For financial advisors, AI has moved past the point of being just another valuable tool that can increase efficiency. It is now an essential tool that empowers advisors to keep pace with the elevated expectations in today’s marketplace.
According to the findings of a survey conducted by InspereX in early 2026, 78% of advisors see AI as a tool that gives firms a competitive advantage, especially when trying to attract and retain next-generation clients. Nearly two-thirds of advisors said AI is giving smaller firms the capability to compete with “much larger firms.”
But the survey’s findings also reveal the limits of AI in financial advisory transactions. It shows that advisors are primarily using AI to drive higher levels of efficiency in their practices by offloading noncritical work to the technology. This is especially valuable for advisors with smaller offices, as it allows them to run at the same capacity as those backed by teams of practice managers and other administrative staff. The less time they need to spend on these non-mission-critical items translates into more time they can devote to high-value tasks like client meetings and relationship management.
“Research and insights” was identified as the top task assigned to AI, with 74% of advisors reporting they used the technology in this area. The next highest uses included “client communications and follow-up” (73%), “meeting preparation and documentation” (72%), “planning and proposal drafting” (56%), “marketing and content creation” (52%), and finally “back-office tasks and operations” (43%). When asked what they would “never” delegate to AI, 26% of advisors gave “product selection” as their top response, followed by 23% saying “portfolio recommendations” and 20% saying “suitability-related analysis.” What’s worrying, though, and something advisors who use AI should be asking themselves, is “how much overlap is there between ‘research and insights’ and ‘product selection’?”
Overall, the survey reveals that advisors feel AI can be used as a value-add tool, but not as a replacement for actual judgment. It can help advisors to move faster, consider more data, and identify more patterns, but outsourcing the full thought process involved in advising is a dangerous practice. With how confident most of today’s models sound, it’ll be important going forward that advisors recognize when the AI tries to replace decision-making tasks and stop it early on.
To avoid pushing AI past its limits, advisors need to use it for preliminary guidance. It can bring clarity in an advising situation involving internal documents that are vague or components that would normally require a check-in with a specialist. But it shouldn’t be seen as the subject matter expert. Rather, it is the tool that gets projects to a better starting point before passing material on to a human advisor who owns the topic.
See also: AI Infrastructure Services: Banking and Financial Services’ New Foundation for Computing
AI without oversight brings the risk of breaching confidentiality
Confidentiality is a core concern in financial advising. Putting client information into third-party AI tools threatens that confidentiality. When sensitive data is fed into a tool like ChatGPT, it can later be used for the tool’s training and ultimately appear in the content it provides to other users.
For example, a busy advisor preparing for a meeting who needs to review a lengthy and complex corporate buy-sell agreement provided by the client might be tempted to copy and paste the document into Claude and ask it for a one-page summary of the beneficiaries and trustee powers in the document. What that advisor may not understand or appreciate is that they have just provided a variety of non-public personal information, personally identifiable information, and/or personal financial information (NPI, PII, and PFI) — including potentially legal names and addresses, bank account details, and more — to a privately owned company that’s free to do whatever it wants with it. Not only could this be a severe violation of an advisor’s third-party disclosure requirements, but it could also constitute various regulatory, state, and federal privacy infractions in addition to violating the trust their clients have placed in them
Using AI to increase efficiency in this way could constitute serious confidentiality and Regulation S-P violations, which can bring penalties as severe as barring an advisor from the industry altogether. Essentially, it puts client information on a server beyond the firm’s control, taking away its ability to understand, explain, or ensure how that information is stored, processed, and secured. That type of behavior threatens both transparency and security.
Overall, the challenge in using AI without overusing it is establishing a healthy balance between convenience and security. ChatGPT, Claude, and similar public tools are fast, easy to use, and grow more capable every day. For an advisor who is under the gun, the risks associated with using those tools can feel abstract while the benefits are immediate. But failing to consider the consequences can lead to high-risk behavior.
One way to balance security and convenience is to use locally hosted AI models or internally controlled AI deployments. This approach keeps sensitive data under the firm’s control and discourages advisors from using tools that pose risks. While it doesn’t eliminate every risk, it reduces them significantly and can almost entirely mitigate any confidentiality concerns if done properly.
Oversight is needed to keep AI hallucinations from misleading advisors
Generative AI isn’t generally designed to provide responses that come across as suspect. When asked to provide an answer or perform a task, it presents results with certainty. The problem, however, is that AI’s confidence can conceal inaccuracies in its responses that are commonly known as “AI hallucinations.”
The danger of AI hallucinations should be obvious to advisors. They can cause a misinterpretation of laws and policies, corrupt a suitability analysis, or errantly shift client recommendations, to name just a few problems that can stem from building upon the wrong information. Even worse, because they can appear within a stream of information that is otherwise correct, errors can be easily missed and only discovered after they’ve already caused a problem.
Advisors should also understand that AI hallucinations are often more than just misstatements or misrepresentations. They can also be a total fabrication. A growing number of attorneys, financial advisors, doctors, and other individuals in professional fields have faced legal sanctions for submitting AI-generated documents. One early example is the attorney who was unfortunate enough to end up citing case law that doesn’t exist. In that instance, the court was not at all satisfied with the explanation that he had misunderstood how these tools function. Financial advisors should be ready for generative AI to cite tax law and other “official documents” that don’t exist.
With that type of risk, the only defense is to ensure that all content is carefully reviewed by someone who can distinguish fact from AI-generated fiction. A good rule of thumb is to treat AI-generated content as draft material. Converting it to the final product requires subjecting it to close inspection by an expert who doesn’t have a history of hallucinating.
To gain the time savings AI can deliver, advisors need to manage the risk. Regardless of whether it is being used for research, client communication, meeting prep, or document drafting, AI’s output should be seen as a first draft that will always need a human-guided revision. Those who use it as a “set it and forget it” tool and don’t subject it to careful cross-checking will inevitably suffer from a hallucination or overconfident mistake. Those who reinvest the time they save in higher-quality human judgment of AI output will gain the benefits without the regulatory and reputational risks that can result from a lack of precaution.
