AI Workflows · Workflow playbook · Updated July 2026
AI Investment Policy Statement Drafting: The Advisor-Built Workflow
Search "IPS software" and you get vendor demos. This is the part that saves you the hour of blank-page drafting: how to turn a real client risk profile into a compliant first-draft Investment Policy Statement with AI, then run that draft through a Reg BI and fiduciary review layer where you, the human, still own suitability. Includes a copy-paste prompt and the exact data you may and may not feed the tool.
Key takeaways
- Structure is the win. AI is fast at the structure of an IPS: objectives, constraints, allocation ranges, rebalancing rules, and review cadence assembled from your inputs. It is not the source of the suitability decision.
- The determination stays human. The best-interest determination under Reg BI, and the fiduciary duty of care for an RIA, are yours. AI drafts the document; you decide whether the strategy fits the client.
- De-identify before you paste. Never put client-identifying data into a consumer AI tool. Your Regulation S-P duty travels with the data.
- Keep your marketing accurate. Do not describe your practice as "AI-powered" in a way you cannot support. The SEC has brought AI-washing cases, and the Marketing Rule governs how you talk about your process.
What an IPS must contain
An Investment Policy Statement is the governance document that records how a portfolio will be run and why that approach suits the client. It is standard practice for fiduciary engagements and effectively expected for ERISA plans and institutional accounts, and it is the paper trail that shows your recommendation was grounded in the client's actual situation. Before you automate anything, you need to know the parts, because AI can only assemble a document as complete as the schema you give it.
A workable IPS covers the client's return objective and the risk they can bear, the time horizon, liquidity needs, and the constraints that shape the portfolio: tax situation, legal or account-type limits, and any personal restrictions such as securities the client will not hold. It sets a target asset allocation with ranges for each class, states a rebalancing policy, defines how performance will be measured and against what benchmarks, and fixes a review cadence. It also names roles: who advises, who has discretion, who approves changes. Miss one of these and the document reads as a template rather than a real policy for a real person.
The IPS is not the strategy. It is the written proof that the strategy was chosen for this client, on purpose, and can be reviewed later.
The AI-assisted IPS drafting workflow
The workflow has a shape that never changes: give the AI clean, de-identified planning facts, ask for a structured draft against a defined schema, then verify and finish the document as the responsible professional. AI handles the assembly. You handle the judgment and the accountability. Sort the work before you start so you are clear on which side of the line each task sits.
| Task | Hand to AI (draft and assemble) | Keep human (judgment and sign-off) |
|---|---|---|
| Turning a risk profile into a first draft | AI structures objectives, constraints, and ranges from your inputs | You confirm the profile is accurate and complete |
| Asset-allocation ranges | AI proposes ranges consistent with the stated risk level | You set and approve the final ranges as suitable |
| Boilerplate and formatting | AI drafts rebalancing, benchmarking, and review-cadence language | You verify it matches how you actually manage the account |
| The best-interest determination | Nothing; this is a professional judgment | Always human, documented under Reg BI |
| Suitability for this specific client | Nothing without your independent review | You own it and you sign it |
| Anything with client-identifying data | Only after de-identifying | You own the Regulation S-P confidentiality decision |
Read the right column as your license at work, not as a limit on the tool. The reason the IPS carries weight is that a credentialed human stands behind the recommendation. AI changes how fast you reach a draft. It does not change who is responsible for the advice.
Turning a client risk profile into a first-draft IPS
Here is the core of the workflow: a copy-paste prompt that converts a de-identified profile into a structured first draft. Fill the bracketed fields from your own notes and risk-tolerance questionnaire. Do not include the client's name, entity name, account numbers, or Social Security number. Use neutral labels instead.
Copy-paste prompt: profile to first-draft IPS
You are helping a financial adviser draft a first-pass Investment Policy Statement. Do not give investment advice or claim any recommendation is suitable. Produce a structured draft from the facts below that the adviser will review and finish.
Client profile (de-identified): Household type: [individual / joint / trust / retirement plan]. Investable assets: [amount]. Time horizon: [years]. Primary objective: [growth / income / preservation / balanced]. Stated risk tolerance: [conservative / moderate / aggressive], from a completed questionnaire. Liquidity needs: [near-term cash needs]. Tax situation: [bracket, account types: taxable / IRA / 401k]. Constraints: [restricted securities, concentrated positions, ESG or other preferences]. Existing holdings of note: [asset classes only, no identifiers].
Draft an IPS with these sections: 1) Purpose and scope. 2) Return objective and risk parameters. 3) Time horizon and liquidity. 4) Constraints (tax, legal, account-type, personal). 5) Target asset allocation with a range for each class, consistent with the stated risk level. 6) Rebalancing policy. 7) Benchmarks and performance measurement. 8) Review cadence and roles. For every allocation range, add a one-line rationale tied to a stated fact. Flag any section where the profile is incomplete and tell me what is missing. Do not invent facts not given.
Two design choices in that prompt matter. First, it forbids the tool from claiming suitability, which keeps the judgment where it belongs. Second, it asks the AI to flag missing inputs rather than paper over them, so gaps in your intake surface on the page instead of hiding in confident prose. When the draft comes back, read the flags first.
Objectives, constraints, asset-allocation ranges
The section AI drafts fastest is also the one you must check hardest: the target allocation and its ranges. A language model will produce ranges that look reasonable for a stated risk level, and reasonable-looking is not the same as suitable. Your job is to confirm each range against the client's real capacity for loss, their liquidity timeline, and their tax situation, then set the final numbers yourself.
Work through the draft in order. Confirm the return objective matches what the client told you and what their assets can realistically support. Check that the risk parameters reflect both willingness and ability to bear risk, which are different things a questionnaire alone does not settle. Verify each constraint is captured: a concentrated low-basis position, a required minimum distribution timeline, a security the client refuses to own. Then look at every allocation range and ask whether you would defend it in front of a regulator or a review committee. If a rationale line reads thin, that is the tool telling you the input was thin. Fix the input, not just the sentence.
The Reg BI and fiduciary review layer
This is the step that makes the workflow defensible, and it cannot be delegated. If you are a broker-dealer or associated person, Regulation Best Interest requires you to act in the retail customer's best interest when you make a recommendation, and to have a reasonable basis for it. If you are an investment adviser, you owe a fiduciary duty of care and loyalty under the Investment Advisers Act. Neither obligation is satisfied by a document a machine assembled. The AI draft is an input to your determination, not the determination itself.
Human-owned review before any IPS is final
- Make the best-interest or fiduciary determination yourself. Decide whether the strategy in the draft actually serves this client, and document your basis. The AI does not weigh costs, alternatives, or the client's full situation the way your duty requires.
- Verify every stated fact against the client file. Objectives, horizon, risk level, constraints, and asset figures must match your records, not just the prompt you typed.
- Confirm the allocation is suitable, not just internally consistent. A range can be coherent on the page and wrong for the person. You set the final numbers.
- Keep the recommendation and the signature human. The IPS goes out under your name and your firm's, and the accountability travels with it.
Handled this way, AI compresses the drafting time while your professional review stays exactly as rigorous as it was before. That is the trade you want. The tool removes the typing, not the duty.
What client data you may and may not feed AI
Your duty to protect client information does not pause when you open an AI tool. Under Regulation S-P, you are responsible for safeguarding nonpublic personal information, and that responsibility follows the data wherever you send it. A consumer chatbot is a third party. Treat it like one.
Set these guardrails before you use AI on any client work
- De-identify before you paste. Names, entity names, account numbers, Social Security numbers, and any detail that could identify the client do not belong in a consumer AI tool. Replace them with neutral labels. Planning facts stated in the abstract are safe; identifiers are not.
- Know where the data goes. Confirm whether the tool trains on your inputs, where it stores them, and who can access them. Read the data terms, not the marketing. Where your firm has an enterprise tier with a no-training setting and contractual protection, use it.
- Verify every number and every claim. No allocation range, benchmark, tax figure, or rule reference goes into a final IPS without independent confirmation. AI can state a figure or a rule with total fluency and be wrong.
- Follow your firm and regulatory policy. Check your compliance department's AI policy, your recordkeeping obligations, and any regulatory guidance before you adopt a tool for client work. When in doubt, get it cleared first.
One more line worth stating plainly: how you describe your use of AI is itself regulated. The SEC has pursued advisers for AI-washing, meaning false or exaggerated claims about how they use artificial intelligence, and the Marketing Rule governs adviser advertising. Using AI to draft an IPS is fine. Telling clients your portfolios are "AI-managed" when a human makes every decision is a claim you cannot support. Keep your marketing as accurate as your file notes.
AI drafting vs IPS software
The vendors ranking for "IPS software" sell a different thing than this workflow, and it is worth being clear on the difference before you pay for either. Dedicated IPS and financial-planning platforms give you templated documents, data integrations with custodians, and compliance-oriented recordkeeping. That is real value, especially at scale, and nothing here argues against it.
What those platforms generally do not give you is flexible, first-draft language from a messy set of intake notes. That is where a general AI tool earns its place. Think of it as the drafting assistant that gets you from a completed risk profile to a structured document you can then refine or drop into your planning software. Many advisers will use both: AI for the fast first draft and the plain-language rationale, the platform for the system of record and the client-facing polish. Neither replaces your review, and neither one is the reason the advice is sound. You are.
Version control and annual-review reuse
An IPS is a living document, and the workflow pays off most on the second pass. Save the de-identified profile and the prompt you used, so next year's review starts from a structured baseline instead of a blank page. When the client's situation changes, a new job, a liquidity event, a shift in risk tolerance, update the profile fields and ask the AI to redraft only the affected sections, then run the same human review layer over the changes.
Keep a short record of your process: that AI was used to assemble the draft, that you verified the facts and set the allocation, and that the best-interest or fiduciary determination is yours. This is good practice for your own quality control, it supports your recordkeeping obligations, and it makes the annual review faster because the reasoning is already on paper. Date every version and note what changed and why. The document that shows its own history is the one that holds up under review.
What AI does not replace
AI took the blank page and the boilerplate, not the advice. An adviser still owns the parts that have no clean prompt: the judgment that a strategy fits a specific client, the best-interest determination the regulator will ask about, the fiduciary duty owed to the person across the table, and the signature that carries the accountability. Used well, AI clears the drafting time so you spend more of your day on the suitability thinking and the client conversation that are the actual job. Used carelessly, it is a fast way to ship a polished document that was never suitable in the first place. The difference is the review discipline, and that discipline is yours.
How we built this playbook
This workflow comes from hands-on testing of the profile-to-draft pattern on de-identified, non-client sample profiles, not from a survey or vendor claims. We do not cite a respondent count or a time-saved percentage, because The Leveraged Years just launched and we will not invent data we do not have. The regulatory references here, Regulation Best Interest, the Investment Advisers Act fiduciary duty, Regulation S-P, the Marketing Rule, and the SEC's AI-washing enforcement, are real obligations you should confirm against the current rules and your own compliance department before you act. Where AI is unreliable, we say so plainly so you verify rather than trust. None of this is legal, tax, or investment advice.
Part of TLY's AI Workflows → workflow playbooks for senior professionals.
Frequently asked questions
Can AI write a compliant Investment Policy Statement?
AI can write a strong first draft of an IPS from a de-identified client profile: objectives, constraints, allocation ranges, rebalancing rules, benchmarks, and review cadence assembled in minutes. It cannot make the document compliant on its own, because compliance rests on your suitability judgment, your best-interest determination under Reg BI or your fiduciary duty of care as an adviser, and your verification of every fact. Use AI for the draft, and finish the document as the responsible professional.
Is it safe to put client data into ChatGPT or Claude to draft an IPS?
Not client-identifying data, and not in a consumer account. Strip names, account numbers, and Social Security numbers, and feed only de-identified planning facts. Your Regulation S-P duty to protect nonpublic personal information follows the data into the tool, so treat the tool as a third party. Where your firm offers an enterprise tier with a no-training setting and contractual protection, use it, and follow your compliance policy either way.
Does using AI change my Reg BI or fiduciary obligations?
No. Regulation Best Interest still requires a broker-dealer to act in the retail customer's best interest with a reasonable basis, and an investment adviser still owes a fiduciary duty of care and loyalty. An AI-drafted IPS is an input to your determination, never a substitute for it. You make the best-interest or fiduciary decision, you document your basis, and you sign the work.
Can I tell clients my practice is AI-powered?
Only if the claim is accurate and you can support it. The SEC has brought AI-washing cases against advisers for false or exaggerated statements about their use of artificial intelligence, and the Marketing Rule governs adviser advertising. Using AI to draft documents is fine to disclose. Describing portfolios as AI-managed when a human makes every decision is a claim you cannot back up, so keep your marketing as accurate as your process.
How is this different from IPS software?
IPS and planning platforms give you templates, custodian data integrations, and compliance recordkeeping, which is valuable at scale. A general AI tool gives you flexible first-draft language and plain-language rationale from messy intake notes, which templates do not. Many advisers use both: AI for the fast draft, the platform for the system of record. Neither replaces your review or your responsibility for the advice.
Build the workflow, not just the opinion
Knowing the shape of the workflow is the start. Running it every time, with the verification and confidentiality discipline baked in, is the skill that compounds. We teach the full method, the prompts, and the guardrails as one repeatable system a senior adviser can defend to a client and a compliance officer alike.
Start with Leveraged Wealth Advisor: the AI operating system for advisory practices New to this? Begin with Leverage Starter, the entry course Join The Leverage Club for $49 and get the prompts, schemas, and review templates Not sure where to start? Take the 2-minute course finderSources: TLY hands-on use of leading general-purpose models on de-identified, non-client sample profiles (July 2026). Regulatory references, current as of drafting: Regulation Best Interest (SEC broker-dealer standard of conduct); fiduciary duty of care and loyalty under the Investment Advisers Act of 1940 for registered investment advisers; Regulation S-P (SEC privacy of consumer financial information); the Investment Advisers Act Marketing Rule (Rule 206(4)-1); and the SEC's enforcement against AI-washing. Capabilities and rules subject to change. This guide is not legal, tax, or investment advice; confirm all obligations against the current rules and your firm's compliance policy.