Maybe AI should help you generate fewer leads

There is a familiar promise appearing in AI lead-generation software at the moment.

Give the system your website. Let it understand what you sell. Ask it to identify hundreds of relevant companies and contacts.

Technically, that is quite impressive. Commercially, I am less convinced it solves the interesting part of the problem.

AI has made finding companies dramatically easier. It can scan markets, monitor company news, interpret websites, identify job changes, track investment and expansion, compare organisations against an ICP and surface possible contacts at a scale that would have required a lot of manual research not very long ago.

That is useful.

But finding a plausible company is only the beginning of B2B prospecting. The harder question is whether there is a sufficiently good reason to contact somebody there.

And the easier AI makes the first problem, the more important the second becomes.

Finding prospects has become cheap

Most traditional lead-generation processes have been constrained by research capacity.

If identifying 100 potentially relevant businesses requires somebody to spend hours searching, reading and checking them, simply producing the list has value. The cost of discovery creates a natural limit on how much of the market can be considered.

AI changes that constraint.

It is now relatively straightforward to monitor large numbers of companies for commercial changes: acquisitions, funding, new leadership, expansion, recruitment, new propositions, market entry and other signals that something may be happening.

You can also ask systems to compare companies against basic qualification criteria. Do they operate in the right market? Are they the right sort of size? Do they appear to sell the right sort of product? Do they resemble existing customers?

The available prospect pool can become very large, very quickly.

That sounds like progress until you remember that sales and marketing teams still have finite attention.

The scarce resource has moved.

Finding another company is increasingly easy. Deciding whether that company deserves human attention is still difficult.

A signal is not a reason to contact somebody

This distinction matters because outbound systems can easily confuse three different things.

A company might fit your ideal customer profile.

It might also exhibit a potentially interesting signal.

Neither necessarily means there is a useful conversation to be had now.

Suppose a company has appointed a new commercial director. That may indicate change. It may even correlate with future marketing activity. But the appointment itself tells you very little about whether there is a problem you can solve, whether the organisation needs outside support or whether contacting that person this week would be useful to either of you.

The same applies to investment announcements, recruitment activity, acquisitions and expansion.

Signals are evidence that deserves interpretation. They are not automatic buying intent. It is the same distinction behind A signal is not a decision: the signal may tell you where to look, but it does not make the decision for you.

This is where some AI outbound propositions feel strangely underdeveloped. They apply increasingly sophisticated technology to generate a larger quantity of possibilities, then treat those possibilities as leads.

The result can be a technically impressive system for creating more things for somebody else to qualify.

The missing layer is rejection

A useful B2B prospecting system needs a layer whose main job is to say no.

In my own business development, I use scheduled searches and AI-assisted monitoring to look for commercial changes across the market. That creates a stream of possible situations.

The useful part happens next.

Those situations are checked against a set of questions designed to determine whether anything commercially interesting might actually be happening.

What has changed?

An acquisition, leadership appointment or recruitment drive may be relevant because it changes the organisation's circumstances. The signal matters because of the commercial context around it, not because it appeared in a news feed.

Is there a plausible marketing problem?

A company can be growing without having a problem I can usefully solve. There needs to be some reasonable connection between the observed situation and a marketing or commercial challenge.

Is there evidence of a capacity gap?

A business might clearly need stronger marketing but already have a capable senior team in place. Equally, rapid expansion or a visible gap in leadership may suggest that additional capability could be useful. This is rarely something one data point can answer.

Is there genuine capability fit?

Could I realistically help with the situation? A prospect is not qualified simply because it has a marketing problem. The problem needs to be one where my experience and operating model are relevant.

What is already known?

The organisation might already be in the CRM. Somebody may have spoken to them before. There may be a previous relationship, an open opportunity or a good reason not to restart a conversation as though none of that history exists.

Why now?

Perhaps the most important question. Is there an actual reason for contact at this point, or have we simply discovered that the company exists?

None of these tests is particularly exotic.

That is partly the point. The value comes from applying ordinary commercial judgement consistently before creating activity. It is also a marketing systems problem: the quality of the outcome depends on how discovery, qualification, CRM history, outreach and learning connect, rather than on the performance of one isolated tool.

Most signals should go nowhere

A system designed this way produces an initially uncomfortable result: most of what it discovers gets rejected.

That can look inefficient if the objective is maximum lead volume.

It looks rather different if the objective is to identify a small number of situations where a relevant conversation might genuinely be useful.

This is one of the more interesting effects I have found from using AI seriously in prospecting. Greater automation has made me more selective, not less.

Because I can scan more of the market, I do not need to treat every vaguely suitable organisation as an opportunity.

Because initial research is cheaper, rejection is cheaper too.

A company can fit the ICP and still go nowhere. A strong signal can turn out to be commercially irrelevant. An apparently interesting situation can disappear once the wider context is checked.

That is healthy.

A low survival rate is not necessarily a problem. If every signal becomes an outreach task, the system is not really qualifying anything.

AI can support judgement without replacing it

There is an obvious temptation to automate this entire layer as well.

Some of it should be automated. AI is useful for gathering evidence, comparing information, spotting inconsistencies, summarising changes and checking whether a situation meets defined criteria.

It can help answer questions such as whether a company has recently recruited for particular roles, whether there is evidence of repositioning, whether a relevant leadership change has occurred or whether the organisation has appeared in an existing pipeline.

The harder question is how much of the resulting commercial judgement you should delegate to it.

Qualification often depends on ambiguity.

A company may appear to have a capacity gap but actually be halfway through recruiting. A strategic change might create an opportunity, or it might make external support less relevant. A marketing problem may be visible, but solving it might require capabilities outside your own.

There is also a subtler question: would contact be useful?

That involves context, timing and a degree of empathy. It means considering whether you have something relevant enough to justify interrupting another person.

AI can make that judgement better informed. I would be wary of assuming that makes the judgement unnecessary.

“More leads” is a surprisingly small ambition

None of this is an argument against AI lead generation.

Using AI to understand markets, discover companies and surface signals is genuinely valuable. I use it extensively myself.

But once those capabilities exist, “generate hundreds more leads” starts to feel like a rather unimaginative destination.

The bigger opportunity is to build better filters.

Use automation to watch more of the market. Use it to gather more context. Let it reject obvious poor fits before anybody spends time on them. Use it to make the remaining situations richer and easier to assess.

Then apply human judgement where judgement is actually required.

A good outbound system does not need to maximise the number of companies entering the CRM. It needs to improve the quality of the decisions made before somebody starts a conversation.

AI gives us the ability to look at far more of the market than before.

There is no requirement to contact all of it.