iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
These 10 AI prompts can help draft sales follow-ups and quotes, sort CRM records, qualify prospects, analyze lost deals, control retrieval agents, audit automation risks, trim reports, flag customer-data duplicates, and capture meeting decisions. They are designed to work from records you provide—not to invent missing details or take consequential actions on their own.
Archit Mittal’s August 15, 2026 article describes these as prompts he uses in practice, not as a measured productivity study. His guiding rule is: “Every one of these prompts drafts, reports or sorts. None of them send.” Read the original prompts.
How to use these prompts safely
Give the model the relevant source material—such as call notes, a rate card, a CRM export, customer replies, or a meeting transcript—and tell it what to do when that material does not answer a question. Constrain the output to a useful format, and keep a person responsible for consequential next steps.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsThe quality and completeness of the input sets a limit on the quality of the output. A model can organize supplied information, but a plausible date, discount, or customer detail is not reliable just because it sounds reasonable. Mittal’s examples are prompt-design guidance; they do not establish measured gains in revenue, conversion, accuracy, or time saved.
#1 Best Overall
Use the templates below as starting points. Replace the bracketed descriptions with actual records and adapt the channel or labels to your process. Review customer-facing drafts before sending, approve prices before quoting, and do not let a suggestion to delete or match a record become an unreviewed action.
What each workflow needs and produces
| Workflow | Input | Output | Review boundary |
|---|---|---|---|
| Sales follow-up | Notes or transcript from the last sales conversation | Three short message drafts | Review before sending |
| Quotation | Customer enquiry and seller’s rate card | Structured quote draft | Approve price and scope before sharing |
| CRM triage | Lead export and column names | REVIVE, PARK, DELETE, or UNKNOWN labels | Review deletion suggestions |
| Qualification | Prospect’s outcome, prior attempts, timing, and budget range | BOOK, NURTURE, or DECLINE recommendation and reply draft | Review decline before sending |
| Lost-deal analysis | Notes on what happened in a lost deal | Stated reason, supported interpretation, turning point, and earlier information need | Interpretation is not verified fact |
| Retrieval guardrail | Documents the agent is allowed to use | Answer grounded in those documents or a fixed fallback | Check important answers against the documents |
| Automation-risk audit | List of team decisions | Automation classification for each decision | Keep high-consequence actions under human control |
| Report cull | Recurring reports and recipients | DELETE, REPLACE, or KEEP recommendation | Ask recipients before removing a report |
| Customer-data cleanup | Customer-list export | Possible variants grouped by confidence, plus uncertain rows | Do not merge records automatically |
| Meeting decision log | Meeting transcript | Decisions, owners, due dates, and open items | Do not turn unresolved discussion into a decision |
1. Draft a three-message follow-up
Use this after a B2B sales conversation when a short sequence is appropriate. The first message should offer new information, not simply repeat the pitch. If there is nothing new, the prompt should stop rather than manufacture a reason to contact the prospect.
Using only these notes or this transcript from our last sales conversation: [insert conversation notes or transcript], draft three short WhatsApp follow-up messages.
- Message 1: Include genuinely new information. If there is none, output exactly: NO NEW INFORMATION — do not send.
- Message 2: Give a real reason for the prospect to make a decision. Do not create urgency or imply a deadline that is not in the notes.
- Message 3: Close the loop politely, without guilt or another pitch.
Supply the change or new information yourself; the model cannot know what has changed since the conversation unless you tell it. Review all three drafts before sending.
2. Prepare a quotation from a rate card
This prompt helps organize a quote while keeping pricing tied to the seller’s actual rate card. It should expose unanswered pricing questions and scope outside the approved list instead of filling gaps with invented numbers.
Rank #2
- COMPREHENSIVE GUIDE: A comprehensive guide covering all aspects of Sales & Operations Planning or Integrated Business Planning (S&OP/ IBP)
- IMPLEMENTATION & OPTIMIZATION: Practical guidance for implementing and optimizing S&OP processes, regardless of experience or maturity level
- HOLISTIC APPROACH: Covers people, processes, analytics, and metrics related to S&OP/ IBP
- EXPERT INSIGHTS: Incorporates perspectives from S&OP thought leaders and the Institute of Business Forecasting (IBF)
- MUST-HAVE RESOURCE: A valuable, must-have resource for anyone that wants to learn or involved in S&OP/IBP
Using only this customer enquiry: [insert enquiry] and this rate card: [insert rate card], draft a concise quotation with these sections: understood need, inclusions, exclusions, price, and validity period. Use only prices stated in the rate card. Flag any requested scope not covered by it. If information needed to price the work is missing, list the questions instead of guessing. Do not create a discount, price, or validity period that the sources do not establish.
A person should approve the price and scope before the quote becomes a customer-facing commitment.
3. Triage dormant CRM leads
For a CRM export, ask the model to label each record using the actual column names. Unknown is a useful result: sparse data should not be converted into a confident guess.
Review this CRM CSV export: [insert export]. The column names are: [list column names]. For each record, return one label—REVIVE, PARK, DELETE, or UNKNOWN—and a one-line reason based only on the available fields. If the record does not support a decision, use UNKNOWN. Draft a follow-up message only for records labeled REVIVE. Treat DELETE as a suggestion for a person to review, not an instruction to remove the record.
Check the underlying records before acting on a deletion recommendation or sending a drafted follow-up.
4. Qualify a prospect before booking a call
First gather four details in the conversation: the prospect’s desired outcome, what they have already tried, their timing, and their budget range. Then use the answers to prepare a recommendation and response.
Here are the prospect's replies about desired outcome, prior attempts, timing, and budget range: [insert replies]. Recommend exactly one next step: BOOK, NURTURE, or DECLINE. Explain the recommendation briefly using only these replies. Draft a response. If declining, be clear and warm; you may suggest another course of action if the replies support one. Do not infer missing budget or timing details.
Review a decline message before sending it, especially if the suggested alternative could be interpreted as a referral or endorsement.
5. Analyze a lost deal without overstating the evidence
A lost-deal review can separate what the notes explicitly say from a possible explanation. That distinction matters: an interpretation is a hypothesis about the supplied account, not independent confirmation of why the customer chose another option.
Analyze this account of a lost deal: [insert notes]. Return: (1) the reason explicitly stated, (2) a possible underlying reason only if the notes support it, (3) the point in the process where the deal appears to have turned, and (4) what would have been useful to know earlier. Separate stated facts from interpretations. If the notes do not support an answer, say so.
6. Make a retrieval agent expose gaps
When an agent must answer from a defined document set, this instruction makes the boundary and fallback explicit:
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Answer only from the documents provided. If the answer is not in them, reply exactly: “I don't have that — let me check and come back.” Never infer, estimate, or fill a gap from general knowledge. Quote the line you used.
This does not eliminate hallucinations. It aims to make missing information visible and gives a reviewer a line to check. Verify consequential answers against the cited document rather than treating the instruction as a guarantee.
Rank #4
7. Audit which team decisions should be automated
Give the model a list of decisions, not a vague request to automate a whole process. Ask it to consider whether each decision is reversible and assign one of three classifications.
For each decision in this list: [insert team decisions], assess whether it is reversible and classify it as one of: automate, automate with human approval, or never automate. Give a short reason for each classification. Treat money leaving the business, filings, and messages sent under my name as never automate unless I have explicitly authorized otherwise.
The categories are a review aid, not permission to execute. In Mittal’s rule of thumb, payments, filings, and messages under the user’s name stay out of autonomous execution unless specifically authorized.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.8. Decide which recurring reports still earn their place
For each report, name the decision it changes. If no decision can be identified, the model can recommend removal; if one number is enough, it can suggest replacing the report with a simpler update.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Review these recurring reports and their recipients: [insert report names, contents, and recipient roles]. For each, state what decision changes because of it. Label it DELETE if no decision can be named, REPLACE if a single number would suffice, or KEEP otherwise. Give a brief reason. Do not assume that a report is unused just because recipients do not discuss it.
Ask the recipients before deleting a report. A person may rely on it as a quiet check even when it does not prompt discussion.
9. Flag possible customer-record variants without merging
A cleanup prompt can surface spelling, abbreviation, phone-number, or city variations for review. The safe output is a set of possible matches and a separate set of records that cannot be judged—not a merged customer list.
Review this customer-list export: [insert export]. Group possible spelling, abbreviation, phone, and city variants. For each possible group, show the records and a confidence level with a short reason. Put rows that cannot be judged in a separate list. Report possible matches only; do not merge, delete, or edit any record.
Check suggested matches against reliable records before making a change. Similar names or contact details do not by themselves establish that two rows represent the same customer.
10. Turn a meeting transcript into a decision log
Ask for decisions that were actually made, with owners and due dates, and keep unresolved matters visible as open items. This avoids turning a long discussion into a list of commitments nobody agreed to.
Free tools Windows power users keep installed
One-click scans. No signup required.
Using only this meeting transcript: [insert transcript], return: decisions actually made, the owner and due date for each if stated, and what remains open. Omit discussion that did not resolve into a decision. If no decision was made, say so instead of summarizing the whole conversation. Do not invent owners or due dates.
Where to keep a person in the loop
These workflows are most useful when they make a person’s review easier, not when they silently turn a draft into an action. A practical boundary is to let the model prepare text, organize supplied records, or surface a recommendation, while a responsible person decides whether to send, quote, delete, merge, pay, or file.
- Check customer-facing messages for factual accuracy, tone, and unintended promises before sending.
- Verify quoted prices, exclusions, and validity against approved commercial terms.
- Review deletion and record-match suggestions against the source records; do not treat a proposed match as permission to merge.
- For retrieval answers, confirm that the quoted line supports the answer and that the relevant documents are current.
- When a source is silent, preserve the gap instead of rewarding a fluent guess.
Mittal’s closing principle captures the distinction: “Draft with AI. Send as yourself.”
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.

