BlogCRM & Sales

How to Use AI for Sales Prospecting: Prompts and Rules

In short

Use AI for the slow, checkable parts of prospecting: researching each account, finding a reason to write now and drafting a first message you then rewrite. Keep choosing who to target and checking every fact yourself. In Salesforce's 2026 State of Sales survey of 4,050 sales professionals, 55% said they use AI for prospecting. LinkedIn and X both restrict automated messaging.

Which prospecting tasks should you hand to AI?

Hand AI the research and the first draft, and keep the targeting decision and the fact check for yourself. The table splits outbound prospecting into its usual steps; if you want the steps themselves explained, start with our plain guide to what sales prospecting is.

Prospecting stepWhat AI does wellWhat you keep
Choosing who to targetSuggests roles, industries and company sizes from your website and past customersThe final call, and who to exclude
Building the listFinds companies and people that match a written descriptionA spot check of 10 names before you use the list
Researching each accountTurns the company site, news, job ads and the person's posts into a short briefOpening every source it cites, and dropping anything you can't find
Finding a reason to write nowPicks the one recent fact that connects to the problem you solveWhether the fact is recent and really about that problem
Drafting the first messageWrites a short draft from the briefThe first line and the question, rewritten in your words
Sending and follow-upSome tools send on a scheduleThe platform rules on automated messages (last section) and opt-out handling
Replies and callsSummarises replies and suggests a next stepThe conversation itself

Inbound work, such as scoring form fills and booking the call, is covered in our seven AI lead generation workflows. This post stays with outbound: finding people who haven't heard of you and giving them a reason to reply.

How many sales teams use AI for prospecting?

Just over half of sales professionals, in Salesforce's 2026 survey. In Salesforce's 2026 State of Sales announcement, 55% of sales professionals said they use AI for prospecting and another 38% said they plan to. The double-anonymous survey ran from August to September 2025 and covered 4,050 sales professionals in 22 countries, including the US, the UK and India.

The same survey gives the reason teams are trying it. Sellers devote nearly one full day of their workweek to prospecting, and 48% say they lack the bandwidth to do adequate cold outreach. Sellers also expect agents, once fully implemented, to cut prospect research time by 34% and email drafting by 36%.

Read these numbers with two cautions. Salesforce sells AI agents for sales, and the 34% is the sellers' own forecast; the page reports no measured saving. Its finding that high performers are 1.7 times more likely to use agents for prospecting compares sellers who substantially grew revenue with those whose revenue stayed flat or fell. That shows the two go together; it doesn't show the agents caused the growth.

Can you trust what AI says about a prospect?

Treat anything AI says about a prospect as unchecked until you have opened the source. OpenAI's help page on ChatGPT's accuracy lists incorrect dates or facts, and fabricated quotes or references, among the errors it can make, and tells users to verify important information from reliable sources.

The largest test we found is about news, which is close to what prospect research looks for: a funding round, a new hire, a launch. In the EBU and BBC study published on 22 October 2025, journalists from 22 public service media organisations in 18 countries rated more than 3,000 answers from ChatGPT, Copilot, Gemini and Perplexity. 45% had at least one significant issue, and 20% had major accuracy issues, including hallucinated details and outdated information. The study tested news questions, not company research, so read it as a warning; it gives no error rate for prospect briefs.

What prompts should you use for AI prospecting?

Use three short prompts in order, in an assistant with web search turned on so it can cite pages. Fill in the brackets; the sample names are made up.

Sample output for Priya at Acme, a sample company: "Hi Priya, saw Acme is hiring two people to run client reporting. We built a tool that pulls those reports together for agencies your size. Worth a look before the new hires start?"

A "no reason yet" answer is useful: put that person on a watch list instead of sending a generic note. Then run the source check and rewrite the first line yourself. Expandi's own advice, covered in our post on AI-written versus human LinkedIn messages, is the same: use AI to draft, then rewrite the draft.

How much time does AI save on 50 prospects a week?

About half the hands-on time in this example, as long as you keep the source check. The minutes are assumptions for a sample founder at Northwind Ltd; time your own first 10 prospects and replace them.

StepBy hand (minutes per prospect)With AI (minutes per prospect)
Research the account and person81 (run prompt 1)
Check the sources0 (done while researching)3
Pick the reason to write21
Write the first message42 (edit the draft)
Total per prospect147
50 prospects a week14 × 50 = 700 minutes, about 11.7 hours7 × 50 = 350 minutes, about 5.8 hours

The saving is 700 − 350 = 350 minutes, about 5.8 hours a week, or 50%.

Now drop the check. The AI column falls to 4 minutes a prospect, saving another 3 × 50 = 150 minutes, or 2.5 hours. But if one brief in ten carries a wrong or stale fact (again an assumption), 50 × 0.1 = 5 people a week get a message built on something untrue. In a niche market of a few hundred buyers, those 5 people are hard to win back, which is why the check stays in.

Which AI sales prospecting tools fit a small team?

Choose by who presses send. Three kinds of tool cover most small teams:

Tool typeExampleWho sendsFits when
Chat assistant with web searchChatGPT with search on, using the prompts aboveYou, by handYou're still testing who to target, or want to write and send every message yourself
AI built into a platformLinkedIn Sales Navigator's Message Assist (AI-written outreach) and Account IQ (AI account insights), which its plan comparison ticks on all three plans, Core, Advanced and Advanced Plus; Core starts at US$119.99 a month per licence, or US$1,079.88 a year (checked 5 Nov 2026)You, inside LinkedInYour buyers are on LinkedIn and you already pay for Sales Navigator
AI SDR agentThe tools in our comparison of AI SDR tools for small businessThe agent, within limits you setYou have a message that gets replies by hand and want it running every day

Siela's Lead Finder is in the third group. It looks for a reason to reach out now, scores each lead hot, warm or cold with a one-line reason, and sends on its own once you press Save and start, within your daily goals. Its "Find leads only" setting saves the leads to the CRM and sends nothing, which fits the by-hand check described above.

What do LinkedIn and X say about AI that sends messages?

Both restrict automated activity, whoever wrote the words. LinkedIn's User Agreement, section 8.2, says members must not use bots or other unauthorised automated methods to send or redirect messages, or to add or download contacts. Its help page on prohibited software and extensions adds that members who use such tools risk having their accounts restricted or shut down. The rule names automated methods; LinkedIn itself offers AI drafts in Sales Navigator.

X's automation rules (updated April 2026) say sending automated replies to posts based on keyword searches alone is not permitted, that you may not send unsolicited direct messages in a bulk or automated manner, and that any AI reply bot needs prior written and explicit approval from X. The page also warns that scripting the X website may result in permanent suspension.

For email in the US, the FTC's CAN-SPAM compliance guide says the law makes no exception for business-to-business email, and that hiring another company to handle your email marketing doesn't remove your own legal responsibility. This is a summary of official pages checked on 5 November 2026, not legal advice. For how LinkedIn enforcement plays out in practice, read what LinkedIn allows and what gets accounts restricted.

Frequently asked questions

Can AI do sales prospecting for you?

It can do most of the research and the first drafts, and AI SDR agents can also send and follow up within limits you set. Choosing who to target, checking each fact and handling replies still need a person. Sending is also where platform rules bite: LinkedIn and X both restrict automated messages, so check them before you let any tool send for you.

What is the best AI for sales prospecting?

There isn't one answer; it depends on who presses send. A chat assistant with web search suits research you send by hand. Sales Navigator's built-in AI suits teams already paying for Sales Navigator. An AI SDR agent suits a message that already gets replies. Compare tools on pricing basis, channels, daily limits and whether you can stop it sending.

Is it legal to use AI to write cold emails?

Who wrote the email doesn't change the rules it must follow. In the US, the FTC says CAN-SPAM makes no exception for business-to-business email, and you must honour an opt-out request within 10 business days. The UK and EU have their own rules for business email. This is a general summary, not legal advice.

How do I stop AI making up facts about a prospect?

Ask for a link and a page date next to every fact, and tell it to write "not found" instead of guessing. Then open each link yourself. Drop anything older than about 90 days or that you can't find on the page. Never let an unchecked fact into the first line, because that is the line the prospect reads first.

How much time can AI save on prospecting?

Sellers in Salesforce's 2026 survey expect agents, once fully implemented, to cut prospect research time by 34%, though that figure is the sellers' own forecast. In our worked example with assumed minutes, AI cut hands-on time from 14 to 7 minutes per prospect, saving about 5.8 hours a week on 50 prospects, with a 3-minute source check kept in.

Sources

  1. Salesforce, State of Sales report announcement 2026 (checked 5 Nov 2026)
  2. OpenAI Help Center, Does ChatGPT tell the truth? (checked 5 Nov 2026)
  3. BBC Media Centre, EBU research on AI assistants and news content (22 October 2025)
  4. LinkedIn Sales Navigator product page (checked 5 Nov 2026)
  5. LinkedIn User Agreement (checked 5 Nov 2026)
  6. LinkedIn Help, Prohibited software and extensions (checked 5 Nov 2026)
  7. X Help Center, Automation rules (checked 5 Nov 2026)
  8. FTC, CAN-SPAM Act: A Compliance Guide for Business (checked 5 Nov 2026)
  9. LinkedIn Sales Navigator, Compare plans (checked 5 Nov 2026)
Danish Khan

Danish Khan

CEO & Founder, Siela

Danish Khan is the CEO and founder of Siela AI, which gives small teams AI marketing agents that find high-intent leads, reach out on LinkedIn, X and email, and run their social.

Connect on LinkedIn

Published