Sales is several jobs, not one

Prospecting, qualifying, scheduling, following up, note-taking, and closing are distinct activities with different requirements. Treating “sales” as a single thing to automate is why these projects either underdeliver or damage the pipeline.

What works: the admin around selling

Research before a call, updating records after it, drafting follow-ups, chasing scheduling, and flagging deals that have gone quiet. This is work salespeople dislike and skip, and automating it returns hours to the part only they can do.

What works: qualifying inbound

Responding immediately to enquiries, asking the qualifying questions, and booking the ones that fit. Speed genuinely matters here, and an always-on responder beats a human who replies the next morning — provided the qualification isn’t front-loaded onto the form.

Automate the work salespeople skip, not the work salespeople are for.

What doesn’t: outbound at volume

Generating large volumes of personalised-looking outreach is technically easy, commercially damaging, and increasingly a deliverability problem. It also trains your market to ignore you, which is expensive and hard to reverse.

What doesn’t: the negotiation

Understanding an unstated objection, judging when to push, and deciding what to concede are exactly the judgment-heavy, low-volume tasks where AI is the wrong component.

Measure pipeline, not activity

Emails sent and leads touched will rise; they always do. The honest measures are qualified conversations created and deals progressed, which is the downstream-confirmed version of resolution applied to revenue.