ChatGPT for financial advisors: what ChatGPT for Financial Services is, and what a solo advisor can use today
An onboarding packet on Monday, an annual review on Wednesday, a client on the phone after a bad week in the market. ChatGPT can take real drafting off that week, and since September OpenAI has an edition built for finance. Here is what it is and who can buy it, what a solo or small-firm advisor can use today, what to keep out of the chat at each client moment, prompts that fit, and the part ChatGPT holds only if you paste it in thread by thread or connect a whole mailbox: what each client sent you this year.
Short answer: ChatGPT for Financial Services is OpenAI's edition for financial institutions, announced on September 10, 2026 and built first for investment banking and equity research, with Morgan Stanley and Evercore as design partners. It is sold to "eligible financial institutions," not to individuals, and has no published price. A solo or small-firm advisor uses ordinary ChatGPT, which still does useful work if you keep client identifiers out of it. For the wider picture, including the questions to put to compliance before any tool, see our guide to AI for financial advisors.
This guide is published by Equerry, a personal assistant for client work. Equerry is not affiliated with OpenAI, and nothing on this page is investment, legal or compliance advice. Every OpenAI statement below was read at its source on 17 October 2026.
What OpenAI launched on September 10, 2026
OpenAI's announcement opens: "We're introducing ChatGPT for Financial Services, a tailored ChatGPT Work experience that combines built-in financial data with GPT‑6 Astra's reasoning to help teams develop research, financial models, and customized client materials." Three parts of it matter to an advisor.
- Who it was built with, and for. "This product has been shaped by our design partnership with Morgan Stanley and Evercore." And: "Our early work with Morgan Stanley and Evercore has helped steer where we have started: investment banking and equity research." OpenAI says its partners "will inform post training, product improvements, and our expansion into other financial services categories." The announcement does not name advisory practices or wealth management. On OpenAI's financial-services page, advisors appear once, in a customer-story card titled "Putting firm knowledge in the hands of financial advisors": the story of Morgan Stanley Wealth Management's internal GPT‑4 tools for its own advisors, including what OpenAI calls "Morgan Stanley's internal chatbot for answering financial advisors' questions."
- The data. It "includes datasets from providers like Daloopa, PitchBook, LSEG News and Crunchbase covering earnings transcripts, financial statements, company fundamentals, private companies, and more," with "no separate contracts to negotiate or connectors to set up." For data a firm already subscribes to, OpenAI says it is "working with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody's on shared sign-in and entitlement integrations."
- The controls. It "builds on ChatGPT Enterprise's SAML SSO, SCIM provisioning, and role-based access controls." "Your firm's business data is not used to train our models by default. It is encrypted at rest and in transit, and admins can configure workspace retention." Compliance teams "can export supported workspace logs through the OpenAI Compliance Platform."
OpenAI's examples are a banker's day (an adjusted EBITDA, valuation models, pitchbooks in firm templates), not a household balance sheet or a review meeting. That is not a criticism; it is who the product is for.
Can a solo advisor get it, and what you can use today
Availability, in OpenAI's words: "ChatGPT for Financial Services is available to eligible financial institutions. If you're interested, please contact us or reach out to your account team." Neither the announcement nor the financial-services page offers an individual sign-up or a price. Inside a bank or a large broker-dealer, the question goes to your firm. In your own practice, or a small one, what you have is ChatGPT:
| Plan | Price (US, as OpenAI states it) | Worth knowing for an advisor |
|---|---|---|
| Free | No charge | "Limited memory and context"; model training "Opt-out available" |
| Go | Shown on the pricing page in local currency | "Longer memory"; "This plan may include ads." |
| Plus | "$20/month (billed monthly)"; no annual billing | "Expanded memory and context"; "Projects, scheduled tasks, and custom GPTs" |
| Pro | "three monthly Pro plans": $100, $200 or $500 a month | "Maximum memory and context" |
| Business | Standard seats "$20 per month billed annually, or $25 per month billed monthly"; Premium $100 or $125 | "at least two paid seats"; workspace content not used for training by default |
On personal plans the pricing page lists training on your content as "Opt-out available" (the switch is below); Business starts from the other default: "By default, OpenAI does not use content from ChatGPT Business, Enterprise, Edu, or ChatGPT for Healthcare workspaces to train its models." With two paid seats required, a one-person practice chooses between a personal plan with training off and paying for a second seat. Neither answers whether client information may go in; that stays with your firm's policy.
Where ChatGPT saves an advisor time, and where it stops
- First drafts. A welcome letter, a meeting recap, the paragraph on why the plan holds two years of cash. You rewrite; it saves the blank page.
- Questionnaires and checklists. Turning your discovery questions into something a client can answer at the kitchen table, or listing what to review after a marriage or a death in the family.
- Rehearsal. Asking it to play the client who wants to sell everything, so the hard call is not the first time you say the words aloud.
- Public material. A regulator's investor bulletin, a plan's published summary, an IRS page you pasted in full, explained in shorter sentences, with each point tied to the passage it came from.
Where it stops. Advice: OpenAI's usage policies bar using its services for the "provision of tailored advice that requires a license, such as legal or medical advice, without appropriate involvement by a licensed professional." Whatever reaches the client is your advice, so read every line as if you wrote it. Your book: ChatGPT sees no custodian, CRM or planning data unless you upload it or connect it. This year's numbers: limits, thresholds and ages change; take them from the source, not the chat.
One more change concerns your clients. OpenAI's consumer Finances feature "is available to users in the U.S. on Free, Go, Plus and Pro plans"; people connect their own accounts "through Plaid," and ChatGPT can then read "balance, transaction, and investment data." Its own page says "ChatGPT is not a replacement for professional financial advice." Some clients will arrive at a review having asked ChatGPT about their portfolio already; asking what it told them is a good opening question.
What to paste, and what never to paste, by client moment
Your firm's policy comes first; if you run your own practice, that means the policy you signed off as its compliance officer. Within it, sort by the moment, because each one carries a different kind of sensitive detail.
| Client moment | Usually workable | Keep out |
|---|---|---|
| Onboarding | Your blank questionnaire, your welcome-letter template, the order of your first three meetings | The completed intake form, ID documents, account transfer paperwork, Social Security numbers, dates of birth |
| The annual review | The agenda headings, round figures with no name ("a couple in their sixties, about 60/40"), the questions you want to ask | The statement or performance report, a holdings list that identifies a household on its own, the client's own written goals |
| A life event | The type of event (a marriage, a death, a business sale) and the general checklist it triggers | Who died, who inherits, beneficiary names, the health detail or the family dispute that came with the news |
| A call after a market drop | Your own talking points, the questions clients are asking this week, a rehearsal with no names | The client's email asking whether to sell, their balances, anything that pairs a name with a decision |
One input deserves its own decision: connected apps. OpenAI's Memory article lists "content from connected apps, such as Gmail" among the sources ChatGPT can draw on, depending on plan and region. A connection is to the account, not to the one thread you had in mind, and an advisor's inbox is mostly other people's finances. If you are weighing it, our guide on whether ChatGPT is confidential sets out what OpenAI publishes about training and retention, plan by plan.
ChatGPT's own data settings, step by step
Three controls, as OpenAI's help center describes them (read 17 October 2026). They decide training, history and memory for your account; they do not answer the compliance question for you.
- Turn off training before the first client question. On the web: open your account menu, select Settings, then Data controls, and turn off Improve the model for everyone. On iPhone and Android: open the sidebar, select your profile icon to open Settings, select Data controls and turn it off. Off, "your new conversations won't be used to train OpenAI models." One caution from the same article: if you give a thumbs up or thumbs down afterwards, "the entire conversation associated with that feedback may be used to train OpenAI models."
- Use a temporary chat for a one-off question. "Open a new chat." "Select Temporary." "Choose Personalized or Unpersonalized before you send your first message." Choose Unpersonalized: it "Does not use memory, custom instructions, or plugins." While temporary, it stays out of your history and memory and is not used to improve OpenAI's models, though "OpenAI may keep a copy of a temporary chat for up to 30 days for safety purposes." More in our guide to ChatGPT Temporary Chat.
- Check what it kept. Open Settings, select Personalization, then Memory. When a client leaves, deleting the chat is not enough on its own: OpenAI says "Deleting a chat alone does not necessarily delete a separate saved memory created from that chat," so delete the saved memory and the chat, and any file in Library.
Prompts for four client moments
Each one works from material with the "keep out" column removed. Read everything before it reaches a client, and treat any figure it gives as a blank to fill from the source.
Onboarding: a questionnaire a client will actually finish
"Here is my firm's new-client discovery questionnaire: [paste the blank form]. Rewrite it so a couple can finish it in about 15 minutes at home. Use plain words, group the questions under family, work and income, what they own, what they owe, and what they want the money to do. Mark each question that needs a document with [DOC]. Keep every question I have; if you think one is missing, list it separately at the end rather than adding it."
The annual review: a one-page worksheet the client fills in first
"Write a one-page 'what changed this year' worksheet I can send a client two weeks before their annual review. Ask about changes in work, health insurance, family, housing, big purchases planned in the next two years, gifts or inheritances, and anything they read or heard that worried them. Ten questions at most, each answerable in one line. No financial terms they would need to look up."
A life event: what to review after a marriage
"A client has just married. List what an advisor typically reviews with them in the next 90 days, grouped under beneficiary designations, account titling, insurance, estate documents, employer benefits and the household budget. For each item, write the question I should ask them, not a recommendation. Note anything that usually involves their attorney or CPA so I can refer it. Use no names."
After a market drop: rehearse the hard call
"Play a retired client who calls after a sharp market fall and wants to move everything to cash. Push back the way a frightened, intelligent person would. After six exchanges, step out of the role and tell me where my answers sounded dismissive, where I made anything that could be heard as a promise about returns, and which of your questions I never answered."
Every one of these is about the kind of client, not the client. None can supply what this household told you in March.
What ChatGPT holds about your clients, and who decides what reaches it
ChatGPT does remember, so forgetting is not the issue; who decides what it sees is. A client's details reach it by one of two routes. You can paste or upload them yourself, which leaves every decision with you but depends on you doing it: the email where a client mentions, three paragraphs in, that she has been offered early retirement comes in while you are with someone else, and nothing of it reaches ChatGPT until you copy it over. Or you can connect a mailbox, which brings that email in without you, along with every other message in the account, the very thing the "keep out" column warned about. And the memory you can review is, in OpenAI's words, "a high-level view" that "does not necessarily include every detail or source that Memory can reference."
The record a review needs sits between the two: what each household sent and said across the year, which you can check before the meeting. The review routine and the compliance questions that go with any tool are in our wider guide to AI for financial advisors; if you use Anthropic's assistant as well, Claude for financial advisors covers the same ground for it.
Where Equerry fits
Equerry is an iPhone personal assistant that helps you get things done, and in an advisory practice its job is narrow: holding on to what each household has told you, built only from the emails, notes, voice notes and files you decide to give it. The drafting stays with ChatGPT. To be clear about the edges: it does not sit in on or record client calls, it does not connect to a CRM, a custodian or Outlook, it never asks for access to your inbox, and it does not give investment advice.
| What you send | The client's email from Tuesday, asking whether they should sell after the drop, forwarded to your Equerry's own address at mail.getequerry.app; the thread with their CPA about next year's withdrawals, with Equerry on the CC line, so the CPA's replies keep arriving; the voice note you recorded in Equerry in the car after their March review; the typed note from the call in July. |
|---|---|
| What you ask | "Before I call the Haddads back about the drop: what did they say they need cash for this year, and what did I tell them in July?" |
| What you get back | The June date for the down payment and the amount they mentioned, from your March voice note; their question from Tuesday's email; what you told them in July, from the typed note, with a source chip on that line. When an answer draws on a note you saved, it carries a source chip: tap it to see the passage Equerry used and open the original. For the forwarded email and the CPA thread: What Equerry remembered from it is listed in Settings → Your memory, each line marked as from that email, and you can delete any line. If you allow calendar access and set up Meeting Prep, a short brief arrives before their next review from the same material. |
The split: ChatGPT drafts the questionnaire, the worksheet and the rehearsal from what you paste; Equerry keeps what came back, from the completed worksheet a client emailed (forward it) to the comparison you promised by the 14th (set a Reminder, and the push arrives on the day you pick with the words you wrote).
What it cannot do. Equerry works from what you send it and, if you allow it, your calendar; an email you never forwarded simply is not there. Anyone on a thread you CC it into will see its address. It is not a books-and-records or archiving system for regulatory purposes, and it makes no claim about any SEC or FINRA rule. It is also a third-party service: to make what you send searchable it processes it, with help from the providers listed on our AI disclosure page, so put it to compliance exactly as you would ChatGPT. The difference is how much any one yes lets in: a single thread or note, never an entire mailbox.
Sources, all read 17 October 2026 (OpenAI's sites refuse automated fetches, so each page was read as rendered text). OpenAI: Introducing ChatGPT for Financial Services (September 10, 2026); OpenAI for financial services and its customer story Morgan Stanley uses AI evals to shape the future of financial services; Usage policies; ChatGPT Finances; ChatGPT pricing (plan features and the training row; prices shown in local currency). OpenAI Help Center: What is ChatGPT Plus?; About ChatGPT Pro tiers; Sign up for ChatGPT Business (seat minimum and seat prices); Data controls in ChatGPT; Temporary chat in ChatGPT; Memory in ChatGPT. OpenAI changes its products and pages often; where its current page disagrees with this one, OpenAI's page is right. This guide is general information, not investment, legal or compliance advice.
Frequently asked questions
What is ChatGPT for Financial Services?
OpenAI's edition of ChatGPT for financial firms, announced on September 10, 2026. OpenAI calls it "a tailored ChatGPT Work experience that combines built-in financial data with GPT‑6 Astra's reasoning to help teams develop research, financial models, and customized client materials." It was shaped by a "design partnership with Morgan Stanley and Evercore," it starts with "investment banking and equity research," and it includes data "from providers like Daloopa, PitchBook, LSEG News and Crunchbase." It builds on ChatGPT Enterprise's security controls. Neither the announcement nor OpenAI's financial-services page publishes a price. The announcement does not name financial advisors or wealth management; on the financial-services page, the one mention of advisors is a customer story, "Putting firm knowledge in the hands of financial advisors," about Morgan Stanley Wealth Management's internal GPT‑4 tools for its own advisors, including what OpenAI calls "Morgan Stanley's internal chatbot for answering financial advisors' questions." (openai.com, read 17 October 2026.)
Can a solo financial advisor use ChatGPT for Financial Services?
Not by signing up for it. OpenAI's announcement ends: "ChatGPT for Financial Services is available to eligible financial institutions. If you're interested, please contact us or reach out to your account team." There is no individual plan and no published price. What a solo advisor can buy today is ordinary ChatGPT: Plus at "$20/month (billed monthly)", one of "three monthly Pro plans" at $100, $200 or $500, or a Business workspace, which needs "at least two paid seats". Whether client information goes into any of them is your firm's decision, or yours as your own compliance officer. (OpenAI, read 17 October 2026.)
What should a financial advisor never paste into ChatGPT?
As a working default, before your firm's policy says otherwise: account numbers, Social Security numbers, dates of birth, a statement or holdings list with a name on it, the client's email itself, and the private things clients tell you in passing (health, a divorce, an inheritance). Questions usually survive without them: "how do I explain a backdoor Roth to a couple in their forties" needs no names and no balances. OpenAI's Memory article adds its own caution: "Sensitive information can appear in Memory if you share it with ChatGPT." This page sorts the rest by client moment: onboarding, the annual review, a life event and a call after a market drop.
Can ChatGPT give financial advice to my clients?
Treat nothing it writes as advice, and never send a client its answer as yours unchecked. OpenAI's usage policies list among things its services may not be used for the "provision of tailored advice that requires a license, such as legal or medical advice, without appropriate involvement by a licensed professional." Its consumer Finances page, which lets people in the US connect their own accounts, says "ChatGPT is not a replacement for professional financial advice." The advice is yours, and so are the duties that come with it. Your firm's policy decides which uses are allowed and how drafts are reviewed. (OpenAI, read 17 October 2026.)
Does ChatGPT remember what each client told me?
It can remember a good deal. OpenAI's Memory article says that when Memory is on, ChatGPT "can remember relevant preferences and details from your chats and other available sources," which, depending on plan and region, include past chats, saved memories, files in Library and "content from connected apps, such as Gmail." So what it knows about a client is what you typed, pasted, uploaded, or reached by connecting a whole account. OpenAI also says "Memory does not retain every detail from every conversation," and the memory summary you can review is "a high-level view" that "does not necessarily include every detail or source that Memory can reference." It is not a client file you can audit line by line. (Read 17 October 2026.)
How much does ChatGPT cost for a financial advisor?
From OpenAI's help center, read 17 October 2026 (US dollars): Plus is "$20/month (billed monthly)", and "we do not support annual billing or the option to pay for multiple months in advance for ChatGPT Plus subscriptions." Pro comes as "three monthly Pro plans": Pro 100 at $100, Pro 200 at $200 and Pro 500 at $500 a month. A Business workspace needs "at least two paid seats"; Standard seats "are $20 per month billed annually, or $25 per month billed monthly," and Premium seats $100 billed annually or $125 billed monthly. The pricing page shows prices in your local currency. ChatGPT for Financial Services has no published price.
Last updated 2026-10-17.
All product names, logos, and brands are property of their respective owners. Equerry is not affiliated with, endorsed by, or sponsored by any other product mentioned; comparisons are provided for informational purposes.