Sales

Sales work splits cleanly into two piles that feel like one. There is the selling, which is judgement about a person and a situation. And there is the administration around it: qualifying, chasing, logging, reminding. The second pile is what stops the first from happening.

Nobody was hired to write the third follow-up email. It still has to be written, because deals are lost to silence far more often than to a competitor.

Qualification is a rule, until it is not

Most inbound qualification runs on criteria you could write on one page: company size, budget indication, whether they have a timeline, whether they are in a market you serve. An agent applies those consistently, at three in the morning, on the enquiry that came in over the weekend.

What it cannot do is recognise the enquiry that breaks the rules and is worth taking anyway. The badly written email from someone who turns out to be the largest account you will sign this year does not score well on criteria. That is why qualification output is a recommendation rather than a decision, and why borderline cases go to a person instead of into a rejection.

Follow-up is where the measurable money is

Follow-up is the least glamorous part of sales and the one with the clearest arithmetic. A sequence that goes out on time, stops the moment someone replies, and picks up again after a defined silence does not require creativity. It requires only that it happens, which is exactly what a busy person cannot guarantee.

An agent handles that reliably and, more importantly, stops reliably. The failure mode of automated follow-up is the sequence that keeps firing after a customer has replied, which reads as carelessness and undoes whatever the sequence was meant to build.

The CRM problem, honestly

Every sales team has a CRM that is out of date, because updating it is work that benefits someone other than the person doing it. An agent that writes back what actually happened after each interaction fixes the symptom.

It does not fix the underlying issue, which is that a CRM full of accurate records is only useful if someone acts on them. Automating the data entry makes the pipeline visible; it does not make anyone call the leads.

What we will not automate here

Anything a customer would feel misled by. An agent that writes as though it were a named salesperson, or that holds a conversation the buyer believes is human, buys short-term efficiency at the cost of the relationship the sale depends on.

Pricing and terms stay with people, for the same reason: a discount given by software is a discount nobody can defend in the next negotiation.

How this gets built

We map your current pipeline, including the stages that exist in practice but not in the CRM. The qualification criteria get written down, which is often the first time they have been. Then one part goes live, usually follow-up, because it is the easiest to measure: either the messages went out on schedule or they did not. The four stages are described in full under how a project runs.

Sales work usually touches more than one system, which is what determines the scope. Book a demo to walk through your pipeline and see which parts are worth moving across.

The question of which processes are worth handing over at all is covered in where AI agents actually save time.