The piece in London Business News puts a single question on the table: "Can AI sales assistants really increase sales conversion rates?" That qualifier — "really" — signals intent to scrutinize vendor claims rather than repeat them.

The question carries weight because sales-conversion claims rank among the most inflated categories in B2B software marketing. Vendors publishing percentages — framed as "X% lift in win rate" or "Y% shorter cycle" — rarely disclose sample size, control group, or the pipeline stage measured. A B2B deal moves through several stages, and an assistant that compresses meeting scheduling can post headline conversion numbers without touching the stage where deals actually close.

The article, as titled, appears positioned to examine that gap between assertion and measurement. The strongest version of that examination would separate three categories: what the assistant touches, how the vendor measured lift, and whether ACV moved alongside the conversion change.

What Conversion Metric Actually Moves?

Conversion rate is not one number. It splits across stage-to-stage progression — MQL to SQL, SQL to opportunity, opportunity to closed-won — and across funnel sources: inbound, outbound, partner. A tool that compresses discovery-call prep changes the number of opportunities a rep can run, which moves the SQL-to-opportunity ratio but does not necessarily move close rate.

Readers looking for a real answer should demand three pieces of evidence:

• The exact funnel stage where lift was measured
• The control group definition (same quarter prior year, same rep before and after, or matched cohort)
• The deal-size distribution, to confirm the tool did not push reps toward smaller, easier deals

Without those three, a posted percentage is not falsifiable.

Why Vendor Methodology Matters

Most published sales-AI case studies are observational, not randomized. The reported uplift therefore conflates the tool's effect with the effect of the team that adopted it — typically the most motivated reps or the accounts most likely to close anyway. A rigorous benchmark would either randomize roll-out or use a difference-in-differences design with a matched control team.

The London Business News framing, by quoting the word "really," suggests a piece more interested in that rigor than in a headline figure.

What This Means for Revenue Teams

For a team of 10 to 50 reps evaluating an AI sales assistant, the immediate practical move is to request the vendor's methodology document. If the vendor cannot produce one with sample size, stage definition, and deal-size distribution, the promised conversion lift cannot be trusted at face value. Run a 30-day internal pilot, pick one measurable stage, and compare against the trailing 90-day baseline.

A tool that survives that pilot is worth adopting. A tool that does not is worth de-installing.

The question London Business News raises is not whether AI sales assistants can help — they can, at obvious workflow tasks — but whether published conversion claims survive the kind of scrutiny the headline's word "really" demands. The next publication that prints the underlying methodology, rather than the wrapped percentage, will set the benchmark for everyone else.