10 June, 2026
ChatGPT for sales: 7 prompts for closing orders and customer care

ChatGPT for sales only works when you hand off work the way you would to a staff member who already knows the product, the customers and the brand voice. Used carelessly, the tool gives generic answers, drifts off in advice and loses your own character. In this article, MONA goes straight to how to use ChatGPT to write consulting scripts, handle objections, draft customer care messages, summarize conversations and create data for your sales team. We also point out the ceiling of standalone tools, so you know when a business needs custom software.
ChatGPT for sales pays off when you give it 3 things: a role, data and a goal

ChatGPT for sales is only worth the money when you do not hand it vague work. A prompt without context produces a generic answer that sounds right but is hard to use for closing an order.
The formula we use inside MONA is simple: role, data and goal. With all 3 layers, ChatGPT can write for the actual sales and customer care situation.
Start with the role. Then add the data. Finally, set the goal.
Assign a role that matches the job: consultant, sales team lead, post-purchase customer care
You need to say clearly who ChatGPT is playing in the conversation. The closer the role is to the real job, the less the answer rambles.
For example, a consultant needs soft, easy-to-understand wording that leads the customer to the next step. A sales team lead needs to spot the bottlenecks in a script. Post-purchase customer care needs to stay calm and protect the customer experience.
- Act as a consultant messaging a new customer from Facebook.
- Act as a sales team lead fixing a script for a new hire.
- Act as a customer care agent handling a customer complaint after purchase.
- Act as a sales trainer and point out which lines are likely to kill the customer's interest.
A role is a frame of thinking. Without one, ChatGPT tends to answer like a sample essay.
Give it real data: product, customer segment, price, offers, return policy
Data decides quality. ChatGPT cannot advise correctly if all you write is “write a better sales message”.
Give it real data before you ask it to write. Data includes the product, customer segment, price level, offers, purchase barriers and return policy.
- Product: name, use, what sets it apart, what needs a careful explanation.
- Customer segment: age, needs, worries, spending level.
- Price: list price, promotional bundles, conditions that apply.
- Offers: gifts, free shipping, warranty, bundled consulting.
- Policy: returns, refunds, processing time, warranty conditions.
We usually advise customers to create a fixed block of data. Each time you need to write a message, you only change the situation and the goal.
Set a clear goal: get a phone number, book a consultation, handle a refusal, win the customer back
The goal has to be clear. A sales reply should not carry too many jobs at once.
Say plainly what result you want after the chat. ChatGPT will pick a different way to lead the conversation for each goal.
- Get a phone number so you can advise faster.
- Book a call in a specific time slot.
- Handle a refusal because the customer says the price is too high.
- Win back a customer after 3 days of silence.
- Suggest an add-on purchase without feeling pushy.
One goal, one answer. This keeps the content tighter and easier to use in practice.
A base prompt template to reuse for every sales conversation
You can save the base prompt template below. After that, you only replace the parts in square brackets for each product and situation.
- Act as [sales or customer care role].
- My product is [product description].
- The customer is [customer segment description].
- The current price and offers are [price, bundle, gift, conditions].
- The return or warranty policy is [policy details].
- The current situation is [new customer, customer asking about price, customer says it is too expensive, customer who already bought].
- The goal of the reply is [get a phone number, book a consultation, handle a refusal, win the customer back].
- Write 3 versions of the reply, natural tone, short, no pressure to buy.
MONA uses ChatGPT and other AI tools internally to speed up drafting, test scripts and train customers to use them well. MONA has operated since 2016 with 14,000+ projects.
But standalone tools always have a ceiling. When sales data is scattered across messages, spreadsheets and internal files, ChatGPT can only support individual tasks.
To make sales, customer care and marketing run more smoothly, a business needs its own system. MONA usually settles on this direction with a website, SEO, CRM, ERP or which automation software built around the business's real data.

A 5-line formula that turns messy messages into a tight, persuasive closing script

Sales messages are often messy because customers raise many points at once. Staff answer by reflex, so the advice runs long but the closing power is weak.
The 5-line formula helps ChatGPT put the advice back in the right order. It is very easy to apply.
You just paste the raw chat into ChatGPT and ask it to turn it into a 5-line script. MONA often uses this when training marketing and customer care teams to write tighter replies.
- Sample prompt: “Read the chat below. Rewrite it as a 5-line closing script, natural tone, no pressure to buy, each line under 25 words.”
- Input chat: paste the customer's message and the earlier advice exactly as is.
- Extra requirements: “Keep the customer's meaning, do not exaggerate benefits, and close with one clear next step.”
Line 1: sum up the customer's real need
The first line has to pull the customer back to the main need. Do not open with a product introduction.
ChatGPT needs to read the chat and boil it down to one short sentence. That sentence shows the customer that your staff understood the problem.
- Weak line: “We have many plans that suit you.”
- Tighter line: “You need a plan that helps your sales team reply to customers faster and more consistently.”
This line cuts the noise. The customer feels heard.
Line 2: name the worry that is blocking the order
A customer who has not bought rarely lacks information. They are stuck on a worry nobody has named yet.
ChatGPT should look in the chat for signals such as price, timing, fit, rollout risk, or fear of buying something that goes unused.
- Extra prompt: “Point out the 1 biggest worry keeping this customer from deciding.”
- Sample line: “What you are weighing is whether your current team will keep using it after rollout.”
Do not dodge this part. Naming the real bottleneck cuts the going in circles.
Line 3: tie the product benefit to a specific situation
Generic benefits make it hard for the customer to picture. ChatGPT needs to attach the benefit to the exact context the customer just described.
Instead of saying “saves time”, the script should say which time, which task, and who is stuck. The more specific, the easier it is to believe.
- Weak line: “The software helps improve sales performance.”
- Tighter line: “When leads come in from the website, your sales team will have ready-made reply templates to handle them faster.”
For campaigns from website and SEO, this line matters a great deal. When a lead arrives on time, the reply has to keep the same pace.
Line 4: give just enough proof, without exaggerating
Proof gives the closing line weight. But exaggeration makes the customer defensive.
Ask ChatGPT to use soft proof if you do not have verified figures. For example the rollout process, screenshots, real feedback, a demo, or warranty principles.
- Extra prompt: “Only use proof that appears in the data I give you, do not make up numbers.”
- Sample line: “We can send you a sample screen so your sales team can preview how it runs.”
MONA always reminds its internal team to separate real data from interpretation. That boundary keeps sales content trustworthy.
Line 5: close with a gentle invitation to the next step
The last line does not need to pressure anyone. The goal is to move the conversation to the next step.
ChatGPT should write a specific, small invitation that is easy to say yes to. A short call, a demo or a comparison table is enough.
- Pushy line: “Close today to get the offer.”
- Softer line: “If it fits, I will send you a demo so your team can see how it is used first.”
A gentle close still has power, because the customer knows exactly what comes next.
This 5-line formula helps ChatGPT turn a long chat into an ordered script. You can use it for web consulting, SEO, advertising, CRM and customer care.
Standalone tools handle individual chats well. But when customer data, the website, CRM and the sales team need to run together smoothly, a business will need custom software. MONA builds that system for real automation, not just manual steps.

Make ChatGPT write 7 versions of a reply so every customer stops getting the same robotic voice

A very common mistake when using ChatGPT for sales is taking the first answer straight away. That answer is usually on target, but tends to sound monotone.
The tone drifts easily. A cold customer finds it too long. A customer comparing prices finds it unconvincing. An annoyed customer finds it lacking in tact.
You should ask ChatGPT for 7 versions, one for each customer mood. Then your sales team picks the right line and adjusts it to your brand voice.
We usually teach sales teams to use this prompt frame before replying to a customer:
- Write 7 versions of a reply for the same sales situation.
- Each version is for a different type of customer.
- Keep the tone natural and polite, with no overstatement.
- Each reply stays under 3 lines, suitable for Zalo.
- End every version with a clear next step.
A sample prompt you can use right away:
- Customer asks: “How much is this product? Another shop is quoting cheaper.”
- Write 7 versions of a reply.
- Include a cold customer, a price-comparing customer, a returning buyer, a customer who needs detailed advice, a customer not ready to buy yet, an annoyed customer and a customer who needs a fast close.
- Sincere advisory tone, no pressure to buy, no putting down competitors.
Short version for customers messaging fast on Zalo
Customers who message fast rarely want to read long replies. A 5-line answer can make them leave the conversation.
What a waste of an opportunity. You need a short, clear line with a direction for what comes next.
- Add this to the prompt: “Write a version under 40 words for Zalo, quick and concise.”
- Ask ChatGPT to avoid long explanations.
- Keep only 1 main benefit and 1 call to action.
Sample reply:
- “Here is our pricing based on your actual usage. If you tell me the quantity you need, I can quote the plan that fits better.”
Do not send it as is. Swap “I” for your salesperson's name if you want it to feel more real.
Detailed advice version for customers who need explanations
Some customers do not mind the price, but they need to understand why they should choose you. If the answer is too short, they feel the advice is superficial.
This group needs structure. ChatGPT should write in 3 parts: the need, why it fits, and the next step.
- Add this to the prompt: “Write a detailed advice version, split into 3 short points.”
- Ask it to give reasons, not just list features.
- Add a sentence confirming the need before making a proposal.
Sample reply:
- “Based on what you described, I think we should put durability and after-sales support first. Our price includes help choosing the right model, usage guidance and support when issues come up. I will send you 2 options so you can compare faster.”
This reply is longer, but it is orderly. A customer who needs explanations will find it easy to follow.
Soft version for customers comparing prices
A customer comparing prices does not always want the cheapest option. Many are just checking whether the quote is reasonable.
If your sales team reacts defensively, the conversation gets tense. If you cut the price too fast, margins wear away.
- Add this to the prompt: “Write a version that does not argue about price and does not criticize competitors.”
- Ask it to explain the difference in terms of value in use.
- Add a lower-tier option if the customer needs to optimize their budget.
Sample reply:
- “I understand you are comparing to pick a reasonable level. We do not go for cheapest at any cost, but focus on fit for your needs and after-sales support. If you want to optimize your budget, I can send a leaner option.”
Holding your ground matters a lot. The reply should be soft without lowering its own value.
Relationship-keeping version for customers not ready to buy yet
A customer who is not buying yet can still come back. But many sales teams end the conversation too abruptly.
This is the point where the relationship is won or lost. A well-judged closing line keeps your brand in the customer's memory.
- Add this to the prompt: “Write a relationship-keeping version, no pressure.”
- Ask it to leave behind a document, a price list or a time to follow up.
- Do not use pressure lines like “offer ending soon” unless it is true.
Sample reply:
- “Take your time to look around and make sure. I will resend the key information so you can check it quickly when needed. If you want to compare options again next week, I am happy to help.”
Close with an action. No pushing. No leaving things hanging.
Apology and resolution version when a customer complains
When a customer complains, a robotic reply makes things heavier. They need to be acknowledged before they hear an explanation.
Your sales team also needs safe wording. Speaking in a hurry can admit fault beyond the real scope, or leave the customer feeling brushed off.
- Add this to the prompt: “Write a version that apologizes, acknowledges, asks for more information and does not shift blame.”
- Ask it to split into 3 steps: apologize, verify, then explain how it will be handled.
- Do not promise what the operations team may not be able to deliver.
Sample reply:
- “I am sorry this experience left you unhappy. I have noted your feedback and need to check the order details to handle it properly. Could you send me your order number or the phone number used to order?”
This reply does not dodge the problem. It does not overpromise either. That is the safe approach when customer care handles a sensitive situation.
Inside MONA, we do not use ChatGPT as a place to copy answers and send them out in bulk. We use it to create variations, check tone and cut drafting time.
But standalone tools always have a ceiling. When customer data is scattered across Zalo, CRM, the website and internal files, staff still do many steps by hand.
For businesses that need smoother selling, MONA usually advises custom software tied to the website, CRM and the customer care process. ChatGPT supports the wording, while only a custom system holds the data, permissions and real automation.

Have the AI score its own answer before you send it, a small trick that cuts errors sharply
ChatGPT writes fast, but sales needs more than speed. One line with the wrong price, an overpromise, or an off-brand tone can leave a customer with the wrong idea.
A safer approach is to have ChatGPT act as a content checker before staff send anything to a customer. This does not replace the salesperson, but it helps you cut basic errors in every message.
Well worth doing. Very easy to apply.
Build a self-scoring table with 5 criteria: accurate info, clear benefit, right tone, no overpromising, a next step
Ask ChatGPT to score a reply against a 5-criteria table. Each criterion should get a score, a short comment and a suggested fix.
- Accurate info: price, policy, timing and support scope are not wrong.
- Clear benefit: the customer understands the real value, not just a generic pitch.
- Right tone: the reply keeps the brand voice, neither too cold nor too eager.
- No overpromising: staff do not commit to what operations may not be able to do.
- Has a next step: the customer knows whether to confirm, book, send information or pay.
You can use this prompt template:
“Score the reply below against 5 criteria: accurate info, clear benefit, right brand tone, no overpromising, has a next step. Give each criterion a score from 1 to 5. Then point out what needs fixing.”
At MONA, we often standardize this kind of checklist for sales, customer care and marketing teams. When everyone uses the same standard, reply quality is steadier.
Ask ChatGPT to flag sentences that risk causing misunderstanding
Many sales mistakes hide in sentences that sound perfectly ordinary. The problem is that the customer reads them differently and then expects more than you can deliver.
Have ChatGPT inspect every sentence that risks being misunderstood. Keep this part separate from the scoring.
- Sentences that could be read as a firm commitment.
- Sentences that make the customer think the price covers everything.
- Sentences that overstate results.
- Sentences missing the conditions that apply.
- Sentences that could blur the brand positioning.
Sample prompt:
“Read this reply as a demanding customer would. Point out sentences that could cause misunderstanding about price, commitments, timing, support scope or results. Briefly explain each risk.”
This part is very important. It lets staff see the error before the customer does.
Have ChatGPT rewrite a safer version that is still persuasive
Making it safe does not mean making it bland. The reply still needs a clear benefit, selling power and goodwill.
Ask ChatGPT to rewrite it to cut risk without cutting persuasiveness. The prompt should state the role and the limits.
Sample prompt:
“Rewrite this reply to be safer legally and operationally. Do not overpromise. Do not overstate results. Keep a friendly advisory tone, a clear benefit and a soft call to action.”
- Replace “guaranteed to increase orders” with “helps improve conversion when the data and process fit”.
- Replace “we do everything” with “we support within the agreed scope”.
- Replace “this price is all-inclusive” with “this price covers the items in the quote”.
- Replace “done in 3 days” with “expected 3 days after we receive all the information”.
This way of writing helps sales teams avoid slips. Customers also get clearer information.
Save the checklist as a template prompt for sales and customer care teams
You should not let each staff member work out their own way of using ChatGPT. When everyone types differently, results drift apart.
Save the checklist as a fixed template prompt. Sales and customer care teams only swap in the message content and run the check before sending.
- Step 1: paste the draft reply for the customer.
- Step 2: ask ChatGPT to score it against the 5 criteria.
- Step 3: ask it to point out sentences that risk causing misunderstanding.
- Step 4: ask it to rewrite a safer version.
- Step 5: the staff member reads it again and takes responsibility before sending.
A combined template you can use right away:
“Check the reply below before it goes to the customer. Score it on 5 criteria: accurate info, clear benefit, right brand tone, no overpromising, has a next step. Then point out sentences that risk causing misunderstanding and rewrite a safer version that is still persuasive.”
MONA uses this in internal training and when guiding customers on applying ChatGPT to their work. We always remind the team that AI only supports the check; the person in charge still makes the final call.
ChatGPT handles individual tasks well, but hits a ceiling when data is scattered. When you need CRM, CSKH, website, SEO and the sales process to connect, custom software is a better fit.
MONA builds custom software, designs websites and delivers SEO services around a business's real cash flow. AI then becomes part of the system, no longer a tool each person uses on their own.
One prompt that turns customer replies into a bank of ideas for SEO, ads and follow-up
Customer feedback is not just for customer care to read and handle case by case. It is real raw material for writing web copy, doing SEO, running ads and following up.
Do not waste this data. It is worth real money.
MONA usually guides you to collect what customers actually say before writing new content. ChatGPT does better when it is given the right sales context.
Paste 20 to 50 customer replies into ChatGPT, grouped by type
Group the feedback into clear categories. Do not paste every message in at once in no particular order.
A simple way to group them:
- Questions before buying.
- Reasons customers hesitate.
- Praise after using the product.
- Complaints or suggestions.
- Messages from customers who did not reply further.
Then paste them into ChatGPT using this structure:
“I sell [product or service]. Below is real feedback from customers. Please read and analyze it from a marketing, sales, SEO and follow-up angle.”
Then paste 20 to 50 replies below. That is enough for ChatGPT to see repeating patterns, but not so long that the answer gets diluted.
A small tip. Do not fix the customer's voice.
If a customer misspells something, keep it as is. The way customers word things is valuable data for headlines, ad copy and FAQ sections.
Ask it to sort into concerns, reasons for not buying, keywords and FAQs
After pasting the data, you need to push ChatGPT to categorize it. If you only ask “analyze this for me”, the result is usually generic.
The prompt we use in MONA's internal training looks like this:
“Sort the feedback above into 4 groups: main concerns, reasons the customer has not bought, keywords customers naturally use, and frequently asked questions. Present it as a table.”
You can ask for 2 extra columns to use right away in marketing:
- Handling priority.
- Suggested content to write.
The “keywords customers naturally use” part is very important for SEO. It is how real customers describe the problem they need solved.
Not every keyword sits in a tool. Many valuable phrases are in sales messages, comments and inboxes.
Turn each group into SEO article ideas and landing page content
Once you have the data groups, ask ChatGPT to turn each one into specific content. At this point the prompt needs to be tied to your web goals.
Sample prompt:
“From the table above, suggest SEO article ideas, H1 headlines, angles, FAQs and landing page copy. Prioritize the language customers actually used.”
For SEO articles, ask ChatGPT to separate clearly:
- Articles that explain the problem.
- Articles that compare options.
- Articles that handle objections.
- How-to articles.
- Articles that prove credibility.
For a sales website, this keeps the landing page from talking one way. The content starts answering exactly what the customer is stuck on.
This is a spot many marketing teams miss. The content looks good but never touches the reason customers have not bought.
MONA often uses this approach when advising on site structure, SEO and conversion content. ChatGPT helps gather ideas fast, but strategy still needs someone who understands the product.
Rewrite into follow-up messages for each customer group
Old feedback can also become a follow-up sequence. You do not need to send the same content to every customer.
Ask ChatGPT to write for each group:
- Customers who asked about price and then went quiet.
- Customers worried about results.
- Customers who do not yet trust the brand.
- Customers who need to check with the decision maker.
- Customers who bought before but have not come back.
Sample prompt:
“For the customer groups above, write 3 follow-up messages for each group. Natural tone, short, no pressure to buy, with a reason for the customer to reply.”
Also ask for versions for Zalo, Messenger and email. Each channel needs a different length.
Shorter is easier to read. Clearer is easier to reply to.
ChatGPT saves your sales and marketing teams many hours of manual idea gathering. But standalone tools still hit a ceiling when data sits apart across the web, CRM, ads and customer care.
When a business needs connected data, automatic grouping and follow-up based on real behavior, MONA goes the custom software route. ChatGPT then becomes one part of the system, not the whole system.
We still use ChatGPT and AI tools inside MONA. We also train customers to use them for the right jobs, then connect them to your web platform, SEO, CRM or automation when your business is ready.

ChatGPT works better for sales with a prompt library for each stage of the customer journey
Sales teams that use ChatGPT in scattered ways tend to produce answers with inconsistent tone, priorities and closing style. A strong person asks one way, a new hire asks another, and managers struggle to control the quality of advice.
What we do at MONA is group prompts by 5 customer stages. New customer asking, customer weighing options, customer about to buy, customer after purchase and returning customer.
This is easy to control. The team just picks the right stage, fills in the right data, and lets ChatGPT help write the first draft of the reply.
Prompt group for new customers to find the real need
New customers rarely state the whole problem. If staff ask rambling questions, the conversation slips away without producing any sales data.
Build a prompt group that helps ChatGPT suggest questions to uncover the need. The goal is to learn what the customer needs, what is holding them back, their budget and whether the purchase is near or far.
- Write 5 short questions for the consultant to understand what a customer asking about this product needs.
- Turn the chat below into a list of needs, worries and buying signals.
- Suggest a follow-up question that gets the customer to state their budget without feeling uncomfortable.
- Summarize the customer profile in 5 lines so the next consultant can pick up the conversation.
Do not ask too broadly. A good prompt includes the industry, product, price range and the channel the customer came from.
Prompt group for handling objections about price, timing and trust
A customer refusing does not mean the chance is gone. Most of the time they just lack a clear enough reason to move forward.
The prompt library should separate 3 common objections. Price too high, no need right now, and not trusting enough yet.
- Write 3 ways to reply when a customer says the price is high, in a soft advisory tone with no pressure to buy.
- Analyze the real reason behind the line “let me think about it”.
- Suggest an answer for a customer who does not trust the brand yet, prioritizing proof and clear commitments.
- Rewrite this reply to sound more natural, without arguing and without cutting the price in a hurry.
This is where orders are easily lost. MONA usually advises sales teams to save both good and bad answers so ChatGPT learns the right pattern.
Prompt group for closing a consultation or an order
Once the customer shows buying signals, staff need to go straight to the next action. Wording that is too long cools the customer off.
Prepare several gentle closing styles in this group. Closing on a call time, on sending a quote, on a deposit, on confirming the order.
- Write 3 closing lines to book a consultation today, polite and clear about the benefit.
- Turn this advice into a short order confirmation message.
- Suggest a way to remind the customer to pay without feeling pushed.
- Write a line that moves from consulting to booking, suited to Zalo and Messenger.
A closing line needs to be short. One message should carry only one main action.
Prompt group for post-purchase care and asking for reviews
A sale should not be the end. This is when the customer remembers the experience best, and is most likely to review and refer.
Have a post-purchase prompt group for each milestone. Confirming receipt, usage guidance, asking about the experience and asking for a review.
- Write a thank-you message after purchase, warm and not cliche.
- Create a 5-step usage guide for the product, easy for new customers to follow.
- Write a message asking for a review after the customer has used the product.
- Suggest how to handle it when a customer gives unhappy feedback.
This part directly affects trust. We usually put the post-purchase prompt group into internal training materials for the customer care team.
Prompt group for winning back past customers with offers and useful content
Past customers are often forgotten because sales teams are busy chasing new ones. This makes the cost of selling climb over time.
The prompt library should include a group for re-engaging past customers. Send offers, and also send content that reminds customers of their need.
- Write a message inviting a past customer back, with an offer but without looking like a clearance sale.
- Suggest 5 pieces of useful content to send to customers who have bought this product.
- Write a script to remind customers to rebuy based on the product's usage cycle.
- Sort the list of past customers into those to call, those to message and those to send content to.
ChatGPT does well at drafting and suggesting angles. But customer data, purchase history and follow-up status still need to live in your own system.
That is why MONA usually uses ChatGPT as a content support layer, not a replacement for CRM or sales software. When a business needs connected data, we design custom software so the sales team can use it consistently.
Standalone tools only speed up individuals; a custom system is what keeps your data and truly automates
Recognize the ceiling of manual use: scattered, dependent on whoever types, hard to control data
ChatGPT can help each salesperson work faster. However, as a business grows, this way of working leads to dependence on whoever enters data by hand and leaves data scattered. Managing and controlling data accurately becomes much more complicated.
Separate what ChatGPT does well from what needs custom software
ChatGPT is very handy for automating frequent customer care replies or drafting ad copy. But to optimize larger processes, such as customer management or full sales tracking, you need dedicated software integrated with a CRM or ERP system.
Connect web, SEO, CRM, customer care and reporting into one smooth process
A system that integrates the web, SEO, CRM, customer care and reporting creates a continuous, seamless flow of information. This saves time and also keeps data managed and updated in sync.
How MONA uses AI internally and trains customers to roll it out in real processes
MONA has built AI into its internal processes to improve work performance and guide customers in applying AI to theirs. This helps businesses reach real automation with secure data that stays in sync.
More related articles from MONA:
- AI that holds your sales team's hand: the director clones themselves

You can use ChatGPT yourself to draft variations of Zalo messages, emails and objection-handling scripts while staff still choose and edit each line. When you need it to read customer records from a CRM, suggest replies according to policy and then save the sales progress, we can build an integrated sales assistant for you; call 1900 636 648 to review one real conversation.
Right data is what closes the right step.
Related articles
Professional website
design services
Own a website with a beautiful, 100% exclusive design and a lifetime warranty, with the ability
to expand features flexibly as your business grows, starting today!































Free AI courses
AI tools at work
VI
EN


