As a fiduciary advisor, I am responsible for every recommendation I make to a client, even when that recommendation is informed by software I did not build and cannot fully explain. That reality is becoming harder to ignore as AI increasingly shapes portfolio construction, risk analysis and even client communications.
Fiduciary duty has always depended on understanding not just what’s happening in financial markets and in a client’s portfolio, but why. AI, on the other hand, is built to find patterns we can’t see, often in ways we can’t fully explain.
I run Axon Capital Management, a fee-only fiduciary firm based in Austin, Texas. Like many advisors, I’m experimenting with AI tools in research and planning. They’re fast, insightful and sometimes startlingly good. But they also make me stop and ask: Where are the boundaries between the algorithm and advisors’ responsibility? And can we reap the benefits of AI without eroding the trust that underpins our profession?
The risks of delegating to AI
AI has a firm foothold in the world of finance. Quant firms like Renaissance have used machine learning to generate trading signals for years. Private equity funds scrape hiring data and satellite images to track growth before it shows up in earnings. Even smaller advisory firms now use tools that summarize client data, run Monte Carlo simulations and draft personalized reports in seconds.
The benefits are real. AI can process enormous amounts of data, highlight inefficiencies and help make portfolios more disciplined. But its strength — pattern recognition — is also its weakness. These models don’t reason. They correlate. When they’re wrong, the errors are often invisible until the damage is done.
This tension is uncomfortable for fiduciaries. Clients deserve to understand why we make a recommendation. If the answer is “because the algorithm said so,” we’ve failed. Regulators agree. With the SEC’s 2023 proposals to regulate advisors’ use of predictive technologies now withdrawn, firms must look to the existing Advisers Act framework and SEC guidance to ensure their use of AI remains accurate, transparent and compliant.
Maintaining fiduciary duty
The fiduciary standard has two pillars: the duty of loyalty and the duty of care. AI is pressure testing both.
Under the duty of loyalty, fiduciaries must ensure that any technology they use serves their client, not the firm or the software provider. For example, if a model steers assets toward products with higher internal fees, it is generating a conflict of interest. Advisors should know how vendors train their systems and what incentives may shape their outcomes.
Under the duty of care, fiduciaries must act with diligence and prudence. When it comes to AI, fiduciaries don’t need to be data scientists, but they do need to understand what an algorithm is doing at a basic level, including the data it relies on, the biases it may carry and how it is validated over time. Blindly following a black-box model cedes responsibility and accountability.
Building a responsible framework
Advisors don’t need to choose between embracing artificial intelligence and protecting their clients. The real responsibility lies in how AI is used — as a tool that enhances professional judgment, not replaces it.
In practice, most advisors are not building proprietary AI models or auditing neural networks line by line. The responsibility is not to become a machine learning engineer, but to understand how a given tool is being used, what role it plays in decision-making and where human judgment must remain firmly in control.
When evaluating any AI-enabled tool, I follow this framework:
Transparency: Can I clearly explain what the tool does, what data it uses and what it does not do? For example, if an AI tool summarizes financial plans or highlights planning opportunities, I need to understand whether it is generating original analysis or simply organizing existing information.
Alignment: Does the tool support better outcomes for the client or primarily benefit the firm through efficiency or scale? Tools that surface planning insights, scenario comparisons or risk flags can enhance advice. Tools that push product recommendations or optimize for sales metrics should be treated with skepticism.
Oversight: AI should not make decisions but inform them. Any output must be reviewed, contextualized and validated by a human advisor who understands the client’s full financial picture. This includes checking assumptions, confirming data accuracy and applying judgment where nuance matters.
Security and data handling: Advisors must understand where client data is stored, how it is encrypted and whether it is used to train external models. Responsible use means choosing platforms with clear data governance policies and avoiding tools that introduce unnecessary privacy risk.
Disclosure: Clients deserve to know when AI tools play a role in their planning. Transparency builds trust and reinforces that the advisor, not the algorithm, remains accountable for every recommendation.
In my own practice, AI is used narrowly and intentionally, for example, to help organize data, stress-test scenarios or identify planning considerations that warrant deeper human review. It is never used to generate investment recommendations on its own or to replace fiduciary judgment. Every output is reviewed, challenged and contextualized before it ever reaches a client.
The hidden bias problem
Every model reflects its data. That means AI often carries forward the same historical biases embedded in financial markets.
For example, a credit-risk model trained on traditional lending data may unintentionally penalize underbanked communities. Similarly, an ESG-scoring system might favor large corporations with sophisticated disclosure teams over smaller, genuinely sustainable businesses. If fiduciaries rely on these systems uncritically, they risk reinforcing inequities that impact investing was designed to correct.
Some firms are beginning to take a more deliberate approach to AI by embedding ethical guardrails directly into how these tools are used, rather than treating them as black boxes. Companies like Anthropic have pioneered ideas such as “constitutional AI,” in which models are guided by defined principles around transparency and accountability, and a similar mindset is starting to emerge in wealth management.
In practice, this doesn’t mean advisors are building AI models themselves, but rather choosing purpose-built tools such as FP Alpha, Holistiplan or RightCapital that use automation to surface planning insights while keeping humans firmly in control of decision-making. The responsibility then shifts to advisors to define clear rules around data usage, oversight and disclosure: understanding what the tool does, what data it touches and where human judgment is needed.
The human in the loop
For all the talk of automation, a lot of financial advice still comes down to human emotion. The rise of AI makes the human side of advising more important than ever.
A client thinking about retiring early needs human reassurance that their life savings will support the next chapter. A client who is panicking about market volatility may need someone with a cooler head to talk to. AI can calculate probabilities, but it can’t read hesitation in someone’s voice or understand why they’re scared to sell a stock that’s tied to their identity.
Fiduciary duty is fundamentally about stewardship and trust — placing human judgment at the center and using the best available tools to make careful, transparent decisions in our clients’ best interests.
Advisors who lead in this space will be the ones who can use AI confidently without hiding behind it and who can pair technological precision with human empathy. They’ll be the ones who remember that trust is earned conversation by conversation and decision by decision.
Technology may change how we deliver advice. But it doesn’t change why clients seek it: because they still want to talk to someone who understands both the math and the meaning behind their money.
Brady Lochte is a fiduciary financial advisor and the founder of Axon Capital Management.
Guest posts on ImpactAlpha represent the opinions of their authors and do not necessarily reflect the views of ImpactAlpha.
