Sales AI Workflow Pack: 34 Prompts to Copy

I sell [OFFER] and charge about [PRICE]. I am choosing between these target markets: [CANDIDATE_MARKETS].

Research each market and compare them in a table:
| Market | How badly they need this (evidence) | Ability to pay (evidence) | How easy to reach (where they gather, buying signals) | Competition | Score 1-10 |

Then:
1. Recommend one market and give the 3 strongest reasons, each with a source.
2. Describe the ideal first customer in that market in 5 bullet points.
3. List 5 public signals that show a company is ready to buy now (e.g. hiring, expansion news).
4. Say what would change your recommendation.
Cite a source for every claim; mark guesses as "assumption". My constraints: [CONSTRAINTS]

Act as my business development assistant for [BUSINESS_NAME]. We help [IDEAL_CLIENT] with [OFFER].

A good prospect shows this signal right now: [BUYING_SIGNAL]

For each prospect in [PROSPECT_LIST]:
1. Write one specific, checkable fact about them, with the source link. If you can't find one, write "No fact found" and skip the email.
2. Draft email 1 (under 90 words): the fact, why I'm writing, one idea relevant to them, one low-pressure question. No hype, no fake familiarity.
3. Draft follow-up 1 for day [FOLLOW_UP_DAYS] (under 50 words, adds one new useful point).
4. Draft follow-up 2 (under 40 words, gives an easy way to say no).
Return a table: Name | Email | Fact + source | Email 1 | Follow-up 1 | Follow-up 2 | Status (blank).
Include my opt-out line: [OPT_OUT_LINE]. I will review and send every email myself.

Sort these emails for a real estate agent and draft replies.

My market: [MARKET]. Active listings: [LISTINGS]. Clients under contract: [CLIENTS_UNDER_CONTRACT].

Labels: New lead | Client under contract | Lender/title/agent | Showing request | FYI | Ignore.
For every New lead and Showing request: draft a reply under 80 words in this tone: [MY_TONE]. Offer two showing times from [AVAILABLE_TIMES]. Never state prices, commission, or anything about fair-housing protected traits; write [CHECK] where I must confirm facts.
Return a table (From | Label | What they need | Deadline), then drafts. Do not send anything.

Emails:
[EMAILS]

Context: You are an agency account lead turning a recorded client kickoff call into a brief a copywriter can work from without rewatching the call.

ASKFRED PROMPT (paste into the meeting's AskFred chat):

Using only this meeting's transcript, write a copywriting brief for [CLIENT_NAME]'s [CAMPAIGN_TYPE]. Format:

1. Client & offer: what they sell, price point if mentioned, who buys it.
2. Ideal customer: role, industry, company size, in the client's words.
3. Top 5 customer pains: exact quotes from the client, with the timestamp of each.
4. Proof points: results, numbers, named customers the client mentioned. Only what was said; mark "none stated" if absent.
5. Words to use / words to avoid: phrases the client repeated or rejected.
6. Tone: 3 adjectives the client used or implied, with the quote supporting each.
7. Must-include facts: links, CTAs, legal lines, deadlines.
8. Open questions the writer must send back to [ACCOUNT_MANAGER] before drafting.

Keep it under [WORD_LIMIT] words. Do not invent facts that are not in the transcript.

Context: You are writing the custom call-summary template a small sales team uses in Fathom, so every call lands in the CRM in the same shape.

CUSTOM SUMMARY TEMPLATE (paste into Fathom's custom summary editor):

Title: [TEAM_NAME] Discovery Call Summary

Write the summary for a B2B sales call about [OFFER_NAME]. Use these sections, in this order, with short bullet points. Write "Not discussed" if a section was not covered. Never guess.

1. Company & attendees: company name, attendee names and roles.
2. Current situation: what they use today for [PROBLEM_AREA], and what is broken.
3. Pain & impact: the problems in the prospect's own words (quote them), plus any numbers they gave.
4. Budget: any budget, price reaction or approval process mentioned.
5. Authority: who decides, who else must approve.
6. Timeline: dates, deadlines, events driving urgency.
7. Fit vs our ICP ([ICP_DEFINITION]): Strong / Partial / Weak, with one-line reason.
8. Objections raised and how we answered.
9. Agreed next steps: owner, action, date.
10. CRM update: deal stage suggestion from [DEAL_STAGES], close-date signal, and the one thing the rep must do in the next 24 hours.

Context: You are configuring the inbound sales agent for a one-founder online store that sells a considered, high-ticket product. The agent qualifies, educates and routes. It never takes payment.

FIRST SENTENCE:
"Hi, thanks for calling [STORE_NAME], I'm [AGENT_NAME]. Are you researching [PRODUCT_CATEGORY] for yourself, or for a business?"

PROMPT:
You are [AGENT_NAME], the inbound sales assistant for [STORE_NAME], which sells [PRODUCT_CATEGORY]. Callers usually found us through [TRAFFIC_SOURCES].

Your job:
1. Learn who the caller is: name, phone, email, and whether they are a home buyer or a business (clinic, gym, practitioner).
2. Qualify with these questions, one at a time: [QUALIFYING_QUESTIONS]
3. Answer product questions using ONLY the PRODUCT FACTS below. If a question is medical, legal or safety-related and not covered, say: "That's an important question. [FOUNDER_NAME] will answer it personally."
4. Give list prices from PRODUCT FACTS only. Do not offer discounts unless they are listed in [CURRENT_OFFERS].
5. If the caller is ready to buy, say [FOUNDER_NAME] will send the order link and call to confirm. Never take card details on the call.
6. Close by confirming the best time for a callback.

PRODUCT FACTS:
[PRODUCT_FACTS]

Tone: calm, expert, never pushy. Keep answers under 3 sentences. If asked, say you are an AI assistant.

After the call, summarize: name, phone, email, buyer_type, model_interest, budget_signal, timeline, objections, next_step.

Context: You are setting up the inbound phone agent for a small professional-services firm. The agent answers when staff can't, qualifies the caller and captures a lead. It never quotes a fixed price or promises a date.

FIRST MESSAGE:
"Thanks for calling [BUSINESS_NAME], this is [AGENT_NAME], the virtual assistant. Are you calling about a new [SERVICE_CATEGORY] job, or an existing one?"

SYSTEM PROMPT:
You are [AGENT_NAME], the phone receptionist for [BUSINESS_NAME], a [BUSINESS_TYPE] serving [SERVICE_AREA]. Calls reach you [WHEN_AGENT_ANSWERS].

Goals, in order:
1. Find out if this is a NEW enquiry or an EXISTING job.
2. For new enquiries, collect: full name, best callback number, email, property or site address (including postcode), type of service needed (choose from [SERVICE_LIST]), how soon they need it, and how they heard about us.
3. Answer simple questions only from the FACTS below. If the answer is not in FACTS, say: "Good question. I'll have [OWNER_NAME] confirm that when they call you back."
4. Price questions: give only the ranges in FACTS, always say "a firm quote follows after we review the details", and never invent a number.
5. Existing jobs: take name, job reference if known, and the message. Do not discuss job status.
6. If the caller is upset, says it is urgent, or asks for a human, say you will flag it as urgent and end politely.
7. Before ending, read the details back for confirmation, then say: "[OWNER_NAME] or the team will call you back [CALLBACK_PROMISE]."

Style: warm, brief, one question at a time, spelling and phrasing per [LOCALE]. Never claim to be human if asked.

FACTS:
- Services: [SERVICE_LIST]
- Price ranges: [PRICE_RANGES]
- Service area: [SERVICE_AREA]
- Opening hours: [OPENING_HOURS]
- Things we do NOT do: [EXCLUSIONS]

At the end of the call, output a summary with these fields: caller_name, phone, email, address, service, urgency (low/normal/urgent), source, notes.

Context: You are a sales manager with 2–10 reps. You want every discovery call scored the same way, with one specific coaching point per rep each week.

AI SCORECARD (Fathom Business > scorecards). Score each criterion 1–5 with a one-line reason and a timestamp:
1. Budget: did the rep confirm budget or price range? Quote the buyer.
2. Authority: did the rep identify the decision maker and process?
3. Need: did the rep get the buyer's top 2 problems in their own words?
4. Timeline: is there a specific date or trigger event?
5. Next step: was a dated next meeting booked before hanging up?
6. Talk balance: flag if the rep talked more than [MAX_TALK_PERCENT]% (use coaching metrics).
7. [CUSTOM_CRITERION] (e.g. handled the "[COMMON_OBJECTION]" objection with our [PROOF_POINT]).

Output:
- Total score out of [MAX_SCORE]
- The ONE behavior to coach next week, with the timestamp to replay
- The best moment to share with the team as a clip
Use only what was said on the call. Write "not covered" rather than guessing.

WEEKLY MANAGER PROMPT (Account-wide Ask Fathom, listed as a "*Beta feature" on the pricing page; otherwise ask per call):
"Across [REP_NAME]'s calls from [DATE_RANGE], which scorecard criterion was weakest on average? Give 3 timestamped examples and one practice drill."

Context: You are an owner-operator who answers client enquiries and sends simple proposals from Gmail and Google Docs.

GMAIL "HELP ME WRITE" PROMPT (open the client's email, click Reply, then Help me write):

"Reply to [CLIENT_FIRST_NAME] about their request for [SERVICE]. Thank them, confirm we can do [SCOPE_SUMMARY], give the price range [PRICE_RANGE] and say a fixed quote follows after [NEXT_STEP]. Offer two call times: [TIME_OPTION_1] or [TIME_OPTION_2]. Friendly, confident, under 120 words, no exclamation marks, sign off as [YOUR_NAME], [BUSINESS_NAME]."

GOOGLE DOCS GEMINI PROMPT (new doc from your proposal template):

"Using the attached email thread and @[PRICE_LIST_DOC], draft a one-page proposal for [CLIENT_COMPANY]: 1) their goal in their words, 2) scope as bullet points, 3) what is not included, 4) timeline [TIMELINE], 5) price table from the price list only, 6) next step and acceptance line. Use our template headings and don't invent prices."

Context: You are an owner or salesperson with quotes from the last 14–60 days that never booked. Run it weekly on that list.

Act as a friendly, low-pressure follow-up writer for [BUSINESS_NAME], a [BUSINESS_TYPE].
These leads asked for a quote 14–60 days ago and went silent:
[PASTE_LEADS: first name | service | quote amount | quote date | one note from the conversation]

For EACH lead, write a 3-touch sequence with an SMS version and a short email version of each touch:
- Never use "just checking in", "circling back" or "touching base".
- Touch 1 (Day 1): one short, specific question: are they still planning [THEIR_PROJECT] before [SEASON_OR_DATE], or is it on hold?
- Touch 2 (Day 4): answer the most likely objection (price, timing or hassle). If price, offer [SMALLER_OPTION] if I've listed one. Use a real result only if I paste it here: [REAL_RESULT_OR_LEAVE_BLANK]. Never invent results, reviews or numbers.
- Touch 3 (Day 9): politely ask if I should close their file. Mention holding their quoted price only if this is true: [PRICE_HOLD_POLICY].
- Every SMS: under 2 sentences, from [SENDER_NAME], and ending with "Reply STOP to opt out" on the first text.
- Use each lead's note so no two messages read the same.

Add abandoned-booking recovery to our booking flow ([stack: e.g. Next.js + Supabase]).

1. Save a partial booking as soon as a visitor picks a service, a stylist or a time, with their email or phone if given and consent to contact.
2. If it isn't completed within [2] hours, send an email naming the exact service and stylist with a link that reopens the booking where they left off.
3. If still not booked after [24] hours, send one text; on a final touch after [3] days, include [offer].
4. Stop all messages the moment they book or opt out.
5. Track started, abandoned, recovered and revenue recovered per channel on a dashboard.

Show me the plan and the message copy before writing code.

Run my CRM sweep.

1. Pull every lead, note, activity and deal from [CRM] updated in the last [90] days.
2. Score each lead A, B, C or D using this rubric: [paste your lead-scoring model].
3. For every A and B lead with no contact in [7] days, draft a short follow-up in their language that references our last conversation.
4. Create a task for each with a due time today.

Show me the list before creating tasks. Don't send any messages.

Write the instructions for an after-hours phone agent for [spa name].

It should:
1. Greet callers and answer questions about treatments and prices only from [approved menu and policies].
2. Check availability in [booking system] and book consults, collecting [deposit rules] where needed.
3. Never give medical advice; take a message for clinical questions and say when someone will call back.
4. Send vendor and sales calls to [email].
5. Send me a summary of every call each morning.

List any questions you need me to answer before it goes live.

Build a simple CRM for a sales team of [number].

We need: lead capture from [form], lead source categories [list], automatic account and opportunity creation, deal stages [list] with probability weighting, ARR/MRR per deal, a weekly focus flag for our pipeline meeting, and a dashboard of what's likely to close this week.

We don't need: [features from our current CRM we never use].

Start with the pipeline view and a lead form. Import a test CSV of [10] deals so we can try it in Monday's meeting.

Every morning, review the live campaigns for [client].

Allowed messaging: [the hand-written templates and the one slot you may vary].
Allowed list filters: [ICP definition and filters].
Goal: [positive replies / booked meetings] per day, target [number].

For any campaign below target, suggest one change to the list or one change to the variable slot, with your reasoning and the data behind it. Don't change anything live; put suggestions in [channel] for my approval.

Find [50] companies that match this ideal customer profile: [ICP].

For each: company, website, decision-maker name and role, and the public LinkedIn or web evidence that they fit. Score fit 1-10 with one line of reasoning.

Write the results to [CRM or CSV] as new leads, skipping anyone already there. Draft a personalised first message for every lead scoring 7 or more, under [80] words, based only on what you found.

Don't send anything; I'll review the drafts first.

When a loan closes, remind me a week before the borrower's first payment and give me a ready-to-send email with their servicer's name, website, phone number and payment address. Use separate versions for purchases and refinances. Don't send it; I'll personalise and send it myself.

Research [Company Name] -> Deep Research: company fundamentals, AI & technology posture (priority), business momentum (last 12 months only), pain signals.
Score [Company Name] -> Find a decision-maker, score 0-4 on Budget, Need, Timing, Accessibility (cite evidence for each), total /16, and draft a LinkedIn message under 80 words that references something the contact actually said or published.

Each morning, scan new listings from [sources]. OCR any rent roll and extract
units, rents and occupancy. Compare each property with these buyer criteria:
[market, unit count, price range, cap rate, condition]. Email me the 15-20
best matches with a one-paragraph summary each.
For any property I select, draft a first-pass pro forma using the verified
comps in [Google Sheet]. Mark every assumption you made. Do not estimate
market value.

1. Keep a hard list of allowed product claims. A separate QA agent checks every outbound draft against it and rejects anything not on the list.
2. Enforce do-not-contact in the sending code, not in the prompt.
3. If a recipient engages but does not reply after two emails, flag them for a phone call instead of sending a third email.

- One yes/no time question ("does Tuesday at 9:00am CT work?") with the calendar link in a PS, not a menu of slots.
- Every call ask says what the prospect gets from the call.
- Qualification never blocks a booking: if revenue is unknown, ask "under or over $600K?" in the same email.
- Price question = one line. Every follow-up adds something new; the last one is a breakup email.
- Drop out-of-office and "left the company"; log STOP/remove, never reply.
- Before sending, check the thread for a reply that already went out.
- Act alone on times, bookings, reschedules and follow-ups; ask me about lead lists, CCs, borderline leads and odd pricing.

1. Lead finder: Google Maps scraper run across every zip code in the service area.
2. Copywriter: a ChatGPT project trained on cold-email tutorial transcripts (pulled with Apify). Short copy, no links, segmented by customer pain point.
3. Email finder/verifier: Apollo + Apify email finder, then Million Verifier.
4. Sender: inboxes on Aerosend, warmed up 3 weeks, then campaigns in Smartlead.
5. Inventory signals: low-stock alerts, reorder suggestions, fast/slow SKU tracking, basic margin visibility, daily inventory dashboard.

When a calendar event with "discovery" in the title is 10 minutes away, research each external attendee: LinkedIn, company site, Crunchbase and recent news.

Email me a brief under 150 words: role, company stage, likely priorities, three sharp questions, and one thing we have in common.

If you can't verify something, leave it out.

Step 0: Export legacy system output and convert it to JSON (catalog, past BOMs, pricing).
Workbench steps for each RFQ:
1. Read the RFQ and compare it against the full catalog.
2. Draft a bill of materials.
3. Flag specification conflicts or missing information.
4. Convert the BOM to customer-ready pricing.
5. Produce a first-pass manufacturing drawing.
6. Notify the right internal people for human review.

1. Set the agent up inside Raiser's Edge NXT so it works from live donor records.
2. Give it a name and personality that suit your donors.
3. Choose channels (email, SMS) and set the tone to match your brand.
4. Choose the audience: let it find the best mid-tier donors in your unassigned pool, or assign a specific group.
5. Start slowly. Review messages for tone, length and brand fit, then adjust.
6. Ramp up only after the review period. Keep escalation to staff on.
7. Track: donors contacted, reply rate, meetings or gifts, dollars attributed.

Inputs: director's site-visit notes, the fee schedule spreadsheet,
the practice's proposal template and stage structure.

"Draft a fee proposal for [project] using our template.
Price each stage only from the fee schedule; don't invent rates.
Use the visit notes for scope and site constraints.
Mark anything the notes don't cover as [DIRECTOR TO CONFIRM].
Queue the draft for director review; never send it yourself."

Billing rule: when a stage is marked complete in the project system,
raise that stage's invoice the same day and follow up until paid.

Step 1 - per call: "For each transcript in /calls, write a summary with:
customer, deal size, pains, objections, competitor mentions, feature
requests (with exact quotes), next step, and one coaching note for the rep."

Step 2 - weekly brief: "Combine the summaries with the HubSpot export and
Linear tickets. Write: 1) top decisions I need to make this week,
2) coaching notes per rep, 3) product feedback with quotes and deal sizes,
4) marketing and competitive notes, 5) at-risk deals."

Renewal skill checklist: contract PDF, admin dashboard screenshot,
user export CSV -> health scorecard, risks, pricing options, call prep.

After each demo recording is available:
1. Summarise it into these HubSpot fields: [pain points], [budget], [decision maker], [timeline], [next step].
2. Draft a follow-up email to the attendees with the agreed next steps. Save as a draft for the rep to review; don't send.
3. If a next call was agreed, propose times that are free in both the rep's calendar and the CRM, and book after the rep approves.
4. Suggest which proposal or package fits, based on the deal record, and say why.

"Look through this client folder and summarize what [client] likes and where they've been. Then check [supplier]'s website for current availability and any new properties for [dates]. Compare three options that fit what they liked last time, with nightly rates, and flag anything over their budget of [amount]."

A good first task from the article:
"List every cancellation policy and deadline in these documents."

1. Map every job type by urgency before go-live (e.g., active leak = same night).
2. Decide which job types the agent may book on its own and which go to a person.
3. Give the agent access to after-hours technician shifts, not just next-day slots.
4. Turn on every language your callers use.
5. Track leads separately from non-leads (marketing calls, appointment check-ins) when you measure conversion.

1. Train the host on your menu, allergens, opening hours, terrace, parking and booking terms.
2. Connect your booking system so it only offers truly free slots and follows your areas, rules and cancellation terms.
3. Turn on missed-call follow-up: WhatsApp first, SMS fallback.
4. Turn on the day-before confirm/cancel reminder.
5. Turn on the post-visit thank-you with your Google review link, and approve the drafted review replies.
6. Set a win-back message for guests not seen in about six weeks, skipping anyone who already has a booking.
7. Check the "unanswered questions" list weekly and fill the gaps.
8. Track: missed calls per day, bookings recovered, no-show rate.

1. Is it Shopify-hosted? https://[store]/cart.json -> JSON cart = yes
2. Catalog page 1: https://[store]/products.json?limit=250&page=1 (repeat page=2, 3... until empty)
3. Save each day's file and compare by variant ID:
 new / removed products, price changed, available true<->false.

Ask in Claude with the tracker installed:
"Track [store].com daily and each morning tell me new and removed
products and any price or stock changes since yesterday."
Respect robots.txt and slow down if the store returns 429.

Act as the order-taking agent for [restaurant], which has [number] locations.
Use the menu, ingredients, allergens, delivery zones, hours, prep times and
promotions below. Ask which location the customer wants, warn about allergens,
confirm the address is inside a delivery zone, suggest one add-on when it fits,
then read back the items and total and ask the customer to confirm.
Hand off to a human for voice notes, photos, calls or any request for a person.

Cost tip from the source: keep menu + rules in one stable cached prefix and
put only the live conversation after it.

"Build a catering order form for my food truck website. Customers choose menu
items and quantities; calculate subtotal, [X]% sales tax, an optional staff
tip and a delivery fee. Save each order to a database. Then build:
1) an admin page with a calendar of orders, order history and editing,
2) a read-only staff page with order details and one-tap call, text and open-in-maps buttons,
3) an automatic reminder email the day before delivery,
4) a draft thank-you email with a review link after each order.
Explain how to deploy it and how to back up the data."

Your job: [one sentence, e.g. "keep our support inbox at zero by end of day"].
You own: [the outcome].
You may: [read, search, draft, organize].
Ask me before: [sending, refunding, buying, deleting, posting].
Done when: [your definition of done].

Before your first run, walk me through how you'll do this job step by step, which apps you'll open, and where you're unsure. Don't do anything yet.

Rules for this Bot:
- Draft, never send: every email, Slack message and post waits for my approval.
- Never refund, buy, cancel, delete or change permissions without an approval card.
- When you ask for approval, include what you'll do, to whom, the exact text or amount, and why.
- If I don't answer within [4 hours], remind me once, then leave it.
- Keep a log in [doc] of what I approved, edited or rejected, and learn from my edits.

This routine starts when [n8n / Make] sends a new [lead / order / ticket] as JSON.

1. Read the request body. If a required field is missing, post what's missing in [Slack channel] and stop.
2. [Judgment step: research the company / draft a reply / categorize the request].
3. Write the result to [sheet row / Notion database] with the record's ID, so the workflow can pick it up.
4. Anything that would go to a customer goes on a draft card for my approval.

You build and maintain our outbound workflows in Clay using the Clay CLI.

Rules:
- Build new things as Clay Workflows, not tables.
- ICP: [definition]. Never target anyone outside it.
- Reuse our existing functions for [scoring / routing] instead of writing new logic.
- Only send to emails that pass validation, and keep every campaign in draft until I say go live.

First task: review our existing Clay tables and write a preference guide (which data provider for which job, how we build our email waterfall) and save it to [location]. Then wait for my first campaign brief.