Artificial intelligence is developing so quickly that it can be hard for business leaders to know where to focus. New tools and capabilities appear constantly, and each seems to arrive with another example of how AI could change the way organizations operate.
I understand why that creates a sense of urgency. No leader wants to ignore a technology that may eventually have a meaningful impact on their industry. At the same time, one idea that stayed with me from the AI Executive Education Program at Stanford was the importance of separating interest in the technology from the business problem an organization is actually trying to address.
During the program, participants discussed the risk of getting distracted by the latest AI capability rather than building the foundation needed to evaluate and implement technology thoughtfully.
That discussion reinforced a fairly simple framework for me: identify the problem, understand the current process, evaluate possible solutions, test them, and then measure what actually changed.
I do not see that framework as an argument for or against AI. It is simply a structured way to approach a rapidly developing technology.
Begin With a Clearly Defined Problem
When evaluating AI, it is tempting to begin by asking what a particular tool can do.
I have found myself asking a different question: What are we actually trying to address?
An organization might identify a process that requires substantial employee time, information that is difficult to organize, repetitive administrative work, or another area where there may be an opportunity to change the existing workflow. Once the issue is clearly defined, leaders can begin evaluating whether technology might be relevant.
That order matters.
Implementing a new technology can introduce considerations that extend well beyond the cost of the software. Depending on the circumstances, firms may need to evaluate employee training, information security, privacy, vendor oversight, recordkeeping, supervision, compliance requirements, and integration with existing systems.
For a registered investment adviser, those considerations can be particularly important because firms operate within regulatory requirements and may handle confidential client information.
As a result, the question is not simply whether an AI application can perform a particular task. Firms also need to determine whether and how a particular application can be used within their own policies, procedures, regulatory obligations, and risk-management framework.
Understand the Existing Process First
Another discussion at Stanford focused on establishing a baseline before introducing AI into a workflow. Participants discussed understanding factors such as the current time, cost, and accuracy associated with a process before evaluating a new approach.
I found that concept useful because it provides a reference point.
If an organization does not know how a process currently performs, it can be difficult to determine whether a new technology has meaningfully changed it.
A baseline will look different depending on the situation. A firm might examine how much employee time a process requires, how many people participate, where delays tend to occur, or what review procedures are currently necessary. Other workflows may require different measures.
The purpose is not necessarily to create an elaborate measurement system. It is to understand the existing process well enough to make a reasonable comparison later.
That can also reveal something else: AI may not always be the appropriate answer.
Sometimes examining a workflow closely exposes an issue that can be addressed through a process change, better use of an existing system, additional training, or another approach. AI may be one possible solution, but its availability does not mean it is appropriate for every problem.
Evaluate the Entire Workflow, Not Just the AI Output
One question I keep coming back to is how firms should evaluate AI when human review remains part of the process.
An AI application might produce an initial output quickly, but that does not necessarily mean the overall workflow has become more efficient. Employees may still need to verify information, correct errors, provide context, or complete additional review.
This is one reason I think firms should consider the entire process rather than measuring only the technology’s speed.
The appropriate level of oversight may also vary significantly by application. An internal administrative task presents different considerations from an activity involving client information or material used in connection with advisory services.
No single framework can determine the appropriate use of AI in every circumstance. Each firm needs to consider its own operations, technology, policies, regulatory obligations, and risk tolerance.
For me, that uncertainty argues for thoughtful evaluation rather than rapid adoption.
Testing Can Reveal What a Demonstration Cannot
The Stanford discussions also addressed testing and iteration. Participants described identifying the problem and desired outcome, developing a possible solution, and then testing and validating it before determining how to proceed.
I think there is an important distinction between seeing what a technology can do in a demonstration and understanding how it performs within an actual organization.
Real workflows tend to be messier.
Information may exist in different formats. Employees may use systems differently. Existing review requirements still apply. A process that looks straightforward in isolation may connect to several other processes that are not immediately obvious.
A limited test can help identify some of those issues before a firm considers broader implementation.
It may also produce an unexpected conclusion. A technology might perform well but require more human review than anticipated. Employees may find that it does not fit naturally into their existing workflow. Alternatively, testing may identify uses that were not initially apparent.
I think it is reasonable to expect some experiments not to proceed beyond that stage. The purpose of testing is to learn, not to prove that a decision already made was correct.
Measurement Should Return to the Original Objective
During the program, participants discussed several ways organizations might evaluate AI initiatives, including time, productivity, employee experience, client experience, accuracy, cost, and return on investment.
Those discussions made me think about how easy it is to reduce AI measurement to hours saved.
Time can certainly be relevant, but it is only one consideration.
If a process becomes faster but requires additional review, that matters. If employees save time but encounter new operational challenges, that matters as well. Depending on the application, firms may also need to consider accuracy, adoption, security, compliance, cost, and other factors.
Client-related measures require particular care. An advisory firm should not assume that introducing AI will produce a particular client outcome simply because a process becomes more efficient.
The appropriate measures will depend on what the firm originally intended to address.
That is why the first step matters so much. If the problem was never clearly defined, determining whether the technology addressed it becomes much more difficult.
AI Adoption Is Also a Leadership Question
I came away from Stanford interested in the possibilities surrounding artificial intelligence, but also aware of how many questions remain.
Those positions can coexist.
RIA leaders can explore new technologies while also recognizing that not every capability will be appropriate for every firm, employee, workflow, or client circumstance. AI tools can also change quickly, which means conclusions reached today may need revisiting as the technology and regulatory environment develop.
For me, the useful discipline is to keep returning to the same questions.
What problem are we trying to address? How does the process work today? What alternatives have we considered? What risks and obligations accompany the proposed technology? How will it be tested? What will we measure? Who remains responsible for reviewing the work?
Those questions are not as exciting as the latest AI demonstration, but they may provide a more practical way for firms to think about adoption.
AI will continue to change. The responsibility for determining where and how it belongs within an advisory business remains with the people leading that business.
Before asking what the newest AI tool can do, it may therefore be worth asking something much simpler: What problem are we trying to solve?
Important Disclosure
This commentary reflects the personal opinions, viewpoints and analyses of the Arden Global Family Offices employees providing such comments, and should not be regarded as a description of advisory services provided by Arden Global Family Offices or performance returns of any Arden Global Family Offices client. The views reflected in the commentary are subject to change at any time without notice. Nothing in this commentary constitutes investment advice, performance data or any recommendation that any particular security, portfolio of securities, transaction or investment strategy is suitable for any specific person. Any mention of a particular security and related performance data is not a recommendation to buy or sell that security. Arden Global Family Offices manages its clients’ accounts using a variety of investment techniques and strategies, which are not necessarily discussed in the commentary. Investments in securities involve the risk of loss. Past performance is no guarantee of future results.
Arden Global Family Offices may not render advice unless a client service agreement is in place.