AI sales prospecting is the use of AI to decide who to contact, when, and with what first sentence, and then to take the dialing, note-taking and logging off the rep so the rep can hold the conversation. It is not AI having the conversation. This guide covers where AI fits in the prospecting workflow, seven principles that lift meeting rates, what to keep human, and the European rules that frame all of it.

Where AI fits in the prospecting workflow

Prospecting is a chain of six jobs. AI is good at four of them and should stay out of the other two.

• Research. Reading the CRM, the website and the market to describe an account. AI does this well and fast.
• Signal detection. Noticing that a champion changed jobs, a lost deal hit its anniversary, a known account read the pricing page. AI is better than humans here because it never forgets to look.
• List building and timing. Ranking accounts by fit, trigger and reachability, and assigning them to the slot when the persona picks up.
• Call prep. A four-line brief and a suggested opener per contact.
• The conversation. Human. The prospect called a human, expects a human, and in Europe has a legal right to know if it is not one.
• Logging and follow-up. AI again: transcript, summary, next step in the CRM, follow-up email drafted.

The pattern is simple: AI on both sides of the call, a person in the middle.

Why this matters more in 2026 than it did in 2023

Two numbers from our benchmark of 1,000,000 B2B cold calls in France. 65% of reps open by explaining what they sell. Trigger-based openers reach a 7-10% meeting rate against 2-3% for pitch-first openers.

The trigger opener was always better. It was just too expensive to prepare for every call. AI removes that cost. That is the real change: not that reps can send more, but that every call can start with a reason.

The other change is volume. Parallel dialing gives reps about 5x more conversations from the same seat. More conversations with a better first sentence is the whole model.

The 7 principles of AI prospecting that works

1. Start from the CRM, not from the market

Reps use about 20% of the signals already in their CRM. Half of Sonik's pipeline came from signals that were already there. Point the AI at your own data before you buy anyone else's.

2. One trigger per contact

The brief states the one thing that changed. Five triggers is a research report. One trigger is an opener.

3. Personalize by situation, not by name

Inserting the first name and the company is not personalization, it is mail merge. The prospect's situation (new role, new hire, lost deal eleven months ago) is what earns the next 20 seconds.

4. Time the call by persona

Founders pick up most around 10am, marketing around 4pm, HR around 1pm. The best slot across personas is 4pm at 15.8%, against 9.8% at 8am. Let the AI assign the slot; let the rep protect it.

5. The AI drafts, the rep edits

Suggested opener, suggested follow-up email. The rep changes a word or two and owns it. A rep who reads an AI script aloud sounds like one.

6. Keep the questions human

Discovery questions starting with "how" lift conversion by 190% in our data. Asking "What do you mean?" before answering an objection lifts it by 197%. Those are judgment calls made in real time. No model makes them for you mid-sentence.

7. Close the loop every week

Call outcomes are training data. Which triggers booked, which personas picked up, which openers died. Feed them back into the list logic on Friday. An AI prospecting system that does not learn from outcomes is a list generator.

What to keep human, and the European rules

Cold outbound by synthetic voice is the wrong tool for B2B in Europe, and not only because prospects dislike it. The EU AI Act requires that people be told when they are interacting with an AI system. GDPR and the French rules on B2B prospecting require legitimate interest, a professional contact and a working opt-out. Call recording for AI coaching requires informing the people on the line.

None of this stops you using AI. It tells you where to put it: before and after the call, with the human in the conversation. A French prospect who picks up at 4pm and hears a person who knows why they are calling will give you 20 seconds. The same prospect hearing a bot will not give you two.

How to measure whether AI prospecting is working

Not dials. Not emails sent. Three numbers, per rep and per week:

1. Conversations per day. Our north star is 20 per SDR.
2. Share of conversations opened on a trigger. Target above 80%. This is the metric AI prep moves directly.
3. Meetings per conversation. If this does not rise with the share of trigger openers, the triggers are weak or the list is wrong.

Give it 30 days against a pre-launch baseline. In our experience European teams see the trigger share jump within a week and the meeting rate follow within a month.

How Sonik approaches AI sales prospecting

Sonik puts AI on both sides of the call and keeps the rep in the middle. Signals from the CRM, the website and the market build the daily list. AI call prep writes the brief and the opener for each contact. The parallel dialer runs the session and shows the brief at connect, with live coaching prompts during the call. After the call, the AI logs the summary and the next step.

What the rep does: talk, ask the "how" question, say "What do you mean?", book the meeting. That is where European teams double their meetings, from 5x more conversations with the right people.

Conclusion

AI prospecting is not automation of the conversation. It is removal of everything around it. Research, signals, timing, prep, logging: automate all of it. The call itself stays a person talking to a person. This week, measure one thing: the share of your team's calls that open on a trigger rather than a pitch. If it is under 50%, that is where AI goes first.