The funnel is taught as fact — but it was falsified 25 years ago. So what should an AI actually optimise for?
The sales funnel and pipeline look precise and scientific. In truth the funnel began as AIDA, sketched around 1898, with the funnel shape added in 1924 — and research falsified its core claim decades ago. Point an AI prediction engine at a tidy funnel and you optimise a journey buyers don’t take.
Video 3 splits AI in sales into three jobs: predictive analytics (who is most likely to buy), generative AI (drafting the outreach — emails, messages, even calls), and automation (moving deals along). It runs on two diagrams: the funnel (awareness → interest → intent → consideration → decision) and the pipeline (lead → qualify → contact → demo → negotiate → close). Fine as checklists.
The funnel didn’t come from a lab. It began as AIDA — attention, interest, desire, action — sketched by advertising man E. St Elmo Lewis around 1898. The funnel shape was added later, in a 1924 sales manual by William Townsend. It was never a peer-reviewed model; it just stuck. The sales pipeline has no academic basis at all — it is CRM-vendor picklist convention.
Keep the funnel as a checklist of stages to serve — not a law of how people buy. Predict from real signals in the mess, and audit any AI-sales vendor’s headline number: no method behind it means marketing, not evidence.
Where did the sales funnel come from, and what is its academic status?
What is the ‘messy middle’, and why does it matter for AI in sales?
Name the three AI jobs in sales and one cautionary tale from the hype.