Use of Technology AI in practice: Answers to advisor questions about artificial intelligence

“For smaller firms, the key point is that AI adoption does not require a massive technology budget,” says Sutton. “A firm can start with a secure business-grade tool, give access to a small number of team members and focus on a few high-value use cases before expanding.” One helpful approach may be to:

1. Start with one tool (e.g., Copilot, Gemini, Claude or another secure enterprise AI)
2. Identify three to five high-impact use cases (emails, meeting prep, marketing)
3. Build repeatable processes using prompts and templates
4. Document compliance guidelines early
5. Scale gradually as comfort increases

AI pricing may evolve over time. Today, much of the pricing is tied to seats and usage. In the future, some AI providers may begin pricing more directly around outcomes — in other words, charging based on the value the tool helps create. That could better align AI vendors with the businesses using their technology because the cost would be connected more closely to the results firms are trying to achieve.

There are also consulting firms and technology vendors that specialize in assessing and implementing AI solutions for RIAs, which may be an option for advisors feeling out of their depth. Keep in mind, even a small amount of AI knowledge can be powerful, and many firms can make meaningful progress without third-party advice or a pricey tech stack.

Insight for advisors: Smaller firms can use business-grade AI tools today. Start with a right-sized subscription, monitor usage and focus first on workflows where the tool can save time or create measurable business value.

Most major AI platforms offer multiple subscription tiers. Larger enterprise plans may be designed for organizations with a minimum number of users, while business or team-level licenses are often available for smaller firms that do not need a full enterprise deployment.

In general, pricing tends to have two components:

1. Per-seat subscription fees: Many tools charge a monthly fee per user. Pricing varies by provider and tier, but the model is often similar to other business software: The firm pays based on the number of people who need access.
2. Usage-based or compute-based costs: Some platforms also include usage limits. If a firm uses the tool heavily — for example, across many workflows or with frequent requests — the company may eventually need to pay for additional usage or expanded access. In other cases, the tool may restrict usage once a certain threshold is reached.

Insight for advisors: A firm’s size shouldn’t stop it from utilizing AI tools. There are a variety of pricing tiers that provide enough flexibility that almost any sized practice should be able to implement some amount of AI.

“Scaling AI should be intentional,” says Sutton. The goal is not to roll out AI everywhere at once. The goal is to build comfort, identify the highest value use cases and expand in a way that creates measurable value.

It’s helpful to think about scaling in three stages.

1. Build broad comfort with the tool

“Before trying to redesign workflows, make sure the team is comfortable using the AI tool the firm has selected,” says Sutton.

The first milestone is getting the team to use AI reactively. In other words, the tool should be open and available during the workday, and team members should begin asking themselves: Could I use AI to help with this?

That is an important behavioral shift. It means the team is beginning to see AI as a practical assistant rather than a separate technology project.

At this stage, teams should focus on:

• Understanding how to use the tool
• Learning how to ask better questions
• Practicing with low-risk tasks
• Seeing where AI is helpful and where it is not
• Building confidence through repetition

The objective is not perfection. The objective is familiarity.

2. Move to top-down implementation

Once the team is broadly comfortable, the firm can begin implementing AI more strategically.

This is where the five-step process becomes important:

1. Prioritize the right workflows
2. Document the current process
3. Map AI tools to the process
4. Implement and test
5. Scale and measure

Leadership should identify one to three priority areas for the next quarter or the next six months. These should be workflows where AI can either save meaningful time or create meaningful business value.

Examples might include:

• Client review preparation
• Meeting follow-ups
• Prospecting support
• Marketing content creation
• Data analysis
• Internal research and summarization

The key is to avoid trying to implement AI across every workflow at once. Focus on a small number of high-value use cases, prove the value and then expand.

3. Get to value quickly

Early AI implementation should be practical and measurable. Firms should ask:

• Where can we save the most time?
• Where can we improve consistency?
• Where can we improve the client or team experience?
• Which workflows are repetitive enough to benefit from AI?
• What can we test within the next quarter?

The faster the team sees value, the more likely adoption will continue.

Insight for advisors: Start by helping the team become comfortable using AI in everyday work. Then shift to a top-down approach by selecting one to three priority workflows, applying the five-step process and measuring the value created before expanding further.

The first step is to speak directly with your compliance department. AI policies, approval processes and acceptable-use guidelines will vary by firm, so advisors should not assume there is a one-size-fits-all answer.

Firms should work with their own compliance teams and, where appropriate, speak directly with their AI vendors to understand how the tools handle data, security, retention, oversight and auditability.

That said, there are several practical guardrails many teams are beginning to build into their AI processes.

1. Involve compliance early

Compliance should be part of the process from the beginning, not brought in after workflows are already built. Advisors should clarify:

• Which AI tools are approved for use
• What types of data can and cannot be entered
• Which outputs require review
• How records should be retained
• Whether AI-generated content should be treated like other drafted communications or materials

This helps the firm innovate while staying within its own compliance framework.

2. Build checks and balances into the workflow

AI-enabled workflows should include a clear process for review and validation.

For example, if a tool is being used to support financial planning-related analysis or client-facing output, the firm may want the tool to explain the logic it used. That explanation can then be captured in a log or retained as part of the process, so the team can review why an output was created, how it was created and whether the reasoning is sound.

More teams are beginning to think this way: not just asking AI for an answer, but asking it to explain the steps behind the answer so those steps can be reviewed.

3. Use AI to evaluate its own work — but do not stop there

AI can be useful for reviewing, scoring or checking its own output. For example, teams can ask the tool to identify potential issues, inconsistencies, missing context or areas that require human review.

However, AI evaluation should not replace human oversight.

AI is a tool that can be used across the business, but the human remains accountable. Advisors and their teams need to remain the final reviewers, especially when the output may influence client communications, planning assumptions or business decisions.

4. Keep a human in the loop

The most important guardrail is human accountability. AI can draft, summarize, analyze and recommend, but it should not operate without oversight in areas that require professional judgment.

A strong AI process should make clear:

• Who reviews the output
• What standard the output is reviewed against
• What happens when the AI output is incomplete, inaccurate or inappropriate
• How the final decision or communication is approved

Insight for advisors: Start with your compliance department and vendor. Then build AI workflows that include documented logic, review steps, output logs and human accountability. AI can support the process, but the advisor and firm remain responsible for how it is used.

RIAs are approaching personally identifiable information, or PII, in different ways. There is a spectrum.

Some firms are not comfortable putting PII into AI tools at all. Others have become more comfortable using enterprise-grade tools with appropriate data protections, vendor agreements and internal policies.

While more firms are becoming comfortable with the PII question, many firms are adopting a “no PII in AI” rule, unless the tool is explicitly approved and secured. The reality is that the approach is not uniform. Each firm needs to make its own decision based on its compliance requirements, technology stack, vendor due diligence and risk tolerance.

“Every firm is handling PII differently,” explains Angelo. “The firms I have engaged are taking PII seriously and moving methodically in terms of when/where/if they are exposing such information. There are steps to be taken in order to secure your data lake while also ensuring that the AI tools you are using are done at an enterprise level and not in an open environment.”

“One pattern we are seeing is that firms that are most focused on security and integration may lean toward Microsoft-based tools,” adds Sutton. “Firms that are more focused on leading-edge AI capability may also evaluate tools from providers such as OpenAI, Anthropic or Google.”

Insight for advisors: There is no single industry standard yet. Firms should work with compliance and technology partners to determine what data can be used, which tools are approved and what safeguards are required.

SEC Regulation S-P requires firms to protect customer information and ensure proper safeguards. When using AI, this translates to:

• Maintaining confidentiality of client data
• Vetting vendors for security controls and data usage policies
• Documenting how AI is used within the firm

Insight for advisors: Regulators seem less focused on banning AI and more focused on how firms use it responsibly.

AI does not substitute for compliance review.

However, AI can be useful as a pre-compliance filter. “Advisors can use AI to review materials before submitting them to compliance by asking the tool to look for issues, flag risky language, check against firm guidelines or suggest required disclosures,” advises Sutton.

This can help improve the quality of materials before they reach compliance, potentially reducing edits and speeding up the review process.

That said, the AI review is not the final review. Compliance teams still need to apply the firm's actual standards and approval process.

Insight for advisors: AI can help prepare materials for compliance review, but it does not replace the compliance function.

With oversight, yes, AI can be used for prospecting in a way that’s compliant. AI can help draft outreach messages and segment audiences, but all communications must still adhere to marketing and advertising rules. Firms should consider:

• Review AI-generated content before sending
• Ensure disclosures and claims are accurate
• Maintain records of communications

Insight for advisors: “AI cannot replace human oversight, and it’s important for a human to always be involved in every process,” says Sutton.

The issue is less about avoiding specific prompts and more about making sure the output is properly reviewed.

Advisors should be thoughtful about what information they provide to an AI tool, especially when it involves client data, financial planning assumptions or regulated communications. But the bigger discipline is validation.

A strong process should include:

• Asking AI to review and critique its own output
• Having a human review the output before it is used
• Checking facts, calculations and assumptions
• Making sure the output aligns with firm policies and compliance requirements

Insight for advisors: The most important safeguard is not a list of banned prompts. It is a strong review process that includes both AI-based checks and human judgment.

An AI hallucination is when an AI system (like ChatGPT, Copilot or other large language models) generates information that sounds correct and convincing but is actually false, misleading or made up. These errors happen because AI is designed to generate what sounds right, not necessarily what is right. When it’s unsure, it may confidently fill in the gaps or invent details — so the result can include plausible-sounding facts, sources or explanations that still need to be double-checked by a human.

The most important safeguard is keeping a human in the loop.

AI-generated outputs should be reviewed before they are used, especially if they will influence client communications, investment discussions, planning assumptions or business decisions.

“We are not in a place where we can simply implement AI, ask it to execute and walk away,” explains Angelo. “Like anything, there is a process to educating the AI system one is using to ensure the output is accurate. Adding a check/review/read-through on the backend is required for most process-oriented or written efforts with AI large language models. As time goes by, you will see improvements in accuracy, which will breed more confidence in the deliverable’s accuracy.”

Another strong practice is to use a separate AI evaluation step. In this approach, one AI workflow creates the initial output, and another AI workflow reviews it. The evaluation step can check the reasoning, compare the output against source material, look for unsupported claims and flag areas that require human review.

This creates an additional quality-control layer before a human makes the final decision.

Insight for advisors: Use AI to help review AI, but do not stop there. Human review remains essential.

AI can reflect bias, but in many practical situations, the bigger risk is that it mirrors the bias in the user's prompt. 

If a user frames a question narrowly, leads the tool toward a preferred answer or only provides one side of an issue, the AI may reinforce that direction. That is why advisors should use AI to challenge their thinking, not just confirm it.

Useful prompts include:

• What are the counterarguments?
• Why might this not work?
• What assumptions am I making?
• What would a skeptical reviewer say?
• Please poke holes in this argument.
• What evidence would change this conclusion?

This is sometimes described as red teaming or steelmanning. The goal is to use AI as a challenging reviewer that helps identify weaknesses, alternative perspectives and blind spots.

Insight for advisors: AI can reinforce confirmation bias if used poorly. Advisors should deliberately ask AI to challenge their assumptions and identify the strongest opposing arguments.

AI should not replace professional judgement, knowledge and experience. Treat it as a tool for ideation and efficiency — not as a source of authoritative advice. Always review and validate outputs before incorporating them into client deliverables.

Insight for advisors: It is important to not only tell the tool what to do, but also how to do it. Be specific about the type of quantitative analysis it is supposed to do for each step and have a separate agent grade the work.

One advisor asked if we could “wave a magic wand,” how would advisors be using AI effectively? There’s no one answer to that question, because it depends on your business and its goals.

“If we could wave a magic wand, advisors would be using AI to replace some of the remedial and low-value but necessary tasks for an advisor team,” advises Angelo. “A few examples would be accurate (client meeting) note-taking, notes logged in customer relationship management (CRM) software in a timely fashion, and actionable summaries created that are ‘client ready.’ One other area of low-hanging fruit for AI implementation tied to efficiency is investment research. Saving time and improving accuracy of information can be achieved in these two areas if we wave the proper AI wand!”

“Advisors should start by understanding some of the spaces where AI is strongest today, in places like coding, research and data analysis,” suggests Sutton. “AI is not a one-for-one replica of human intelligence. It has different strengths and weaknesses, and those will continue to evolve as the technology improves. Practices may want to consider using AI to reduce administrative friction and enhance client communication.” Today, that includes:

• Drafting client emails, meeting summaries and follow-ups
• Generating marketing content (blogs, newsletters, social posts)
• Preparing for client meetings with quick research and agenda creation
• Summarizing long documents such as earnings reports or tax updates

Adoption is expected to expand into more integrated workflows — automated CRM updates, deeper personalization of client communication, and early-stage AI “agents” that can handle recurring tasks, like onboarding or meeting prep.

Insight for advisors: Start with areas where AI already has clear strengths: research and data analysis. These are often practical, high-value places to begin.

AI agents are systems designed to perform specific tasks with minimal supervision — think of them as digital assistants that follow defined workflows. In an advisory context, that could include:

• An onboarding agent that gathers client data and drafts initial plans
• A service agent that prepares quarterly review materials
• A marketing agent that generates and schedules content
• An agent that does meeting preparation, drafts follow-ups and helps prepare materials

For advisors, the opportunity is to think about repeatable tasks where an associate might normally be involved. Customization is key. And like us, AI agents get better with practice. Consider starting with clearly defined tasks, ensure compliance oversight and prioritize human review.

Insight for advisors: AI agents are moving from answering questions to completing tasks. Advisors should watch this space closely, but apply it first to clearly defined workflows with human oversight.

Practical learning may be more effective than theoretical study. Advisors can build fluency through industry articles and webinars (like Capital Group’s recent AI webinar), hands-on experimentation and peer-to-peer sharing. “However, those looking to take their education a step further should consider the education resources from OpenAI and Anthropic,” suggests Sutton.

“Both companies offer training content and practical examples for users at different levels, from beginners to more advanced business users,” says Sutton. These resources can help advisors understand how to use AI more effectively, how to write better prompts and how to apply the tools to real workflows.

“Advisors do not need to become technologists,” Sutton adds. “The goal is to become fluent enough to understand what AI can do, where it fits in the business and how to use it safely and productively.”

Insight for advisors: Start with practical training from leading AI providers, then apply what you learn to real business workflows.

For portfolio analysis, it is important to distinguish between dedicated portfolio analytics systems and general AI tools.

AI should not be viewed as a replacement for trusted portfolio analytics software. However, general AI tools can be useful for supporting research, summarizing data, identifying themes, creating client-friendly explanations and helping advisors think through analytical questions.

From a general AI capability perspective, OpenAI and Anthropic are two of the current leaders to watch, says Sutton, particularly because of the broader capabilities around data analysis, reasoning and tool use. Tools such as Codex and Claude Code also show how important the surrounding harness is becoming.

The model itself is only part of the equation. The harness — the software environment, tools, instructions, integrations and workflow around the model — increasingly determines how useful the system is. If the model is the engine, the harness is the car around it.

Insight for advisors: There is no single AI platform that replaces portfolio analytics tools. Advisors should use trusted portfolio systems for core analysis, and use AI to augment research, summarization, data interpretation and client communication.

AI can potentially help across almost any part of a business.

While many advisor conversations focus on client communication or practice management, AI can also support areas such as institutional work, retirement plans, internal operations, marketing, research, reporting and business development.

The important question is not: Where can AI help? The better question is: Where do we have repeatable work, meaningful time spent or high-value decisions that could be improved with better research, analysis or communication?

Insight for advisors: AI is broadly applicable. The best use cases usually appear where there is repetitive work, heavy research, data analysis or a need to communicate complex information more clearly.

Given the rapid pace of innovation, flexibility is an advantage. 

One trend we are seeing is that some firms are moving toward a multi-platform approach. Rather than relying on only one provider, firms may use different AI tools for different purposes or keep multiple options available as the technology evolves.

So far, many AI workflows have been relatively portable. Prompts, use cases and process designs can often be adapted from one platform to another without starting over completely. 

Insight for advisors: Avoid overcommitting too early. The AI landscape is changing quickly, and flexibility may be more valuable than locking in to one provider.

AI is transforming asset management, but the most important foundation for young professionals remains strong financial and market knowledge. Understanding valuation, portfolio construction, macroeconomic trends and risk is essential. AI tools can enhance analysis, but they don’t replace judgment — professionals who can interpret results and identify when something doesn’t make sense will have a clear edge.

At the same time, technical and data literacy is becoming increasingly important. Basic proficiency in tools like Python, SQL and data analysis frameworks allows professionals to work effectively with large datasets and AI-driven models. You don’t need to be a full-scale data scientist, but understanding how models are built, tested and used in practice is critical for contributing meaningfully in modern investment teams.

Equally important is the ability to think critically and communicate clearly. AI produces vast amounts of information, so success depends on filtering signal from noise and forming coherent, evidence-based investment views. Being able to explain complex, data-driven insights in simple, compelling terms — whether in memos, presentations or discussions — is a key differentiator.

Finally, adaptability and curiosity are essential traits. The tools and techniques shaping investing will continue to evolve rapidly. Professionals who embrace continuous learning, stay informed about new technologies and bridge finance with broader disciplines (like technology and geopolitics) will be best positioned to stand out and succeed.

Insight for advisors: AI is transforming quickly, but it remains important to use the technology, understand its strengths and apply it there. At the same time, understand AI weaknesses and gaps. That is where you should invest in career/team/firm development as those gaps become more valuable.

A useful way to think about AI and careers is to break business activity into three broad functions:

1. Building and improving the product or service: This includes innovation, advice delivery, client experience and the core value proposition.
2. Selling and distributing the product or service: This includes business development, client acquisition, marketing, relationship management and getting the service into clients' hands.
3. Measurement and operations: This includes many of the back-office, administrative, reporting, measurement and process-oriented functions that help the organization run.

AI appears to be more complementary in the first two areas. It can help people build better services, improve client experiences, communicate more effectively and develop new business. In those areas, AI may make strong professionals more productive and valuable.

The third area may experience more automation and replacement over time. There will still be people involved, but more routine operational, measurement and administrative tasks may be handled by AI-enabled systems.

For advisors, this suggests that relationship skills, judgment, client trust, business development and the ability to apply technology thoughtfully will become even more important.

Insight for advisors: AI is unlikely to remove the need for advisors. But it will change where human value shows up. The advisor of the future will likely spend less time on routine tasks and more time on relationship-building, judgment, business growth and differentiated client service.