AI for lead generation, follow-ups and email

AI earns its place in lead generation at four points: sourcing and verifying contacts, scoring inbound leads by behaviour, running follow-up sequences nobody has time to track by hand, and deciding which message goes out when. It does not write the message that closes the deal, and the pipeline still needs a person in it.

The first thing we ask to see on a sales engagement is the call list, in the order the rep will actually work it. It is almost always sorted by when the lead arrived.

Nothing else in the system matters as much as that ordering, and almost nobody is looking at it.

How much of a sales rep's week actually goes to selling?

About a quarter of it. Salesforce's fifth State of Sales report, published in December 2022 from 7,775 sales professionals across 38 countries, found reps spend 28% of the week selling. The other 72% goes to deal admin, data entry, internal meetings and research.

That number is three years old and worth treating as a shape rather than a measurement. The shape is the useful part: the majority of a sales week is spent on work that repeats, follows a predictable logic, and produces a checkable output. That description is also the definition of what automation handles well, which is why this is the function where AI has the least to prove.

Where does AI genuinely help in lead generation?

In sourcing and in ordering, which are different jobs.

Sourcing means pulling contact data, verifying it, and matching it against your ideal customer profile before a rep sees any of it. The rep receives a qualified list rather than a raw one. Calls to people who were never going to buy fall away, and the ones that remain are not better calls, there are simply fewer wasted ones.

Ordering means ranking inbound leads by what they actually did, pages visited, time on site, emails opened, forms submitted, rather than by arrival time or by whoever is at the top of the inbox. The rep calls the lead who spent twenty minutes on the pricing page on Thursday before the one who downloaded something nine months ago and never returned.

Same number of calls, different order, measurably different week. This is the cheapest improvement available to most sales teams and the one that gets skipped, because reordering a queue looks like less work than buying a tool.

What does lead nurturing automation actually do?

It solves an arithmetic problem. A rep holding forty active leads at different stages and temperatures, each needing a different message at a different interval, will drop some. That is capacity, not carelessness, and no amount of discipline fixes it.

A nurture sequence follows every lead on schedule without anyone tracking it by hand. The warm lead who went quiet three weeks ago gets a check-in. The cold one from January joins a long-term drip. The inbound enquiry gets a response in minutes.

What separates a sequence that feels like attentive follow-through from one that feels like being pestered by a company that has forgotten who you are is the map underneath it: where a lead sits at each stage, and what should happen next. That map is a workflow document, not a feature of the tool you bought.

What does AI change about email marketing?

Three things, and segmentation is the one that compounds.

Segmenting by behaviour rather than by industry and job title means the right message reaches the right person, which finally makes your open rates a measurement of the content rather than of the list. Subject-line and send-time testing removes two large guesswork variables. Triggered sequences respond to what a contact does, so the lead who clicks a case study gets more of that, and the lead who has opened nothing in sixty days gets a re-engagement attempt rather than a seventh campaign.

None of it improves the writing. An AI-assisted email that nobody wanted is an email nobody wanted, delivered at a better hour. The question of how the message itself gets written is separate, and it is the one that decides whether any of this works.

What does sending automated email at volume actually cost you?

Your deliverability, if you get it wrong, and the thresholds are published rather than mysterious.

Since 1 February 2024, Google's sender guidelines have required anyone sending more than 5,000 messages a day to a Gmail address to authenticate with SPF, DKIM and DMARC, to support one-click unsubscribe, and to keep the spam rate reported in Postmaster Tools below 0.30%. Google's own recommendation is to stay under 0.10% and never to touch 0.30% at all.

That is the number that should govern how aggressive a sequence gets. Automation makes it trivial to send more, and the complaint rate is the thing that quietly decides whether the next campaign reaches the inbox at all. Volume is the easy lever and it is the one with a published ceiling.

What has to be connected for any of this to work?

The CRM, to everything else. Sourcing that does not write back to it produces data nobody uses. A nurture sequence that cannot see deal stage fires the wrong message at someone already talking to a rep. A campaign tool blind to sales activity emails people mid-negotiation.

The integration layer is less interesting than the sequences and it is what decides whether this is a system or four tools in the same building. It is also the part that breaks quietly, which is the argument for knowing what the whole thing does on its worst day before it is running your pipeline.

Start by sorting the call list by behaviour instead of by date. Everything above is an elaboration of that.

MorningLabs AI builds the agents, automations and internal systems a business runs on, and audits AI systems that are already live.