ChatGPT prompts for accountants, bookkeepers, lawyers and insurance agents: 14 copy-paste prompts from 13 real firms (2026)

What ChatGPT prompts help with client emails and questions?

Five firms use it on what clients read. Libby Walklett drafts a difficult email herself, with names and confidential details removed, and asks ChatGPT to rewrite it as sensitively as possible; when a client asked for a payment plan she doesn't normally offer, the client was really happy with the reply. At GWCPA, staff research a client's quick question in Blue J and draft the reply for a partner to verify, and a custom ChatGPT built on the firm's SOPs answers staff questions about how things are done. Gloria Hebert has ChatGPT make educational client worksheets that used to take an hour or two. At Jon Peters Insurance Agency, staff draft and polish client emails in ChatGPT for Business and build HTML email templates with CRM merge fields. Top O' Michigan Insurance used generative AI to write a message for each client moment, such as a renewal document arriving in the portal, and built them into automated workflows in Applied Epic, refreshing the wording with ChatGPT or Claude.

Here is my draft email to a bookkeeping client, with names and confidential details removed: {DRAFT}. The situation: {SITUATION, e.g. they can't pay and asked for a payment plan}. My decision: {DECISION, e.g. a payment plan for this year only, as an exception}. Rewrite it to be as kind and clear as possible without changing the decision, dates or amounts. Keep it under 150 words, in plain English. Don't add promises I didn't make.

Draft a reply to this client question for {FIRM NAME}. Tax year: {TAX YEAR}. Jurisdiction: {STATE/FEDERAL}.
Question: """{CLIENT EMAIL}"""
Research notes and citations from our tax research tool: """{RESEARCH NOTES}"""
Write under 150 words in plain English, cite the rule, note any assumption, and end with 'Let us know if your situation differs.' Mark anything uncertain as [PARTNER CHECK].

Create a one-page client worksheet for {FIRM NAME} on {TOPIC} for {AUDIENCE, e.g. new self-employed clients}. Include a short plain-English explanation (under 120 words), a fill-in table or checklist they complete with their own numbers, 3 common mistakes and 'questions to ask your bookkeeper'. Reading level: {GRADE LEVEL}. Don't state tax rates, deadlines or thresholds unless I give them here: {RATES AND DATES}; otherwise write [CHECK].

You write client emails for {AGENCY NAME}, a hometown independent insurance agency in {TOWN}. Our voice: {3 WORDS}. Two emails we have sent that sound like us: """{PAST EMAILS}"""
Write a {TOPIC, e.g. renewal reminder or storm-season coverage check} email under 150 words, with a subject line and one clear next step. Use only these CRM merge fields: {FIELDS}. Then give me the same email as a simple, mobile-friendly HTML template in our brand color {HEX}. Do not state any coverage, price or carrier detail I have not given you.

You write automated client messages for {AGENCY NAME}, a small independent insurance agency. Moment: {MOMENT, e.g. a renewal document just arrived in the client portal}. Write a friendly email under 90 words and a text under 160 characters that tell the client what happened and the one thing they can do next ({ACTION, e.g. log in to the portal to review it}), and invite them to reply with questions. Use only these merge fields: {FIELDS}. Do not mention premiums, coverage changes or carrier decisions.

What ChatGPT prompts help turn research into client deliverables?

Brian Davis runs a client's scenario through BlueJ for cited tax research, makes the decisions himself, then uses ChatGPT with a saved prompt library to turn the findings into slide decks, road maps and one-page PDF overviews. Evergreen Insurance's leadership team uses ChatGPT for Business for client communications, proposals and coverage comparisons. Ciara Siegel has Gemini transcribe her client conversations and drops the transcripts and questionnaire answers into ChatGPT, which pulls out how often and in what ways the client mentioned a word or idea.

Turn this CPA research summary into a one-page client overview with: situation, options, recommended next step, and open questions. Plain language, no jargon. Summary: {PASTE}

You help {AGENCY NAME}, an independent insurance agency, prepare a coverage comparison for a commercial client ({INDUSTRY}, no names). Quote summaries from {N} carriers: """{QUOTE DETAILS: carrier, premium, limits, deductibles, key exclusions and endorsements}"""
Build a side-by-side table (rows: premium, each limit, deductibles, notable exclusions and endorsements), then a plain-English summary under 150 words of the real differences. Use only the numbers above, mark anything unclear as [CHECK], and do not recommend a carrier.

You are helping me analyze research for {CLIENT NAME}, a {CLIENT TYPE}. Use only the uploaded interview transcripts and questionnaire responses. 1) List the 10 words or phrases they use most about {TOPIC, e.g. their customers or competitors}, with counts. 2) For {SPECIFIC WORD OR IDEA}, quote every mention with the speaker and file name. 3) Group the quotes into 3-5 themes and note any contradictions between people. Don't paraphrase quotes or add anything that isn't in the files.

What ChatGPT prompts help build SOPs and firm systems?

Katherine Pomerantz fed her client call recordings, emails and brainstorming sessions into ChatGPT, which turned her patterns into procedures; when a client asks a question like whether to use a Roth IRA, her administrator asks it for an overview of when she would recommend that strategy, and she then discusses the decision with the client. After getting a precise result, Brian Davis asks ChatGPT to turn it into a reusable prompt so next time takes one step. Lisa Chensvold told ChatGPT she works alone and values simplicity, and it helped her design a Notion database she uses as a "scrappy CRM", with Claude cross-checking the plan. Natasha Everard has a ChatGPT agent review her Outlook calendar each morning and report the gaps.

You are documenting how {OWNER NAME} at {FIRM NAME} works so a new {ROLE, e.g. administrator} can follow it. From the uploaded transcripts, emails and notes about {TASK OR CLIENT QUESTION}, write: 1) a numbered SOP with who does what, the tools used and how long each step takes; 2) the questions {OWNER NAME} always asks the client; 3) a 'how I decide' section listing the factors and when she'd recommend {OPTION A} vs {OPTION B}; 4) what must always go to {OWNER NAME} for approval. Cite the source file for each rule and don't add rules that aren't in the files.

You just helped me produce this deliverable: {OUTPUT}. Turn our conversation into a reusable prompt with placeholders so next time I get a similar result in one step.

I run {BUSINESS NAME}, a one-person {TYPE OF FIRM}. I want the simplest possible way to track prospects and clients in Notion. My stages are {STAGES, e.g. inquiry, call booked, proposal sent, signed, active, done}. Leads come from {LEAD SOURCES}. Suggest the database properties (name and type for each), two views and a weekly 10-minute review routine. Then write my client onboarding checklist from signed contract to kickoff using the steps I already do: {CURRENT STEPS}. Keep everything doable by one person.

Here is my calendar for today and this week: {CALENDAR}. List open blocks of 30+ minutes and suggest which of these tasks fits each: {TASKS}.

What ChatGPT prompts help a law or accounting firm with content and follow-ups?

Angel Reyes & Associates, a personal-injury firm, plans blog topic clusters around the questions people actually ask at intake; a six-step pipeline of custom GPTs drafts a post in about three minutes, with an 8,000-word style prompt for the firm's voice, and a human editor always has the last step. Mitch Zoll photographs a business card with the ChatGPT phone app, and his custom GPT reads it, looks at the person's role for a connection to his practice and drafts a follow-up email he reviews and sends.

You're drafting content for {FIRM}, a {PRACTICE AREA} law firm in {STATE}. Attached: our style guide and a list of real client questions from intake. Step 1 only: group the questions into 5 topic clusters. For each cluster give a pillar article title, 8–10 supporting article titles that each answer one specific question in plain language, and the intake question each one answers. Don't give legal advice or promise results. Flag any topic where the answer depends on state law so an attorney can review it.

I'm {YOUR NAME}, {WHAT YOUR FIRM DOES, e.g. a business lawyer in Austin who forms companies and reviews contracts}. Attached is a photo of a business card from {EVENT NAME}. 1) Pull out the contact details as: name, title, company, email, phone, website. Mark anything you can't read clearly as [CHECK]. 2) In one sentence, say how this person's role could connect to my work (referrals, their clients, or their own business). 3) Draft a follow-up email under 120 words from me that mentions {WHAT WE TALKED ABOUT}, offers to stay in touch and is not pushy about my services. Don't open with "I hope this finds you well" and don't invent details about the person or the conversation.

What should accountants, lawyers and insurance agents watch out for with ChatGPT?

• It can state the law wrong. Amanda Butler Schley of Business Law Group says a ChatGPT summary of a CLE invented the law, the post went live, and she had to pull it after another lawyer flagged it; she now questions every sentence (case).
• It tends to agree with you. Schley also says models tend to adopt your stance and feed it back, and the same question asked ten times gives different answers.
• Confidentiality. Todd McMurtry warns that free versions aren't confidential, so client information goes only into a paid version (case); Everard keeps client data out of ChatGPT and tells clients about her AI use in her engagement letter.
• Rules, rates and coverage. A partner checks every GWCPA reply, Davis makes the final decisions himself, and the insurance prompts here tell ChatGPT not to state coverage, price or carrier details you didn't give it.
• Getting the team to use it. Patrick Villafranca says getting the rest of Evergreen's team on board, and learning how to use it best, were the hard parts; Jack Tribble says hands-off automation is something his agency hasn't figured out yet (case).