40 Questions Financial Advisers Should Ask Before Trusting AI Recommendations
When Aladdin suggests a sector rotation or ChatGPT drafts client communication, you need to know whether you are following sound reasoning or outsourcing your thinking. These questions help you stay in control of your advice and protect your relationship with clients.
These are suggestions. Use the ones that fit your situation.
1When Morningstar AI flags a fund for removal from the portfolio, what specific metric changed and over what time period did it decline?
2If Aladdin recommends increasing emerging market exposure, what is the confidence level of the currency forecasts underlying that suggestion?
3Does the AI tool show you which holdings it excluded from its analysis, and why?
4When Bloomberg AI suggests a bond duration adjustment, is that recommendation based on current yield curves or on a predictive model of rate movements?
5Can you identify which data feeds the AI is drawing from, and when those feeds were last updated?
6If the AI ranks funds by past performance, what lookback period is it using and does that match your client's investment timeline?
7Does the recommendation account for your client's existing tax position, or is it a generic suggestion applied to all portfolios in that asset class?
8When Vanguard AI suggests a portfolio rebalance, is it optimising for the goals you set or for a default objective in the software?
9Has the AI tested its recommendation against historical periods when that strategy would have underperformed?
10If you disagree with the AI's analysis, can you export the underlying data to review it independently or does the tool lock you into its conclusions?
Fiduciary Responsibility and Compliance
11If a client disputes an investment decision later, can you document that you personally evaluated the AI recommendation before acting on it?
12Does your firm's compliance team know which AI tools you use for investment advice and have they approved your process for validating outputs?
13When ChatGPT generates explanation text for why you recommended a particular fund, have you fact-checked the statements it made about that fund's strategy?
14If the AI tool makes an error in suitability assessment, who is liable: the software vendor, your firm or you personally?
15Are you able to produce a clear audit trail showing which recommendations came from the AI and which were your own override?
16Does the AI disclose its data sources and methodologies in a way that would satisfy a regulator or ombudsman asking why you followed its advice?
17If you use AI to analyse suitability, are you still speaking directly with the client about their circumstances or relying on outdated profile information?
18When Aladdin produces a portfolio analysis, does it flag assumptions about inflation, growth or volatility that differ from your firm's view?
19Have you checked whether the AI tool's model was trained on a time period that includes market conditions relevant to your clients' situations today?
20Does your engagement letter tell clients that AI is involved in producing their investment recommendations?
Client Communication and Relationship Risk
21When you send a client email drafted by ChatGPT, can you articulate the investment reasoning in your own words if they ask a follow-up question?
22Does the AI-generated communication reflect your firm's voice and values or does it sound generic enough that the client could get similar advice elsewhere?
23If you use AI to write portfolio commentary, are you explaining the 'why' behind market movements or just describing what happened?
24When the AI suggests which clients to contact about a market opportunity, are you reaching out because they genuinely need the advice or because the algorithm flagged them?
25If a client asks why you made a particular recommendation, will they accept 'the model suggested it' as a sufficient answer?
26Are you using AI to reduce the number of one-to-one conversations with clients, and if so, are you noticing them asking fewer questions in reviews?
27When Morningstar AI suggests personalised fund recommendations for a client, do you know how recent their goals and risk profile data is in the system?
28If you draft a suitability letter using AI assistance, have you checked that it accurately reflects what you actually discussed with that specific client?
29Does the client know that an algorithm had input into their recommendation, or would they be surprised to learn that?
30When you rely on AI for quarterly client updates, are you still building the interpretive skill that clients pay you for?
Your Own Judgement and Expertise
31Before accepting an AI recommendation, can you identify what information or market view the AI does not have access to that you do?
32When you choose to override an AI suggestion, can you explain your reasoning without referring back to the AI's reasoning?
33Are you still regularly reading investment research directly, or have you stopped because the AI summarises it for you?
34If the AI tool stopped working tomorrow, how much of your investment process would break?
35Can you remember the last time you changed your mind about a client's strategy based on your own analysis rather than a tool recommendation?
36When you analyse a portfolio, are you asking your own questions first or are you waiting for the AI to flag issues?
37Do you understand the statistical methods that Aladdin or Bloomberg AI is using, or are you trusting the output on authority?
38If a client asked you to explain a recommendation without mentioning any AI tools, could you do it?
39Are you maintaining the skill to challenge an AI recommendation or are you becoming dependent on it being right?
40When you tell a junior adviser to check something, are you asking them to verify the AI's output or just to review what the AI produced?
How to use these questions
Test the AI tool on a decision you have already made independently. Does it reach the same conclusion and for the same reasons? If not, which source are you more confident in and why?
Ask your compliance team to review a recent recommendation that used AI support. Can they clearly see where your judgement was applied to the output?
Write down your investment thesis for a client before running it through the AI tool. If the AI suggests something different, investigate whether it found something you missed or whether it is using different assumptions.
Schedule a quarterly conversation with a client in which you explain one recent recommendation without referring to any software output. Notice what you remember and what you have to look up.
Document one specific instance each month where you disagreed with an AI recommendation and explain what you did instead. This becomes your evidence of active management.