes-effect
6. Kennedy, R., & Kennedy, R. (2024, December 12). BIS highlights AI risks for finance. The Global Treasurer. https://www.theglobaltreasurer.com/2024/12/12/bis-highlights-ai-risks-for-finance/