The short version
- AI is good at the writing around advice: meeting notes, follow-up drafts, summaries and meeting prep. It isn't a source of advice.
- Keep NRIC numbers, medical details, account numbers and policy numbers out of any AI tool your firm hasn't approved.
- You're responsible for every word that reaches a client, whoever drafted it. Read it before it goes.
- In Singapore, the Financial Advisers Act requires a reasonable basis for a recommendation, built on the client's own objectives, finances and needs. An AI answer doesn't give you one.
- In the US, FINRA says its rules apply when firms use generative AI, the same way they apply to any other tool.
Where AI cuts an advisor's admin, and what you still check
Most of the time AI saves is in the writing before and after a client meeting. Here's where it helps, and what you still check:
| Task | What the AI does | What you check before it's used |
|---|---|---|
| Meeting notes | Turns your rough notes or a recording into a clean summary | Every figure, date and promise against what was said |
| Follow-up message | Drafts the email or WhatsApp after a meeting | The tone, the facts, and that nothing sounds like a recommendation you didn't make |
| Meeting prep | Summarises what you already know about the client from your own notes | That it only used your notes, and nothing is missing |
| Explaining a concept | Rewrites how a rider or a waiting period works in plain words | That it's accurate for the actual policy terms |
| Admin | Turns notes into tasks and follow-up dates | That the dates are right |
| Your own week | Summarises your activity and what's overdue | Nothing, it's for you |
| Posts and newsletters | Drafts content for social media or email | Your firm's approval, before anything is published |
The pattern is the same in every row: AI drafts, you check, and you send. The time saved is in the first draft, never in the check.
What to keep away from AI
- Keep NRIC numbers, medical reports, bank and policy numbers out of any chatbot your firm hasn't approved, even to "just tidy up the notes".
- Leave the recommendation to yourself. Which product suits a client is your judgement from their fact-find, and the rule on that is below.
- Take premiums, illustrations and benefit figures from the insurer's own system and documents. An AI answer can be confidently wrong about a number.
- Treat an AI-written post as your own advertisement, because it is one, and the same approval rules apply.
- Read every message before it reaches a client. The draft that goes out automatically with nobody reading it is the one that goes wrong.
- Tell the client before you record a meeting, and stop if they don't want it.
A follow-up drafted with AI and sent on WhatsApp also falls under the Do Not Call rules if it markets something, which the WhatsApp guide for financial advisors in Singapore covers.
What the rules say in Singapore
This is a summary of the official rules, not legal advice. Your firm's compliance team has the final word on which tools you can use.
- The Personal Data Protection Act says an organisation must protect personal data in its possession or under its control by making reasonable security arrangements against unauthorised access, use or disclosure (section 24). Putting client data into a tool with no control over where it goes is the risk this covers.
- The same Act restricts transferring personal data outside Singapore unless it gets protection comparable to the PDPA (section 26). Many AI tools process data overseas, so ask where yours does.
- The PDPC published Advisory Guidelines on Use of Personal Data in Generative AI on 20 July 2026, to be read with its 2024 guidelines on AI recommendation and decision systems.
- The Financial Advisers Act says a licensed financial adviser must not make a recommendation without a reasonable basis, meaning it considered and investigated the recommendation having regard to the client's investment objectives, financial situation and particular needs (section 36).
- MAS consulted from 13 November 2025 to 31 January 2026 on Guidelines on AI Risk Management for all financial institutions. The proposals cover generative AI and AI agents, with expectations on oversight, an inventory of AI use, human oversight and third-party risk, applied in proportion to each firm's size and risk. When we searched mas.gov.sg on 28 September 2026 we found the consultation, closed, and no final version yet, so watch your firm's AI policy for it.
- MAS's Guidelines on Standards of Conduct for Digital Advertising Activities took effect on 25 March 2026. They apply to financial institutions and their digital marketers, including representatives, so an AI-drafted post goes through the same approval as any other.
What the rules say in the US
In Regulatory Notice 24-09 (27 June 2024), FINRA reminded member firms that its rules are technology neutral and apply when firms use generative AI, just as they apply to any other tool. It gives two examples:
- Rule 3110 (Supervision): a firm using AI as part of its supervisory system needs policies that cover technology governance, including model risk management, data privacy and integrity, and the reliability and accuracy of the AI model.
- Rule 2210 (Communications with the Public): the content standards apply whether a communication was written by a person or generated by a technology tool.
FINRA also says firms should evaluate an AI tool before they deploy it. That includes AI built into third-party products they already use.
Before you use an AI tool with client information
Ask these, and get the answers from your firm or the vendor in writing:
- Has your firm approved it for client information?
- Where is the data processed and stored, and in which country?
- Is what you type kept, and is it used to train the model?
- Can you delete a client's data when you need to?
- Who else can see it, at the vendor and at your firm?
- Does it keep the record your firm needs of what was sent to a client?
If you can't get a clear answer to the first two, use the tool only for work with no client information in it.
Prompts you can use with no client details in them
- "Rewrite this explanation of a waiting period in plain English for someone with no finance background: [your own explanation]."
- "Turn these meeting notes into a short summary with action items and dates: [notes with names replaced by Client A]."
- "Draft a friendly WhatsApp to confirm a meeting on [day] at [time] at [place], under 40 words."
- "Give me five questions to ask a new client about their goals for the next ten years."
- "Here are the objections I heard this week: [list]. Suggest a calm, honest reply to each."
Swap names for "Client A" and leave out anything that identifies the person. The draft is as good without them. For the messages you send every week, the WhatsApp message templates for financial advisors need no AI at all.
AI mistakes advisors make
- Treating a confident answer as a correct one, especially about product terms or tax.
- Pasting a whole fact-find into a free chatbot to get a summary of it.
- Letting a note-taker's summary stand as the record of the meeting when nobody has read it.
- Posting AI-written content before your firm has approved it.
- Asking AI to write a recommendation, then signing it as your own.
Ask about your week, get the follow-up drafted
This guide is free. ActivityTracker, the app we make, is paid. Its assistant, Ask AI, answers from your own work, and the rules above still apply to what you put into it.
- Ask AI answers from your own tasks, goals, calendar, notes, documents and logged activity
- Tell it what a client asked for and it drafts the follow-up for you to tidy and send
- Buttons under an answer add the steps as tasks, with Undo, or set them as a goal
- For bigger jobs it lists every change first, and makes them only after you tap Confirm