← Back to blog

AI Agents

AI Agents for Lead Generation: What to Automate First

Glass funnel filtering glowing prospects down to a single lead that lands on a booked-call calendar card

AI agents for lead generation take over the repetitive pipeline work: finding prospects, researching them, qualifying them against your criteria, drafting outreach and following up. Your team keeps the conversations and the decisions. Build in this order: research, qualification, drafted outreach, follow-ups, and only then automatic sending.

This is a build guide, not an overview. You will write your inputs, fill in an ICP and signals worksheet, build a list and pilot it on 10 to 20 leads, score leads with a rubric, write first lines and sequences, set reply and handover rules, and go live with a small batch. It needs one short AI instruction (the first-line prompt in Step 5) and everything else is a worksheet, rubric, template or rule table. If you are new to the term, read what an AI agent is first, then come back.

The whole build in 20 seconds. Define your ICP and signals, pilot on 10 to 20 leads, qualify with a rubric, write the messages, set reply rules and go live with a small batch.

Disclosure: some tool links in this guide are referral links from our AI Stack page. If you sign up through them we may earn a commission at no extra cost to you. We list tools we use or have vetted, and you can run every step here with other tools.

Who this is for: B2B founders, consultants and agency owners who sell a considered service (typically a deal worth thousands, not a $20 subscription). If your average deal is small, the economics of researched, personalised outreach rarely work, and a self-serve funnel is a better bet.

What you will have at the end

  • A one-page ICP and signals worksheet that a stranger could use to say yes or no to any company.
  • A pilot table of 10 to 20 scored leads, with a source link behind every signal.
  • A qualification rubric with a decision rule, checked against your own judgement.
  • A first-line instruction, a four-step email sequence and a LinkedIn sequence.
  • Reply rules that stop unsubscribes automatically and give every interested reply a named owner.
  • A KPI table, a pre-launch checklist and a go-live plan for your first week.

How AI agents for lead generation work

A lead generation agent is a chain of small jobs. Each job has one input, one output and one check. Anthropic's engineering team describes this as a workflow built from patterns such as prompt chaining (each step feeds the next, with checks in between) and routing, as opposed to open-ended agents that decide their own steps. For lead generation the chained, checked approach is right, because you want predictable output you can review. (Source: Anthropic, "Building effective agents", accessed 6 November 2026.)

In order, the chain is: find companies that match your filters, research each one, qualify it against your rubric, write a personalised first line and fill your sequence, then label replies and route them to a person. Every stage has a human check, and the steps below build them in that order.

Before you start: what you need

  • A document or spreadsheet for the worksheets and the pilot table.
  • A source for your first list: a sales navigator export, a database you pay for, a client list or event attendees.
  • An AI model you can give instructions to, ideally with your own API key.
  • A secondary sending domain with mailboxes, if you will send email (see the deliverability section).
  • One named person who owns replies and handovers.

Tools that can run these steps

Descriptions follow our AI Stack page. Check each vendor's documentation for current features, limits and pricing before you commit. You can do the planning and pilot steps with a spreadsheet and an AI chat window, so the post works without any of these tools.

StepToolHow it fits
Build the list and find signalsDeeplineGTM data infrastructure built for AI agents: waterfall enrichment, lead and account discovery, and signal detection
Build the list and find signalsTheirStackTech stack and job posting data for precise, intent-based account targeting
Find and verify emailsFindymailFinds and verifies decision-maker emails so your lists convert instead of bouncing
Verify emails again before sendingBouncify or ZerobounceEmail verification that keeps bounce rates low; we run a second layer of validation before any list goes live
Sending domains and inboxesZapmailDone-for-you inbox infrastructure so your sending domains stay warm and out of spam
Email sendingInstantly or LemlistInstantly is cold email sending infrastructure with built-in deliverability; Lemlist combines email, LinkedIn and calling in one sequence builder
LinkedIn outreachHeyReach or GetSalesHeyReach is multi-account LinkedIn outreach built for agencies; GetSales handles connection requests, messaging and CRM sync in one place
Workflow gluen8n or MakeThe workflow layer that moves rows between the list, the AI step, your sending tool and your CRM
CRMPipedrive or BreakcoldPipedrive is a simple, visual CRM pipeline; Breakcold is a social-selling CRM built around LinkedIn and Twitter engagement
BookingCal.comOur preferred booking tool, for the link in your reply drafts

The build, step by step

Step 1: Write your four inputs

Input: what you already know about your best clients. Do: fill in the worksheet below in a document before you touch any tool. Agents amplify whatever you give them, so fix the inputs first. Output: one page. Check: a colleague could apply your "qualified" definition in 30 seconds.

Asset 1: Inputs worksheet
InputWhat to writeYour answer
ICP in plain languageIndustry, company size range, country or region, buyer job titles, and the two or three signs that someone needs what you sell
Definition of "qualified"A checklist a human could apply in 30 seconds (Step 4 gives you a rubric to copy)
Offer in one paragraphThe problem, the outcome, who it is for, and what a first conversation covers
Exclusion listExisting clients, competitors, partners and anyone who has asked not to be contacted. Check it before anything is drafted

Step 2: Fill in the ICP and signals worksheet

Input: the offer and ICP from Step 1. Do: write up to three segments, one row each. A signal is a publicly verifiable sign the company needs your offer now, such as hiring for a relevant role, a leadership change, a new location, a product launch or a funding round. Output: a completed worksheet plus 10 exclusion rules (types of company you never contact). Check: a colleague who has never heard of your business can read it and say yes or no to any company you name. If they cannot, tighten the signals and the qualifying question.

Asset 2: ICP and signals worksheet
SegmentJob titles (primary, secondary)Industries and sizeSignals to look forCore problem, in their wordsQualifying question
1 (example, fictional): Warehousing firmsFounder or head of operations; HR managerLogistics and warehousing, 50 to 200 staff, GCCHiring several operational roles at once; new warehouse; new operations director"We are growing faster than our hiring process"Who runs hiring when you open a second site?
2 (yours)
3 (yours)

Step 3: Build the list and pilot on 10 to 20 leads

Input: the worksheet and your list source. Do: resist the urge to start big. In our setup the rule is to build the whole table end to end on a sample of 10 to 20 leads first, then estimate the cost of running it on the next batch of 100 to 300, and only then scale to the next few hundred. Enrichment and AI steps cost money per row, so a flaw multiplied across thousands of rows is expensive. Output: a pilot table with the columns below, in this order. Check: every signal has a source link and a date, and you have read every row.

Asset 3: Pilot table columns
#Column groupColumns
1ContactCompany, website, contact name, job title, profile link
2Clean nameNormalised company name (drop "LLC", "Ltd", "Inc" and legal suffixes so messages read naturally)
3ResearchCompany summary (two sentences from the website)
4SignalOne recent, verifiable fact, with the source link and date
5QualificationQualified (yes, no or maybe), Score, Reason
6OutreachFirst line, Email draft, Status

Where your tooling lets you, use your own API key for the AI model rather than a platform's bundled credits, and check usage after the pilot. Re-run steps only on rows that failed or came back empty, never on the whole table.

Step 4: Qualify with a rubric that demands evidence

Input: the pilot rows with summary and signal. Do: score each lead out of 10 using criteria you can verify from public information. The most common failure in AI lead generation is a vague "is this a good fit?" check, so the rubric forces evidence. Apply it yourself on the pilot, or give the same table to your AI step and keep these rules: use only the information in the row, write "unknown" and score 0 when a fact is missing, and never guess. Output: a score, a verdict and a one-sentence reason with evidence for every row. Check: pick five rows you know well. The verdict should match your own judgement on at least four. If it does not, the rubric or the input text is the problem, not the model.

Asset 4: Qualification rubric
CriterionPointsEvidence that must be cited
Industry matches an ICP segment0 to 2Wording from the company's own site
Size within range0 to 2Team page or profile headcount
Region matches0 to 1Contact or office page
Contact is a decision maker0 to 2Job title
At least one buying signal0 to 3Source link and date
TotalVerdictWhat happens
Matches an exclusion ruleExcludedDropped before scoring, reason on file
8 to 10QualifiedGoes to first lines
5 to 7MaybeA human reviews it
0 to 4Not qualifiedSkipped, reason on file

Step 5: Write personalised first lines that are not generic

Input: the signal and summary for each qualified row. Do: personalisation that only inserts a name and a company is easy to spot. A good first line shows you read something specific and says nothing about you. This is the only AI prompt you need in this build: run it on qualified rows only, and read every line from the pilot yourself. Output: one first line per qualified row. Check: every line passes the rule: specific fact, no flattery, no pitch.

Asset 5: First-line prompt (the only prompt in this guide)
Write ONE opening line for a cold message.

Rules:
- Under 18 words.
- Start from a specific, verifiable fact about the company: <SIGNAL>
- Positive and factual in tone. No flattery words such as "impressive", "amazing" or "love".
- Do not mention our company, our offer or any pitch.
- Do not use exclamation marks or em dashes.

Company: <COMPANY_NORMALISED>
Contact first name: <FIRST_NAME>
Summary: <SUMMARY>

Output only the line.

A weak line: "Love what you are doing at Harbor Freight Co!" A strong line: "Saw Harbor Freight Co opened a second warehouse in Dammam this quarter." Another strong one: "Noticed you are hiring three operations coordinators across two sites." The company is fictional and used only to show the pattern.

Step 6: Build the outreach sequences

Input: qualified rows with first lines. Do: use short messages. For email, keep subject lines under 7 words and avoid images and tracking-heavy formatting. For LinkedIn, keep the connection note to one sentence under 300 characters and DMs under 80 words. The cadence follows the one we run in our own LinkedIn sequences: a connection request, then follow-ups roughly 2 to 3 days, 4 to 5 days and 7 to 10 days after the previous step. Fill the templates below (these are templates, not prompts). Lock the structure: only the first line and the specific insight may change per lead, never the ask or the opt-out language. Output: drafts loaded in your sending tool, not yet sent. Check: read five full drafts as the recipient would, and confirm every email has an opt-out line.

Asset 6: Four-step email sequence
EMAIL 1 (day 1)
Subject: <COMPANY> and <PROBLEM_AREA>

<FIRST_LINE>

We help <ICP_TYPE> with <PROBLEM>. Teams usually come to us when <TRIGGER>.

Worth a short conversation, or is someone else the right person?

<SENDER_NAME>

EMAIL 2 (day 4) - proof or insight
Subject: Re: <COMPANY> and <PROBLEM_AREA>

One pattern we see with <ICP_TYPE>: <SPECIFIC_INSIGHT_IN_ONE_SENTENCE>.
If useful, I can send the two-page checklist we use for it.

EMAIL 3 (day 9) - what is working
Subject: <PROBLEM_AREA> checklist

<ONE_PRACTICAL_TIP_THEY_CAN_USE_WITHOUT_YOU>
Happy to share the full version. Want it?

EMAIL 4 (day 16) - close the loop
Subject: Should I close this out?

I will assume this is not a priority right now. If <TRIGGER> comes up later, reply here and I will pick it up. If you would rather not hear from me, tell me and I will remove you.
Asset 7: LinkedIn sequence
CONNECTION NOTE (under 300 characters)
Hi <FIRST_NAME>, I work with <ICP_TYPE> on <PROBLEM_AREA> and saw <SIGNAL>. Thought it would be useful to connect.

FOLLOW-UP 1 (2 to 3 days after they accept)
Thanks for connecting. Are you currently doing anything about <PROBLEM_AREA>? A lot of <ICP_TYPE> we speak to find it harder than expected.

FOLLOW-UP 2 (4 to 5 days later)
<FIRST_NAME>, most <ICP_TYPE> do not need more of <THING>, they need a better <PROCESS>. We built a simple way to do that. Want a quick walkthrough?

FOLLOW-UP 3 (7 to 10 days later)
No problem if now is not the right time. I will leave it here. If <TRIGGER> becomes a priority, message me.

Step 7: Set reply rules and the handover

Input: the live inbox and LinkedIn replies. Do: replies are where agents earn their keep and where they do the most damage if left unchecked. Label every reply with one of the labels below, using the first rule that matches, then decide who acts. A person can apply this table by hand in week one; check your tool's documentation before automating it. Output: every reply has a label, an owner and a next step. Check: an unsubscribe stops all sequences immediately, and an "interested" reply gets a human response inside your target time. The agent never answers pricing or scope questions on its own.

Asset 8: Reply rules and handover table (first match wins)
RuleIf the reply...LabelWhat happensWho acts
1Asks to stop contact in any way, whatever else it saysunsubscribeStop all sequences, add to the suppression listAutomatic, immediate
2Is an automatic away messageout_of_officePause, resume after the return dateAutomatic
3Shows interest or asks to talkinterestedDraft reply with booking link preparedA person sends
4Asks a question about the offerquestionDraft an answer from your offer paragraph, flag pricing and scopeA person sends
5Says not nownot_nowAdd to a nurture list with a follow-up dateAutomatic, human reviews weekly
6Points to someone elsewrong_person or referralExtract the new contact and restart the sequence for themA person approves
7Says it is not a fitnot_a_fitHold, note the reasonA person decides
8Matches none of the above, or you are unsureunclearHoldA person decides

Step 8: Set up tracking and go live

Input: the pilot results, the drafts and the reply rules. Do: authenticate your sending domain (see the next section), connect the suppression list, work through the pre-launch checklist below, then send a small first batch (10 to 20) after you approve it. Log each reply label and outcome in the KPI table and decide the next batch size from the cost you saw in the pilot. Output: a live, small, monitored campaign. Check: the spam complaint rate in your mailbox tools stays low and every interested reply was answered inside your target time.

A worked example (illustrative)

This example is illustrative. The company and all names are fictional, and the numbers are invented to show how the steps connect. They are not a client result.

The seller: a four-person HR advisory firm that helps growing companies set up hiring processes. The ICP (Step 2): logistics and warehousing firms with 50 to 200 staff in the GCC, buyer is the founder or head of operations, signal is hiring for several operational roles at once.

Company (fictional)Signal foundScoreVerdictFirst line produced
Harbor Freight CoOpened a second warehouse, hiring 3 coordinators9qualifiedSaw Harbor Freight Co opened a second warehouse and is hiring three coordinators.
Palm Courier LtdNone found; right size and region6maybe(held for human review)
Desert Cold StorageNew operations director appointed8qualifiedNoticed Desert Cold Storage recently appointed a new operations director.
Atlas Print WorksManufacturing, not logistics3not_qualified(skipped)

Of four leads, two go to drafted outreach, one is held for the founder to judge, and one is dropped with a reason on file. When Harbor Freight Co replies "can you send details?", rule 4 of the reply table labels it question, a draft with a booking link is prepared and the founder sends it.

Deliverability and compliance: the non-negotiables

Strong messages do not matter if they land in spam or your domain gets blocked. Google publishes sender requirements for mail to Gmail accounts. As of 6 November 2026 its guidance says:

  • All senders should set up SPF or DKIM authentication for their sending domains, use a TLS connection, and keep spam rates reported in Postmaster Tools below 0.30 percent.
  • Senders who send more than 5,000 messages a day to Gmail accounts must set up SPF, DKIM and DMARC (the DMARC policy can be set to none), keep the sending domain aligned with SPF or DKIM, and support one-click unsubscribe for marketing and subscribed messages.
  • Google advises keeping spam rates below 0.10 percent and never reaching 0.30 percent or higher.

(Source: Google, "Email sender guidelines", accessed 6 November 2026.)

Practical rules we apply regardless of volume: send from a secondary domain, never your main one; warm up new mailboxes gradually; include a plain opt-out line in every email; stop all sequences the moment someone replies or unsubscribes; and run a spam-check on each email draft before launch. Rules on unsolicited messages differ by country and channel. We are not giving legal advice, so check the requirements for each market you contact.

Troubleshooting

ProblemLikely causeFix
More than 30 percent of leads come back "maybe"Rubric criteria are too vagueAdd evidence requirements, tighten the signal definition
First lines sound alikeThe signal field is genericRequire a dated source link for every signal
Replies are mostly "wrong person"Title filter is too broadNarrow job titles, add a secondary contact step
Opens fall off a cliff after a few daysPossible deliverability problemPause, check authentication and spam rate, reduce volume
The AI step invents factsInstructions allow guessingAdd "use only the information in the row; write unknown otherwise" and score unknown as 0
Costs creep upRe-running whole tablesRe-run only failed rows, pilot before scaling
Verdicts disagree with your own judgementRubric or input text is weakFix the rubric criteria or the research summary, then re-run only the rows that disagree

What to measure and how to read it

Asset 9: KPI table
MetricHow to calculateWhat it tells youIf it is weak
Qualification pass ratequalified / leads processedQuality of your list source and ICPFix sourcing or ICP before copy
Reply ratereplies / delivered messagesTargeting and first lineTest new signals and first lines
Positive reply rateinterested + question / repliesOffer and message fitRework the offer line and the ask
Booked callscalls on calendar / qualified leads contactedWhole-pipeline outputCheck handover speed first
Show ratecalls held / calls bookedLead intent and remindersAdd reminders, confirm by message
Spam complaint rateas reported in your mailbox provider toolsSafety of your domainPause and fix before anything else

Review these weekly on a sample of at least a few hundred sends before drawing conclusions. Change one thing at a time (the first line, the subject, the ask) so you know what moved the result.

Pre-launch QA checklist

  • The pilot ran end to end on 10 to 20 leads and I read every output.
  • Five spot-checked verdicts matched my own judgement on at least four.
  • Every signal has a source link and a date.
  • The exclusion list is applied before drafting.
  • Every first line passes the rule: specific fact, no flattery, no pitch.
  • Sending domain is authenticated (SPF, DKIM, DMARC) and is not my main domain.
  • Every email has an opt-out line; the suppression list is connected.
  • The reply rules route unsubscribes to stop sequences automatically.
  • A named person owns every "interested" reply and has a response-time target.
  • Costs for the next batch are estimated from the pilot.

Your first week, day by day

  1. Day 1: Write the four inputs (Step 1) and fill in the ICP and signals worksheet until a colleague can use it (Step 2).
  2. Day 2: Pull 20 leads and build the pilot table columns (Step 3). Score all 20 with the rubric (Step 4).
  3. Day 3: Review verdicts, fix the rubric, re-run only the rows that disagree with you.
  4. Day 4: Generate first lines for qualified rows (Step 5). Edit by hand until you trust the pattern.
  5. Day 5: Load the email and LinkedIn sequences as drafts (Step 6). Set up authentication and the suppression list. Send a first small batch (10 to 20) after approval.
  6. Day 6: Apply the reply rules to the first replies and write the handover owner and response time (Step 7).
  7. Day 7: Review the KPI table, note what to change, and decide the size of next week's batch from the cost you saw (Step 8).

Frequently asked questions

Can AI agents generate leads on their own?

They can handle the repetitive parts of lead generation, such as finding prospects, researching them, scoring them and drafting outreach. Results still depend on a clear ICP, a good offer and a person who reviews the output and owns the system.

What is the first thing to automate in lead generation?

Start with prospect research and qualification. A mistake there is cheap to catch because nothing is sent to a prospect until a human has reviewed the list. Automate drafted outreach next, follow-ups after that, and automatic sending last.

How many leads should I test before scaling?

Run the full table end to end on 10 to 20 leads first. Read every output, estimate the cost of the next batch of 100 to 300 from what the pilot used, and only then scale up.

Is AI lead generation the same as AI outbound?

They overlap. AI outbound usually means using agents for cold email and LinkedIn outreach, while AI lead generation also covers finding, researching and qualifying prospects before any message is sent.

Do I still need a salesperson if I use AI agents for lead generation?

Yes. Agents handle volume work, but conversations, negotiation and judgment calls are best handled by a person. Think of the agent as removing admin from your sales process, not replacing the people who close.

Is automated outreach legal and safe for my domain?

Rules differ by country and channel, and this is not legal advice, so check the requirements for each market you contact. For email, Google's sender guidelines ask senders to authenticate their domains, keep spam rates low and support unsubscribing. Always include an opt-out and stop sequences when someone replies or unsubscribes.

Sources and further reading

  • Anthropic, "Building effective agents" (workflows versus agents, prompt chaining, routing). Accessed 6 November 2026.
  • Google, "Email sender guidelines" (authentication, spam rate and one-click unsubscribe requirements). Accessed 6 November 2026.
  • SFC AI Stack page (tool descriptions; some links are referral links).
  • SFC internal prompt library and enrichment SOP (methods adapted and anonymised).

What to do next

Do Day 1 today: write the four inputs and fill in the ICP and signals worksheet. If you would rather have the system built and running in your own accounts, we build AI outbound agents that source, qualify and reach out to prospects and push qualified leads into your CRM, and you can book a short call to talk through your pipeline. Once leads reply, our guide to the AI appointment setter covers turning conversations into booked calls, and signal-based outbound goes deeper on choosing the signals that trigger outreach.

Want help instead of just reading about it?

Tell us what you're trying to do and we'll show you the path, whether that's AI automation, business setup, or growth.