AI for Wealth Management: Meeting Prep, Review Packs and Compliance for Advisers and RIAs

AI for wealth management: a natural fit, and a cautious one

An advisory practice runs on client context. Before every review, someone gathers the last meeting’s notes, current holdings, performance against the plan, cash needs, tax events, beneficiary changes and any open service tickets. After every meeting, someone writes the notes, the follow-up email and the CRM tasks. Across a book of a few hundred households, that preparation takes more adviser and associate time than the meetings themselves.

Language models are good at exactly this kind of gathering and drafting. The difficulty is that every input is personal financial data, and much of the output either becomes a regulated communication or a required record. That combination decides which tools a firm can use and how.

This guide is written for registered investment advisers, broker-dealer wealth divisions, private banks and UK wealth managers. It covers the work first, then the rules, then the buying decision. Regulatory sources were read on regulators’ own sites on 25 October 2026.

AI tools for financial advisors: seven jobs worth starting with

Meeting preparation from CRM and portfolio data

The highest-value starting point for most firms is a one-page brief per client meeting. A model reads the CRM history, the portfolio management system’s holdings and performance, the financial plan’s goals and the last review’s action items, then drafts a page that says what changed since the last meeting and what the adviser should raise. The brief should link each statement to its source record so the adviser can check a figure in one click.

Client review packs

Quarterly or annual review packs combine performance, allocation against target, fees paid, realised gains and planning updates. Most of the numbers already come from the portfolio system. AI helps with the commentary: a plain-language explanation of why the portfolio moved, written to the client’s level of sophistication, with every number pulled from the system of record rather than generated.

Proposals and investment policy statements

A proposal for a prospect, or a revised IPS for an existing client, follows a firm template. AI can fill the template from the risk questionnaire, the fact-find and the firm’s model portfolio descriptions, and flag gaps such as a missing time horizon or liquidity need. The recommendation inside the document remains the adviser’s decision under the duty of care.

Research summaries

Advisers receive more house views, fund factsheets and manager commentaries than they can read. Turning that material into talking points is a low-risk use when the source set is the firm’s own approved research.

Compliance pre-review of client communications

Marketing pieces, newsletters, social posts and client letters go through compliance before they go out. A model can do a first pass against the firm’s checklist, for example unbalanced performance claims, missing disclosures, testimonials without the required statements, or promissory language, and send the marked-up draft to a human reviewer. It speeds the queue without replacing the approval.

Onboarding and KYC paperwork

New accounts generate identity documents, trust deeds, powers of attorney, beneficial ownership forms and source-of-wealth evidence. Extracting the fields, checking them against the account application and listing what is missing is structured document work that models handle well. The KYC onboarding agent shows how that review is laid out with a human decision at the end.

Meeting notes and follow-ups

Transcribing a client meeting and drafting the notes, CRM update and follow-up email is the most popular adviser use and the one with the most legal strings attached. It gets its own section below.

AI agents for wealth management: where an agent beats a chat window

A chat assistant answers one question at a time. An agent can run a sequence: pull the client record, read the holdings, compare the allocation with the IPS, check whether a required minimum distribution or a large cash movement is due, and draft the agenda. That is useful because most adviser preparation is a fixed sequence across several systems.

Three design rules keep agents within an adviser’s supervisory obligations:

1. Read and draft, never act alone. An agent can propose a rebalance or draft a client email. Placing a trade, moving money or sending the email needs a named person to approve it.
2. Narrow tool grants. A meeting-prep agent needs read access to the CRM and portfolio system. It does not need the ability to change account details or reach the trading system.
3. A trace for every run. Record which data the agent read, which model produced the draft and who approved it, so compliance can reconstruct any output later.

The wealth advisory agent prepares advisor briefs, suitability context and portfolio commentary on this pattern, and the wider set of AI agents for banking and wealth teams covers onboarding, servicing and compliance monitoring alongside it.

Recording and transcribing client meetings: consent and records

AI note-takers have spread quickly through advisory firms because they remove the most tedious part of the day. Two obligations apply before the first recording.

Consent. In the United States, the federal wiretap statute permits recording when one party to the conversation consents. State law can be stricter: Washington, for example, makes it unlawful to record a private conversation without first obtaining the consent of all the persons engaged in it, and other states have similar all-party rules. Clients join calls from many states, so most firms adopt a single practice of telling every client in advance and getting agreement on the record. In the UK and EU, data protection law requires the firm to explain what it records, why, how long it keeps the recording and who processes it, including the note-taker vendor.

Records. Rule 204-2(a)(7) under the Advisers Act requires an adviser to keep written communications sent or received that relate to any recommendation or advice given or proposed. Records must be kept for at least five years from the end of the fiscal year of the last entry, the first two in an appropriate office of the adviser. An AI summary emailed to a client, or a transcript that captures a recommendation, can fall inside that rule. Broker-dealers have their own retention rules under SEA Rule 17a-4, which our guide to FINRA’s expectations for generative AI covers in detail.

The SEC’s off-channel communications cases show what happens when required records sit in tools a firm does not capture. On 13 January 2025 the SEC announced settlements with nine investment advisers and three broker-dealers, $63.1 million combined, for failing to preserve electronic communications. The same reasoning applies to a note-taker that stores transcripts in a vendor account nobody archives. In February 2026 the director of the SEC’s Division of Investment Management said the treatment of electronic communications under the books and records rule remains unclear and the rule reflects a paper-based era. Until it changes, the safe reading is to keep anything that relates to advice.

Practical controls:

• approve specific note-taker tools and block the rest;
• configure retention so that transcripts and summaries flow into the firm’s archive;
• have the adviser review and correct the AI summary before it is filed or sent, because a wrong summary in the file is worse than none;
• decide whether raw audio is kept or deleted once the reviewed notes exist, and write that into the policy.

AI compliance for RIAs: the rules that apply in 2026

The SEC has no AI-specific rule for investment advisers. In June 2025 it withdrew its 2023 proposal on conflicts of interest from predictive data analytics, effective 17 June 2025, and said it does not intend to finalise it. The obligations below carry the weight instead.

Fiduciary duty and Regulation Best Interest

The Commission’s 2019 interpretation of the adviser standard of conduct describes a fiduciary duty made up of a duty of care and a duty of loyalty, including a reasonable belief that advice is in the client’s best interest based on a reasonable understanding of the client’s objectives. Broker-dealers recommending securities to retail customers have been subject to Regulation Best Interest since 30 June 2020, which cannot be satisfied by disclosure alone. Neither standard changes because a model drafted the recommendation. If an AI-built proposal relies on a stale risk profile, the adviser who signed it owns the problem.

The marketing rule and AI washing

On 18 March 2024 the SEC settled charges against Delphia (USA) and Global Predictions for false and misleading statements about their use of AI. Delphia agreed to a $225,000 penalty and Global Predictions to $175,000, and both orders found violations of the marketing rule, Rule 206(4)-1. Any statement on a website, pitch deck or social post that the firm uses AI to manage portfolios or plan finances is an advertisement and needs evidence behind it.

What examiners will look at

The Division of Examinations’ fiscal 2026 priorities say examiners will look at firms offering automated investment advisory services and related tools, including whether representations are fair and accurate, whether operations match disclosures, whether algorithms produce advice consistent with investors’ profiles, and whether controls keep automated advice consistent with obligations to retail and older investors. On AI specifically, examiners will check the accuracy of what firms say about their AI capabilities and whether policies and procedures exist to monitor or supervise AI use.

Regulation S-P

The 2024 amendments require covered institutions, including SEC-registered advisers, to run an incident response program, notify affected individuals no later than 30 days after becoming aware of an incident involving sensitive customer information, and oversee service providers, who must notify the firm within 72 hours of a breach affecting customer information systems they maintain. Larger entities had to comply by 3 December 2025 and smaller ones by 3 June 2026. A note-taker, CRM assistant or AI vendor holding client data is a service provider for this purpose.

FINRA for dual registrants and broker-dealer advisers

FINRA’s Regulatory Notice 24-09 of June 2024 says its rules are technology neutral and that generative AI use could implicate virtually every area of a member’s obligations, naming supervision under Rule 3110 and communications under Rule 2210. The GenAI section of FINRA’s 2026 oversight report lists summarisation and information extraction as the most common use and adds risks specific to AI agents, such as acting without human validation or beyond the user’s intended authority.

UK wealth managers: Consumer Duty and the FCA’s approach

UK firms work under the Consumer Duty, which applied to open products and services from 31 July 2023 and to closed books from 31 July 2024. The FCA said on 27 January 2026, launching the Mills Review into AI and retail financial services, that it does not plan to introduce AI-specific regulation.

The FCA’s good practice publication on AI in financial services, published on 19 October 2026, shows how the existing rules apply. It asks firms to define the consumer need before deploying AI, test outcomes across customer groups, keep clear senior accountability and audit trails, and avoid depending on a single vendor without an exit plan. It also names weak practices: excluding vulnerable customers from AI-supported service by default, overstating what AI can do, and tick-box human review. For a wealth manager, that translates into testing AI-drafted suitability reports and client letters for clarity and fairness, not just accuracy.

Which client data can go to a cloud assistant

The table assumes an enterprise subscription with terms that exclude training on customer data. Treat it as a starting point for your own data classification, not as advice on a particular vendor contract.

Choosing wealth management AI tools

Four categories compete for the same budget.

Point solutions for advisers. Note-takers and meeting assistants built for financial advisers, often connected to the CRM. They are quick to deploy. Check where audio and transcripts live, for how long, and whether they reach your archive.

AI inside existing platforms. CRM, planning and portfolio vendors are adding assistants to their products. Data stays with a vendor you already oversee, but each feature runs under that vendor’s model provider and terms, which you need to review.

General enterprise assistants. Strong at drafting and summarising public material. They become harder to justify once client holdings or identity documents are involved, unless your policy and vendor review cover that data.

A platform in your own environment. For firms that want one governed layer across CRM, portfolio data, research and document stores, with agents and an audit trail they control. This is the route for larger RIAs, private banks and firms with strict data residency needs.

Whichever you choose, ask for the same evidence: data flows and subprocessors, retention and deletion, training terms, export of prompts and outputs for your records, access controls by role, and an incident notification commitment that fits Regulation S-P. The finance executive brief sets out the governance questions boards ask, and our comparison of open models for financial analysis helps when the plan is to run models in-house.

The future of AI in wealth management

The near-term direction is visible in what regulators have said, not in forecasts. The SEC withdrew its predictive analytics proposal and examines AI through existing rules. Its investment management director told an industry audience in February 2026 that liability concerns should not be insurmountable obstacles to adoption, and invited firms to seek staff guidance. The FCA is relying on the Consumer Duty and senior accountability while the Mills Review considers what advanced AI means for retail markets.

For firms, the advantage goes to those who can evidence control: an approved tool list, data classification, review steps, retention, and a log of who used which model on which client’s data.

How VDF AI fits a wealth management firm

VDF AI is installed in the firm’s own data centre, a private cloud tenancy or an isolated network with no internet route. When it is configured for local models only, client notes, holdings extracts, onboarding documents, retrieval indexes and logs never leave that environment. VDF AI holds no SEC, FINRA or FCA approval or certification of its own; it operates under the firm’s compliance program, vendor oversight and Regulation S-P procedures.

In practice, research libraries, compliance manuals, model portfolio documents and client files sit in separate knowledge collections, each with its own access list. Retrieval respects those permissions, so an associate in one adviser team cannot surface another team’s households through a question. Governed agents handle the multi-step work described above, meeting briefs, review pack commentary, onboarding document checks and communication pre-review, with each role granted only the tools it needs and an approval step before a draft becomes a client communication. The log for every turn shows the user, the model, the source passages and any tool calls, which gives compliance the reconstruction it needs for a books and records review.

The VDF AI Router puts local open-weight models served by Ollama or vLLM behind one policy layer and can admit a cloud model for low-risk tasks such as public market commentary where the firm allows it, with budgets and a per-call audit record. Advisers sign in with Microsoft Entra ID single sign-on on-premises; firms on Okta or another SAML or OIDC provider connect through an SSO-aware reverse proxy, and role-based access control is part of every plan. A wealth firm can pilot the setup on one adviser team through a three-month on-premises platform trial before deciding on seats.

The finance and banking solution and the hub of AI use cases for finance describe the wider setup. To see where your firm stands on data access, governance and skills before a pilot, take the free AI readiness assessment.

This guide is general information for planning and is not legal or compliance advice.