How to automate lead generation with AI

Winning new clients is the engine of any company, but it is also one of the most time-consuming tasks: finding contacts, writing emails one by one, following up, noting who replied… Manual, repetitive work that is easy to abandon when busy weeks arrive. Artificial intelligence lets you automate much of this process and keep a steady stream of opportunities. Here is how.

Automating lead generation has a bad reputation, and for good reason: most of what is done under that name is sending the same email to a thousand people and calling it personalisation. This article is about the other thing: which parts of the sales process automate well, which do not, and where the line is.

What automates and what does not

Lead generation has four phases and only two are automatable.

Identifying who to target: almost entirely automatable.
Making contact for the first time: automatable with care.
Conversing when they reply: for people.
Closing: for people.

Most disasters come from trying to automate the third. A system that holds the conversation for you works until the other person asks something off-script, and then it becomes obvious.

Identifying: building the list

It is the most profitable part to automate and the lowest risk, because it does not touch the customer yet.

It means defining precisely who you want to reach (sector, size, area, some specific signal) and building a system that finds those companies and gathers the public information needed to approach them meaningfully.

A real example: for ticwebapp we built a system that identifies businesses with no website or an outdated one. The signal is objective and verifiable, and it defines a niche where their offer fits obviously.

The key is the signal. “Barcelona companies with 10 to 50 employees” is not a signal, it is a demographic filter. “Companies that have just opened a second site” is, because it indicates a specific moment when they have a specific problem.

Making contact: where almost everyone gets it wrong

Here is the difference between a system that works and templated spam.

Personalising is not putting the name in the subject line. Writing “Hi Marta” and continuing with the same text five hundred others receive fools nobody. Real personalisation means the reason you are writing is specific to that company: something you have seen, something they do, something happening to them.

If the email you send could go to any other company just by changing the name, it is not personalised.

Volume does not fix a bad message, it amplifies it. With a weak message, sending ten times more multiplies unsubscribes and complaints tenfold, and burns your email domain. It is the mistake that turns a legitimate tool into a deliverability problem taking months to fix.

Send fewer and better. Fifty well-targeted emails a day yield more than five hundred generic ones, and they do not burn your domain.

The legal side, which is not optional

For B2B commercial communications in Spain you must consider the GDPR and the Spanish e-commerce law. The points that cannot be ignored:

• Identify yourself clearly. Who you are and on whose behalf you are writing.
• A simple, working unsubscribe route. In every send, and one that actually works.
• Legal basis for processing. If you are going to process professional contact data, you must know on what basis and be able to justify it.
• Handle unsubscribes and rights requests. And do it quickly.

None of this prevents prospecting. What it prevents is doing it crudely. If you have doubts about your case, check with whoever handles your data protection: it comes out far cheaper than a complaint.

Prioritising: what pays off most and is done least

If you are only going to automate one thing in the sales process, make it this.

A system that scores each contact by what they have done (which pages they viewed, what they opened, what they replied) and by what they are (sector, size, fit with your ideal client) and sorts the list by real likelihood of closing.

It does not change what the team does, only the order in which they do it. And order is exactly what has the most impact when there are more contacts than hours available.

Follow-up: the money left on the table

Most opportunities are not lost on price or product: they are lost because nobody followed up.

Automating follow-up is among the most profitable and least risky things: internal reminders of who to write to next, alerts on proposals sent without a reply, notifications when a contact returns to the site.

Note the nuance: these are reminders for you, not automated emails to the customer. A person does the following up; what gets automated is remembering.

How to know whether it works

Forget the number of emails sent: it means nothing. Look at these:

Reply rate, not open rate. Opens are increasingly unreliable because of how image filters work.

Meetings booked per hundred contacts. It is the only number that turns into money.

Unsubscribes and complaints. If they rise, the message or the list is wrong, and it is best to stop before it affects the deliverability of all your email.

Where to start

First, prioritise what you already have. Almost every company has unworked contacts in the CRM, and sorting that list is faster and cheaper than going out to find more.

Then follow-up. And only when those two work, cold outreach.

Doing it the other way round (starting with mass sending) is what produces the cases that gave all this a bad name.

If you want to see which other processes pay off, here are the ones that work best. And to work out how much time manual prospecting is costing you today, try the savings calculator.