Five admin jobs are worth trying first in a small wealth management firm: meeting notes, follow-up emails, annual review packs, onboarding checklists and newsletter drafts. Start with the ones that involve no personal advice, review every output before it is used or sent, and decide with your compliance lead which AI-generated material counts as part of the client record.
None of these jobs asks the AI to advise anyone. Each takes a repetitive piece of preparation or writing and produces a draft, and each has a person who reads it. The examples below follow a firm of about eight people looking after 220 households, with one lead adviser, two further advisers, two paraprofessionals, one operations manager and two administrators. The time figures are illustrative, so measure your own.
The firm and the week we are trying to shrink
Take the 220 households. Each gets two meetings a year, so roughly 440 meetings, plus ad hoc calls. After each meeting an adviser writes a note and a follow-up email, which takes about 35 minutes together. Before each annual review the team assembles a pack: portfolio summary, changes since last year, open actions and agenda. That takes about 60 minutes. New clients arrive at a rate of three or four a month and need an onboarding checklist chased. Twice a quarter someone writes a client newsletter.
Do the sum for the first two: 440 meetings at 35 minutes is about 257 hours a year, and 220 review packs at 60 minutes is 220 hours. That is nearly 480 hours of admin, equal to about three months of one person's working time, before anything is automated. If AI cuts even a third without lowering quality, the saving is real. The point of the examples is to test whether it does.
Example 1: Meeting notes that an adviser approves
The note is the job with the biggest payoff and the biggest care. Specialist tools for advisers record or ingest the meeting and draft a note; the tutorials on Jump, Zocks and Wealthbox and on the Wealthbox notetaker against standalone options compare them. You can also paste a transcript into a business-plan assistant. Either way, tell the client you are recording and why, and use a tool whose data terms you have read.
A 45-minute review, timed
Say an adviser has a 45-minute review with a client. Writing the note by hand takes 20 minutes, and 10 more for the follow-up email. With a notetaker, the draft appears within minutes. The adviser spends 8 minutes correcting it. The email draft, built from the note, takes another 4. The total is 12 minutes against 30, saving 18 minutes per meeting. Across 440 meetings that is about 132 hours, if the quality holds.
A sample structure to ask for, with the contents made up for illustration:
Meeting note - annual review - [client first names]
Attendees: [adviser], [client 1], [client 2]. Duration 45 minutes.
Client circumstances discussed: change of role, second child due in June, wants to review emergency fund.
Client stated objectives: keep current risk approach; clarify school-fees plan.
Information given to client: explained the cash holding and the review schedule (no recommendation made).
Actions: adviser to send fee summary by Friday; client to send new payslip; follow-up call in 6 weeks.
Items needing adviser attention: client said "I might need access to some of it next year" - clarify.The adviser checks that every statement is attributed correctly, that no advice appears that was not given, and that any change in circumstances or objectives is recorded in the client's own words. The line "Items needing adviser attention" is where the tool earns its keep, flagging a half-heard comment that matters.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowThe recordkeeping caveat
Decide, in writing, what is kept: recording, transcript, AI draft, approved note. Rules on records and on what must be retained, and for how long, differ by firm type and place, so ask your compliance lead. A common position is that the adviser-approved note is the record, the recording follows a defined retention period, and AI drafts are deleted after approval, but that is a decision for your firm, not a default. Whatever you choose, check that the tool stores data where you expect, that you can delete or export it, and that leaving the vendor does not strand your records.
Example 2: Follow-up emails that match the note
The follow-up is where drafts can drift from what was agreed. Generate it from the approved note, not from the raw transcript, so the email states only what the adviser has already checked. The prompt is short:
Write a follow-up email from [adviser name] to [client first names] using only the approved meeting note below.
Include: a thank-you line, a bulleted list of agreed actions with owners and dates, and the date of the next contact.
Do not add any recommendation, projection, product suggestion or opinion. Under 150 words. Plain English.
[approved note pasted here]An illustrative result thanks the clients for their time, lists three actions (fee summary by Friday, payslip from the client, call in six weeks) and confirms the next contact. The adviser confirms every date and owner, and checks that the sentence about the emergency fund says only that it was discussed. A realistic slip: the draft writes "as we agreed, we will move $10,000 to cash" when the note said the client would think about it. Catch it by reading each action against the note, and by banning verbs such as "agreed" unless the note uses them.
If you keep a library of approved wording for common follow-ups, such as "documents we need for onboarding" or "how to access your portal", the assistant can personalise those and your team reviews less. That is the same approach as a template folder, only faster.
Example 3: Review packs assembled from facts you supply
An annual review pack gathers a lot of material. The assistant does not generate portfolio figures, and you should never let it; your reporting system does that. What it can do is arrange the facts, last year's actions and the client's stated goals into a one-page agenda and a set of questions to ask. The tutorial on how advisers use AI to prepare for annual client reviews goes into the full pack, and this is the lean version.
One household's review pack
Say a household's file shows: goals (retire at 62, fund a deposit for a child in four years), last year's actions (three of four completed, the pension consolidation outstanding), changes since last review (a new job and a pay rise), and the figures from your reporting tool for the portfolio's value and the contributions made. You paste those facts, with the household's name removed, and ask for an agenda and questions.
Using only the facts below, produce a one-page review agenda for an adviser meeting:
1) items from last year still open; 2) changes in circumstances to confirm; 3) topics the client has said matter to them; 4) five questions for the adviser to ask.
Do not comment on performance, suitability or what the client should do. Flag anything missing from the facts that an adviser would normally need.
Facts: [pasted]The sample output lists the outstanding pension consolidation first, asks whether the new job changes pension contributions, notes the deposit goal as a four-year horizon and flags that the file contains no recent expenditure figures. The adviser reads it as a prompt, not a plan. Time per pack falls from about 60 minutes to about 25, mostly spent checking the figures from the reporting tool against the agenda.
The checks: every number comes from your reporting system, nothing is described as performance or a recommendation, and the missing-information list is acted on before the meeting. A pack that looks complete but omits the pension is worse than a messy one, which is why the open-items section comes from last year's action list and not from the assistant's memory.
Example 4: Onboarding checklists and chasers
New-client onboarding is a sequence of documents and checks. Your firm already has a list; the job is to turn it into a personalised checklist and chase what is outstanding. Identity checks and the decisions made on them are covered separately in which KYC checks stay human, and the same rule applies here: the assistant organises and reminds, the person verifies.
Worked example: a new client in week one
A couple becomes a client on Monday. Your checklist has fourteen items: signed terms, fact-find questionnaire, identity documents, proof of address, source-of-funds evidence, existing pension statements, bank details for fees, and so on. By Thursday six have arrived. The administrator pastes the list and what has been received (item names only) and asks for a message that lists what remains in order of urgency, with a plain explanation of why each is needed and a reminder date.
An illustrative output groups the remaining eight under "needed before we can start" (source of funds, existing pension statements), "needed within two weeks" and "helpful but not blocking". It suggests reminders on Monday and the following Thursday. The administrator checks the groups against the firm's real order of dependency, because a document that blocks the start must not be filed under "helpful", then sends the message and diarises the reminders.
Mistake to watch for: the assistant explains why an item is needed using plausible but wrong reasons, such as citing a rule you do not rely on. Supply the reasons yourself in the checklist and tell it not to add any. A one-line reason column in the checklist costs you an hour once and saves explaining forever.
Example 5: Newsletter and general-education drafts
This is the lowest-risk job because it needs no client data and contains no personal advice. A quarterly newsletter might cover how to read a statement, what an annual review covers, how pension contributions work in general terms, or what to have to hand for a meeting. The assistant can draft from your outline in your voice. For a general piece you still need a compliance read, since communications to clients and the public are often regulated even when they are not personal.
Worked example: a 400-word newsletter piece
Give an outline: topic (preparing for an annual review), five bullet points the firm wants to make, the tone (calm, plain, no jargon), the word limit and a sentence on what to avoid: no forecasts, no performance, no product mentions, no statements about returns. Ask for two versions of the headline. A sample opening line might be: "A good review starts before the meeting. Five minutes of preparation means we spend the hour on what matters to you." The compliance lead reads the draft against your communications rules, removes anything that could read as a promise or a recommendation, and approves a final version for the archive.
Two tips. Keep one standing instruction in the prompt: "Do not state facts or figures; where a figure is needed, leave a bracketed placeholder for me to fill from a source." And keep the approved version, the date and the approver, because a newsletter is a communication and may need to be shown later.
Where the five examples sit on a risk scale
| Job | Client data involved | Time saved (illustrative) | Main risk | Reviewer |
|---|---|---|---|---|
| Newsletter drafts | None | About 90 minutes per issue | Wording that reads as advice | Compliance lead |
| Onboarding checklists | Document names only | About 20 minutes per client | Wrong dependency order | Operations manager |
| Follow-up emails | From approved note | About 6 minutes each | Overstating what was agreed | Adviser |
| Review packs | Goals and actions | About 35 minutes each | Missing open items | Adviser |
| Meeting notes | Full conversation | About 18 minutes each | Misattribution, records rules | Adviser and compliance |
Run them in that order, easiest first, and give each a month. The wider rollout plan, with governance steps, is in the 90-day rollout plan for small wealth firms. If your clients are very high-net-worth, the privacy demands go up, and AI tools for UHNW wealth managers covers seven jobs with the privacy rules that apply.
A second household: how the five jobs chain together
The five jobs are easier to judge when you follow one household through them. Say a couple in their early fifties become clients in January. The administrator runs the onboarding checklist in week one (Example 4). The adviser holds the first meeting in week three, and the notetaker drafts the note, which the adviser corrects and approves in ten minutes (Example 1). The approved note produces the follow-up email, sent the same afternoon (Example 2). The couple appear in the March newsletter mailing, which was drafted weeks earlier (Example 5). Eleven months later, the review pack is built from that approved note, the year's action list and the reporting system's figures (Example 3).
Notice that the approved note is the hinge. The follow-up, the pack and the next year's agenda all grow out of it, so an error left in the note travels. If the note says the couple want to retire at 62 when they said they might consider 62, every later document inherits the firmer wording. That is why the adviser's review of the note is the one step to protect, even when the firm is busy, and why a hasty approval is more costly than a slow one.
Testing a notetaker before you trust it
Before any tool touches a real client, run a short bench test. Take three meetings from your own team, role-played or already completed with permission, and have the tool produce notes. Then score each note against a simple list: were all attendees identified correctly, were figures and dates right, were decisions separated from discussions, were any statements attributed to the wrong person, and did it add anything nobody said? Count the errors per meeting. A tool that makes one or two small slips you can fix in minutes is usable. A tool that invents a decision, or loses a client's expression of worry, is not.
Repeat the test with the awkward cases: a noisy room, two people talking at once, a client with a strong accent or a quiet voice, and a meeting where the client changes their mind halfway. Notetakers handle neat conversations well and messy ones less so. Write down the settings you chose, such as language, speaker labels and where the recording is stored, so that every adviser uses the same ones.
A one-page AI-use policy you can write this week
A small firm does not need a long policy to start; it needs one page that everyone has read. Headings that work: approved tools (named, with the plan type); what may be pasted (no names, account numbers or national identifiers unless the tool is approved for client data); what is never delegated (advice, suitability reasoning, recommendations, KYC decisions); review (every output read by a named person before use); records (what is kept, where and for how long, as agreed with compliance); and incidents (what to do if client data goes into the wrong tool or an error reaches a client).
An illustrative line for the incidents section: "If you paste client information into a tool that is not on the approved list, tell the operations manager the same day, do not delete the conversation until they have seen it, and note what was shared." That wording matters because people hide mistakes when the penalty is unclear and report them when the process is. Review the page after the first quarter, when you know which rules were too tight or too loose, and again whenever you change tools.
Keep the policy and the tool choices consistent with what your clients are told. If your terms of business say nothing about AI, ask your compliance lead whether to add a short paragraph on AI-assisted administration and human review.
Why none of this gives advice, and how to keep it that way
An assistant that suggests how to invest, summarises a client's situation into a recommendation or drafts a suitability argument has crossed the line, even if the output sounds modest. Keep three habits. Write prompts that forbid recommendations, forecasts and product names. Have the adviser read every output with the client's file open. And keep a short written policy for AI use, naming the approved tools, what data may be pasted, what is retained and who signs off. The adviser, not the tool, is responsible for what the client receives.
If you would rather map this with someone, that is what my AI implementation consultation covers: picking the admin jobs that pay off in your firm and handing over a routine your team can run. The firm above would reasonably start with newsletters and onboarding, and move to notes once recordkeeping is settled.
Questions from small wealth firms
Which admin job is safest to start with?
Newsletter and general-education drafts, because they carry no client data and no personal advice. Meeting notes give the biggest time saving but involve client information and records rules, so start them once your tool choice and retention approach are settled and a person reviews every note.
Are AI meeting notes part of the client record?
They may be, depending on your rules, so ask your compliance lead. Decide whether the recording, the transcript, the AI summary and the final adviser-approved note are each kept, for how long and where. Many firms treat only the reviewed note as the record and keep the others under a defined schedule.
Can AI write the suitability report or recommendation?
It should not. Recommendations rest on the client's circumstances and the adviser's judgement and responsibility. AI can organise facts you have gathered, draft neutral sections such as background, and check completeness against a template, but the adviser writes and owns the reasoning and the recommendation.
How much does this cost a small firm?
A standard seat on ChatGPT Business or Claude Team is $20 a month billed annually or $25 monthly. Adviser notetakers cost more: in November 2026 Jump listed its Meet product at $100 per adviser a month, Zocks started at $67 per user a month billed annually, and Wealthbox's AI Notetaker add-on was $49 per user a month on top of the CRM. Check current prices before you budget.
Further reads
- AI for Wealth Managers: Drafting Proposals With a Compliance Review — Proposal drafts with a compliance review.
- AI Software for Wealth Management: A Small Firm's Evaluation Checklist — A checklist for evaluating wealth-firm AI software.
- AI Note Takers for Financial Advisers: Compliant Meeting Records — Compliant meeting records for advisers.
- Can Financial Advisers Use ChatGPT? What Compliance Allows — What compliance allows with a general assistant.
- AI Governance Checklist for Small Financial Advice Firms — A governance checklist for small advice firms.
- FP Alpha vs Holistiplan: Tax Report Tools for Planners Compared — Per-household against per-credit pricing, a feature table from both vendors' pages, and a test you can run on one synthetic return before paying.
- Holistiplan Pricing: Basic vs Premium Tax and Who Needs Which — Holistiplan Basic Tax vs Premium Tax priced by household tier from 30 to 750, with a Premium break-even, the billing rules and a 7-day trial checklist.
- AI Tools for Investment Managers: Client Reports and Fact Sheets — How a small investment manager can draft quarterly reports and fact sheets with AI while the numbers stay locked and compliance gets a reviewable packet.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: general advice-firm practice; ChatGPT Business, Claude, Jump, Zocks and Wealthbox pricing pages (checked 4 November 2026). No regulator guidance is quoted; confirm recordkeeping and communications rules with your compliance lead.