Cookies on softitland.com

We use essential storage to run this site. With your permission we would also like to use analytics cookies to see which pages help visitors, and marketing cookies to measure our ads. You can change your choice at any time with “Cookie settings” in the footer. Privacy Policy

AI & automation

Lead-generation AI agents: what they can do and the rules they must follow

What a lead-generation AI agent can do, from sourcing to CRM routing, and the GDPR, CAN-SPAM and LinkedIn rules it must follow before any outreach.

Cover image for the article Lead-generation AI agents: what they can do and the rules they must follow
Key takeaways
  • A lead-generation agent can source, research, enrich, score and route prospects into your CRM, but a person should approve every first message.
  • Under the GDPR, people whose data you did not collect from them must be told who you are, why you hold their data and where it came from, at the latest in your first message.
  • Anyone can object to direct marketing at any time under GDPR Article 21, so an automatic suppression list is part of the build, not an extra.
  • CAN-SPAM makes no exception for business-to-business email: identify yourself, include a postal address and honour opt-outs within 10 business days.
  • LinkedIn's User Agreement bans scraping profiles and using bots to add contacts or send messages, so keep prospecting agents off LinkedIn.

What can a lead-generation AI agent do, and which rules apply?

A lead-generation AI agent can find companies that fit your ideal customer, research them, enrich and score the records, and route the good ones into your CRM with a summary. It should not send first messages on its own: a person approves every first outreach. The rules come from data protection law such as the GDPR, anti-spam law such as CAN-SPAM, and the terms of every platform the agent touches, such as LinkedIn's User Agreement.

This guide covers outbound prospecting. For leads that arrive through your own forms, see our guide to the first tasks to give AI agents. This is general information, not legal advice.

What does the prospecting workflow look like?

A reliable agent runs as a chain of small steps, each with a clear input, output and log entry, rather than one long prompt.

StepWhat the agent doesWhat a person decides
SourcesPulls companies from your CRM history, a licensed data provider or public company websitesWhich sources are approved
ResearchReads the company's public pages and notes what it sells, who it serves and why it might fitWhat counts as a buying signal
EnrichmentAdds company fields and a business contact, recording the source of every fieldWhich fields you actually need
ScoringScores the company against your written ideal customer profile and explains the scoreThe criteria and the cut-off
CRM routingCreates or updates the record, assigns an owner and attaches the research summaryWho owns which segment
DraftWrites a first message with the required notices and an opt-outSend, edit or reject

A typical build runs on n8n or Make, with an AI model for research and drafting and the CRM's API for routing. Our workflow automation team builds the pipeline, and our CRM and lead systems team sets up the fields, owners and pipelines in HubSpot or GoHighLevel.

Why should a person approve every first message?

A model can misread a website or mix up two companies, and a false claim in a first email loses the prospect. The legal notices in that message must be right every time. And an agent that sends at volume can break anti-spam rules and platform terms before anyone notices.

So the agent puts each draft in a review queue. The salesperson checks the research, edits the message and sends it from their own mailbox, or rejects the lead with a reason you can use to tune the scoring. One step should be fully automatic: every objection or unsubscribe goes onto a suppression list the agent checks before it researches or drafts anything.

Which data sources can the agent use?

Only sources whose terms allow your use. Licensed data providers have contracts that say what you may do with their data, so read them before connecting anything.

LinkedIn is the common trap. In section 8.2 of LinkedIn's User Agreement (effective 3 November 2025, current as of November 2026), members agree not to use software, scripts, robots, crawlers or browser plugins to scrape or copy the service, including profiles. The same section bans bots that add or download contacts or send messages, and bans copying LinkedIn information, including through data brokers, without the content owner's consent, unless LinkedIn permits it in a separate writing. So keep the agent off LinkedIn: a person can read a profile, but the agent should not scrape, connect or message there.

What does the GDPR require when you prospect people?

Under Article 3(2) of the GDPR, the regulation can apply to a business outside the EU that offers goods or services to people in the EU. Personal data means information about an identified or identifiable natural person (Article 4(1)), so a named contact at a company counts.

  • A lawful basis. Article 6(1)(f) allows processing necessary for your legitimate interests, unless the person's interests or rights override them. Recital 47 says direct marketing may be regarded as a legitimate interest, but that this needs careful assessment, including whether the person can reasonably expect it.
  • Only the data you need. Article 5(1)(c) limits data to what is necessary for the purpose, so do not enrich every field a provider offers.
  • Telling people. When the data did not come from the person, Article 14 requires you to tell them who you are, why you process their data, your legal basis, their rights and which source the data came from, at the latest in your first communication if you contact them.
  • The right to object. Under Article 21, a person can object to direct marketing at any time, including related profiling, and their data must then no longer be used for it. Your first communication must point out this right clearly and separately.

Fines for breaking these articles can reach EUR 20 million or 4% of worldwide annual turnover, whichever is higher, under Article 83(5) of the GDPR text on EUR-Lex (as of November 2026).

Is business-to-business email treated differently in the EU?

Often, but it depends on the country. Article 13 of the ePrivacy Directive says marketing email needs the recipient's prior consent, with an exception for marketing your own similar products to existing customers. That consent rule applies to subscribers who are natural persons; for companies, each member state must make sure their legitimate interests are sufficiently protected. So the rules for emailing a company address depend on national law. Every marketing email must also identify the sender and give a valid address for opt-out requests.

What does CAN-SPAM require for commercial email?

The FTC's CAN-SPAM compliance guide says the law covers all commercial messages and makes no exception for business-to-business email. As of November 2026, the guide requires you to:

  • Avoid false or misleading headers and deceptive subject lines.
  • Disclose clearly that the message is an ad, and include your valid physical postal address.
  • Explain how to opt out, and keep the opt-out working for at least 30 days after sending, per the FTC guide.
  • Honour opt-outs within 10 business days and never sell or transfer the address of someone who opted out.
  • Monitor anyone who sends on your behalf: hiring another company does not move your legal responsibility.

The same guide says each separate email that breaks the law can bring penalties of up to $53,088, as of November 2026.

Do Gmail's sender rules matter for outreach?

They apply to mail sent to personal Gmail accounts. As of November 2026, Google's email sender guidelines require every sender to set up SPF or DKIM, use TLS, have valid forward and reverse DNS and keep the spam rate reported in Postmaster Tools below 0.3%. Senders of more than 5,000 messages a day also need SPF, DKIM and DMARC, plus one-click unsubscribe on marketing messages. Our guide to SPF, DKIM, DMARC and the Gmail sender rules explains the setup.

How should the agent handle prospect data safely?

  • Store the source and collection date of every field, so the first message can name the source.
  • Keep the suppression list separate from leads, so deleting a lead never deletes an opt-out.
  • Set a retention period and delete prospects that never qualify; Article 14 also requires you to tell people how long you keep their data.
  • Give the agent its own CRM user with minimal permissions (see our guide to guardrails for AI agents that take actions), log every change, and send the AI model only the fields each step needs. Our guide to keeping customer data safe with AI APIs covers provider settings and masking.

If you want an agent like this connected to your CRM, our AI agents and integrations team agrees the sources, scoring rules and review steps with you before anything is built.

Frequently asked questions

Can we buy a lead list and load it into the agent?

Only if the provider's contract allows your use. Ask how the data was collected, which lawful basis the provider relies on and when records were last checked. Load the list into a staging area, run it against your suppression list and your ideal customer criteria, and keep the provider's name as the source on every record.

Which CRM works best with a prospecting agent?

Any CRM with a solid API and custom fields can work, including HubSpot and GoHighLevel. The setup matters more than the brand: fields for source, score and score reason, a suppression property that blocks outreach, owner rules, and a queue where drafts wait for approval. If your CRM cannot hold those, fix that first.

What does the agent need from us before it can start?

A written ideal customer profile with scoring criteria, a list of approved data sources, a CRM user with limited permissions, your privacy notice and opt-out wording, your current suppression list and a few examples of good first messages. Without written criteria, the agent cannot explain its scores.

How personal should AI-written first messages be?

Personal enough to show real research on the company, such as a product it launched or a market it serves, but based on business facts, not the person's private life. Every claim in the draft should link to the page the agent read, so the reviewer can check it before sending.

What should happen when a prospect replies?

The reply goes to the salesperson who sent the message. The agent can label replies as interested, not now, wrong person or unsubscribe and route each one. Any unsubscribe or objection should update the suppression list automatically and stop all further outreach to that person.

Last updated

Start a project

Have a project in mind? Let’s build it.

Tell us what you need. You get a clear plan, a fixed quote and a timeline before any work starts.

Prefer email? info@softitland.com