AI lead generation, from one sentence to the first replies.

What AI lead generation is, and the chatbot it is not

Search for AI lead generation and about half of what comes back is inbound: a chatbot that greets visitors on your site, a form that scores whoever fills it in, a tool that routes the lead to a rep. Those help with people who already found you. This guide covers the other half, outbound, where you choose the companies and the people and write first.

In outbound the slow part is everything before the first message. Someone turns an idea of the buyer into a list, finds a work email for each person, checks who really fits and looks for something worth saying. Later someone keeps up with the replies on two channels. AI now does most of that legwork well, and the parts below follow it in the order a run takes.

The order matters more than the tool. A list nobody checked wastes a good opener, and a good list gets nothing from a message that reads like a template. When AI outbound goes wrong, it is usually a skipped step: no fit check, or replies left unread for two days.

GetSales works the outbound side. Say who you sell to, typed or said, and Creator plans the search, the table, the work emails, the first lines, the workflow and the AI SDR.

Find people from a description, not a grid of filters

The usual way to build a list is a search form: tick the titles, an industry, a headcount band and a country, then scroll. It works, but the form only holds what it offers, and the buyer in your head, heads of sales at UK software companies big enough to have a team and too small for a sales ops hire, gets cut down to four dropdowns.

AI search starts from the sentence instead. You describe the buyer the way you would brief a new colleague, the tool turns it into the filters it can run, and it shows you those filters. Read them before the first page comes back. A word the tool could not match should be named, not quietly guessed, because a guessed filter fills the list with people you never asked for.

Write the sentence from the deals you won, not the market you hope for. Leave out who to exclude: customers, open deals and competitors belong on a do-not-contact list that every search checks, not in the description. A good description names three things.

• The role, in the words people put in their titles: Head of Sales, VP Sales
• The company: its industry, its size and its country
• One sign they could buy now, such as a sales team that is hiring, kept for the fit check

In GetSales, AI Search turns your sentence into a search filter and shows you what it built. A word it cannot match is named, not guessed, and nobody already in your tables comes in twice.

Score the fit and find work emails before anyone is written to

A search returns everyone who matches the filters, and filters are blunt. A title and a headcount cannot tell you whether a company sells to businesses, whether it just hired its first SDRs or whether it already runs a tool like yours. That is the job of a fit check: an AI column that reads each row and answers one question in your words, with the reason for its answer.

Write the question as a rule a colleague could apply, then try it on ten rows and read every reason. A wrong reason means a wrong prompt, so fix the words before the column runs on the whole list. Ask for a short scale, such as strong, good or weak, rather than a score out of 100 that looks precise and is not.

Then pay for data only where it will be used. A waterfall asks one email finder after another until one returns an address, and a verifier checks that the address takes mail before anyone sends to it. Run both on the people who pass the fit check, and write the first line last, from one fact on the row, for the same people. The order, in four steps:

1. A fit check on every row, with its reason
2. A work email for each good fit, asked of one finder after another
3. A verifier on each address before it is used
4. A first line for each fit with a checked email, from one fact on their row

In GetSales a Use AI column scores each row and keeps its reasoning with how sure it is. Find work email asks finder after finder in the order you set, and a miss costs nothing.

Let a trigger choose the moment

The same message reads differently in a quiet week and in the week someone starts a new job. A trigger is an event that makes a message timely: a new role, a new post, a return visit to your pricing page after an earlier email, an accepted connection request. AI cannot make up a reason to write, but it can watch for these events and either tell you or take the step you set.

Choose two or three triggers that fit how you sell and decide in advance what each one does. A job change might move the person into a short sequence for new leaders. A visit to your pricing page from someone you wrote to might pause their follow‑ups so a person can write by hand. Use the trigger to pick the day and keep it out of the message: nobody wants to read that you saw them on your pricing page.

Be careful with bought intent data. Topic intent from a third party can help you choose accounts, but it is a guess about a company, not something a person did. First-party signals, from your own links, site and messages, are stronger, and news about people you already hold, such as a new job, is easy to check. Four triggers worth watching:

• A new job, when people often look again at the tools they inherited
• A new post on the problem you solve
• A return visit to your site after you wrote
• An accepted connection request, which opens the first real message

In GetSales, Signals shows website visits, posts and job changes from the people you write to, and a play takes the next step or asks you first.

Reach them on LinkedIn and email from your own accounts

A buyer who ignores a connection request may still answer an email, and the other way round, so a sequence on both channels gives each person two ways to reply. A common shape: visit the profile, send a connection request with a short note or none, then a message to the people who accept and an email to the rest a few days later. Plan one or two follow‑ups on each path, each with a new reason to answer.

Send from the accounts of the person whose name is on the message, their own LinkedIn account and mailboxes, and keep each account within its limits. In the US a cold B2B email is commercial email under CAN-SPAM: a true sender and subject line, your postal address and an opt-out honored within 10 business days, in the FTC's compliance guide. Google's sender guidelines, in force since February 2024, ask every sender for SPF or DKIM and a spam rate under 0.3%.

Here AI helps most with the part that changes per person. It can write the first line from a fact on the row and keep it to one sentence. It cannot tell whether your offer is worth the reader's time, so write the body yourself, keep it short and let the AI fill only the line that changes. Three messages to start from:

In GetSales a workflow puts LinkedIn and email steps in one sequence, sent from your sender profiles and paced within each account's limits. By default a reply pauses the follow‑ups.

Answer every reply, with you in the loop

A reply is what the run was for, and it is where AI tools differ the most. Some answer on their own from the first day, others draft and wait. Start with drafts that wait for your approval, read a few dozen, and let the AI send on its own only where its drafts have stopped needing your edits.

Some replies always belong to a person: a price question, a request for a call or a time, anything about legal or security, and any reply the AI is unsure how to read. Answer the same day if you can. When they say yes, offer one booking link, and if they name a time instead, confirm it yourself and send the invite.

Sort each reply before you answer it, since each kind wants its own next step:

• Interested or asking for a meeting: answer what they asked, then offer one next step
• Wants more: send the short version of what they asked for
• Not now: agree, and write again on the date they named
• Wrong person: ask who owns it, then write to that person
• An objection: answer with one proof, not three
• Do not contact: on your stoplist the same day, for good

In GetSales the AI SDR labels each reply with how sure it is and why, and drafts the answer. Each scenario sets whether it sends on its own or waits for your OK, and price, call and legal questions come to you.

Where AI helps, and where a person decides

AI lead generation moves the legwork to software: the search, the lookups, the checks and the first drafts. The judgment stays with people, and it should, because the costly mistakes in outbound are judgment calls: the wrong market, a careless offer, a good reply answered badly.

Keep two decisions with a person on every run: the go-ahead before the first send, and the answer to any reply that matters. Everything before the go-ahead can be checked on a sample of rows. Everything after a reply is a conversation with a real buyer.

That is also the honest answer to whether AI takes over the SDR's job. It takes the tasks that are the same for every lead. Knowing the market, picking the accounts and running the call are still the job.

In GetSales a Creator plan makes nothing and spends nothing until you run it, and going live is a second yes. Nothing is sent before it.