In this article
AI for sales prospecting saves real time on research, first drafts and call prep, as long as you feed it true facts about each prospect and read everything it writes. Used as a machine to blast thousands of generic emails, it does the opposite of what you want: messages nobody reads and a damaged sender reputation. Below: where AI helps, where it hurts, a before/after example and a prompt template you can copy into any chatbot.
Where AI genuinely helps
- Researching the target: list the typical pains of a trade (a garage, a beauty salon, a restaurant), the months when it buys, the words owners actually use.
- Personalised first drafts: from facts you give it (rating, number of reviews, no website, city), it writes in ten seconds what would take you five minutes.
- Subject line variants: five or six short options to test, which you then shortlist.
- Call scripts and objection prep: "I'm busy", "I already have someone", "just send me an email". The AI lists likely objections and a one-line answer to each.
- Summarising replies: paste a long reply and ask "interested, not now, or no?" plus the next action.
- Sorting a list: group records by priority on criteria you set (no website, many reviews, high rating).
Where it backfires
Invented facts
A language model fills gaps. Ask for an email to "a garage in Lille" with nothing else and it may invent an owner's name, a speciality or a customer review. One false detail and the prospect knows nobody looked at their business. Simple rule: every fact in the message comes from your data, not from the model.
Generic tone
"I hope this email finds you well", "innovative solution", "boost your online presence": default AI output all sounds the same, and shop owners already get plenty of it. Fix the style in the prompt (short, plain, no adjectives) and in your own edit.
Volume for its own sake
A large volume of loosely targeted messages drags reply rates down and can push your emails into spam. Thirty precise messages a day beat 500 rough ones; our piece on cold email reply rates explains why.
GDPR and public chatbots
Don't paste personal data you have no reason to share into a consumer chatbot: a named email address, an owner's name and mobile number, a private email thread. Under French CNIL guidance, a generic address like contact@ is not personal data; firstname.lastname@ is. Work from the business's public facts (trade, city, rating, reviews), and see our guide to GDPR and B2B cold email in France. This is not legal advice.
Worked example: one garage, two emails
Take an invented business: Garage Delcourt Auto, an independent garage rated 4.6 from 87 reviews, with no website. You sell simple websites to tradespeople.
The generic email ("write a cold email selling a website to a garage"):
Subject: Boost your online presence!
Hi, I hope this email finds you well. In today's digital age, every business needs a strong online presence. Our agency offers innovative, tailored solutions to help you reach your goals. Would you be available for a 30-minute call this week?
Nothing in it is about this garage: straight to the bin.
The email built on real facts:
Subject: 87 reviews, no website
Hello, Garage Delcourt Auto has 4.6 stars from 87 reviews, so your customers clearly recommend you. But when someone nearby looks for a garage, there is no site where they can see your prices or book. I build one-page sites for garages: opening hours, services, your best reviews and a call button, live within a week. Shall I send you an example I made for another garage?
A checkable fact in the first line, a problem that follows from it, a small ask. More structures in our cold email templates.
This situation is common: in our database, 28% of garages have no website.
A prompt template to copy
This works in any general-purpose AI assistant. Replace the brackets and use public facts only.
You are a salesperson writing to owners of small local businesses. Write a cold email in English, 80 words maximum, no filler phrases and no marketing adjectives. What I sell: [YOUR OFFER IN ONE SENTENCE]. The business: [NAME], [TRADE] in [CITY], rated [RATING] from [NUMBER] reviews, website: [YES/NO]. Use only these facts and invent nothing. First sentence: a fact about the business. Then one concrete problem linked to that fact, then a simple yes/no question as the call to action. Also give 3 subject lines under 6 words.
For a phone call, swap the instruction for "write a 30-second call script and the 3 most likely objections with a one-sentence answer each". For Instagram, ask for two sentences and no link; our guide to Instagram DM prospecting covers the rest.
Task by task: what AI does, what you check
| Task | AI useful? | Human check |
|---|---|---|
| Research on a trade or a city | Yes, for a first overview | Cross-check with what you know on the ground |
| First draft of an email | Yes, with facts you supply | Re-read every fact and the tone |
| Subject line variants | Yes | Pick, then measure replies |
| Call script and objections | Yes | Say it out loud, cut what sounds fake |
| Summarising replies | Yes | Confirm the intent before filing |
| Sorting a list | Yes, on criteria you set | Spot-check a sample |
| Choosing recipients and volume | No | That is your call |
| Fully automatic mass sending | No | Stay at a volume you can review |
How Prospya does it from the business record
The slowest part with a chatbot is gathering the facts. In Prospya they are already on the business record: trade, city, rating, number of reviews, whether it has a website. The AI message writer uses them directly. You describe once what you sell, then on any record you generate a cold email, an Instagram DM or a call script built on those real facts.
- 30 messages a day, free plan included (25 free credits every month, no card).
- You read and edit before sending: nothing goes out on its own.
- Every email ends with an unsubscribe link.
Find your targets first with the free filters, for example garages without a website in Lyon or beauty salons in Paris. Email campaigns from Prospya need a paid plan, from €19 per month for 1,000 credits, no commitment, cancel anytime; details on the pricing page.
FAQ
Can I use ChatGPT for sales prospecting?
Yes, for first drafts, subject line variants, call scripts and summarising replies. The quality depends on the facts you give it: without real data about the prospect, it writes something generic or invents details. Read everything before sending, and don't paste unnecessary personal data, such as a named email address or a mobile number, into a consumer chatbot.
Does an AI cold email generator replace a salesperson?
No. It saves writing time, but choosing the target, shaping the offer, setting the volume, editing and above all the call or meeting stay human. The best results come from someone who feeds the tool precise facts, cuts what sounds fake and sends fewer, well-targeted messages rather than a large automated volume.
Can prospects tell an email was written by AI?
Often, yes, when the text is generic: empty pleasantries, marketing adjectives and no detail about the business. A short message that cites a checkable fact, like the rating or the missing website, and asks a simple question reads like a hand-written email, especially once you have reviewed and adjusted it.
How much does Prospya's AI writer cost?
It is included in every plan, the free one too, up to 30 messages a day. Unlocking a business's contact details costs 1 credit, never twice. The free plan gives 25 credits every month; paid plans start at €19 per month with no commitment, and unused credits carry over for up to 2 months.