Sales automation tools

Most sales automation tools are bought to send more first messages, which is the one job automation is worst at. The money is in the opposite direction: the follow-up nobody made, the quote nobody chased, the answer that took two days. Nothing about that is glamorous, and all of it is measurable.

The safe half and the dangerous half

Sales work splits into things a machine may do because they are about timing, and things it must not do because they are about judgement. The split is not obvious from a demo, so here it is plainly.

Safe: everything after someone has already engaged. The reminder on day three, the quote that was never answered, the customer who has not ordered in ninety days, the summary of what was said last time. This is scheduling and memory. It is boring, it works, and it is where the recovered revenue actually is.
Safe: preparation. Pulling the company's own numbers, the order history and the last conversation into one page before a call. Ten minutes per call, back, with no risk attached to it.
Dangerous: the first message to a stranger. An AI SDR sending volume is the fastest way to burn a domain and a name. It also runs straight into consent rules in Spain and the EU, which cold outreach automation does not get to opt out of.
Dangerous: anything that pretends to be a person. A fake name on an automated message is a small lie that costs a big amount exactly once, when someone notices. Say what it is.

What to build first, and it is not the CRM

Ask what happens today to a lead that says 'send me the details and I will look next week'. In most businesses the honest answer is: nothing, unless someone remembers.

A single list of what is open and who owes the next move. Not a pipeline picture — a list of overdue moves.
Automatic follow-up on your own quotes, on a schedule you set, stopping the moment a human replies.
The dormant-customer sweep: who used to buy and stopped. Usually the highest-return list in the business and nobody has time to build it.
A short brief before each call, assembled from what you already hold.
Only then the reporting. Sales automation software that starts with dashboards produces a prettier view of the same missed follow-ups.

How this goes wrong

The tool becomes the job. Fields to fill so the automation can fire. If the rep is now feeding the machine, the machine is not working for the rep, and they will quietly stop using it.
The sequence keeps running after a reply. The single most common defect in AI sales automation and the most damaging. A human answers and gets nudged twice more. Every sequence needs a stop condition wired to a real reply, and it needs testing.
Personalisation that is obviously generated. A first line assembled from a company description reads as generated to the person it is about. It is worse than no personalisation, because it announces the volume.
Nobody deletes the rule that stopped making sense. Rules accumulate. Once a quarter someone has to read all of them and delete some. Put it in a calendar or it will not happen.

Skip this if

You have fewer than a couple of dozen open opportunities. A written list and a weekly hour will beat any tool, and cost nothing.
Your sales cycle is one conversation, one decision. There is no follow-up to automate.
You want it mainly to send cold outreach at volume. I will not build that, and in Spain and the EU the consent rules make it a poor idea on its own terms.
Your customer records are so incomplete that the first automated message would be wrong. Fix the records; that is the cheaper project and it helps everything else.

The chasing, already automated

Advertising manager — roughly 600 campaigns pulled, merged and judged every morning, each with a plain verdict: making money, wasting it, or worth more budget.
Sales funnel report — about 750 products read every day, with only the ones that have a real problem flagged and a plain diagnosis written for each.
New-arrivals radar — the market scanned on a schedule so the movers surface early — the same discipline applied to what to sell rather than to whom.

Honest answers

Which sales automation tools are worth it for a small team?

The question is which job, not which product. Follow-up on open quotes, the dormant-customer sweep and a pre-call brief pay back almost immediately for a team of any size. Sequenced cold outreach at volume rarely does, and carries consent risk in the EU.

Can AI write the outreach?

It can draft. Whether it sends is a different decision, and I would keep a human on the first message to a stranger. Where it earns its keep without argument is the fourth follow-up and the summary of the last conversation — the messages that otherwise never get written at all.

What about an AI SDR that books meetings for us?

Booking a meeting with someone who already raised their hand is fine and mostly a calendar problem. Generating that hand-raise at volume from cold contacts is what the term usually means, and that is the part I would not build for a business whose name matters more than its send volume.

Do we need a CRM first?

No, and starting there is a common way to spend six months without a result. Start with the list of overdue moves. If a spreadsheet holds it, a spreadsheet is fine until it visibly stops being fine.

How do you stop a sequence annoying someone who replied?

The stop condition is wired to a real inbound reply and it gets tested deliberately, including replies from a different address and out-of-office bounces. This is the defect I look for first in an existing setup, because it is nearly always there.

Tell me what happens to a lead that goes quiet

Describe your last month honestly: what was quoted, what was chased, what was forgotten. I will point at the one automation worth building first and tell you what it would recover. If the honest answer is a weekly hour and a written list, you will get that instead — the first hour with me is free.

Book that hour →