In a May 21, 2026 column, Lisa Larson — Kelley warns that AI speeds product marketing but often yields polished, shallow positioning. She advises feeding models real evidence, forcing specificity, and changing workflows so claims are tied to data.
Lisa Larson — Kelley warned in a May 21, 2026 column that AI has accelerated product marketing workflows but frequently produces polished yet shallow positioning that can mislead buyers and teams. That matters because unchecked, AI-generated claims can shape buyer perceptions and derail sales conversations if they lack verifiable backing. Marketers and product teams must link messaging to evidence or risk shipping generic, untrustworthy content.
Larson — Kelley lays out three concrete fixes for AI-assisted messaging: do not assume models understand your product or buyer; feed AI evidence instead of empty prompts; and force uncomfortable specificity that exposes unsupported claims. These are practical, procedural measures aimed at turning AI from a guess generator into a drafting tool anchored in fact.
She points to familiar hollow phrases—"built for modern teams," "streamline workflows," "unlock efficiency at scale"—as symptoms of speed without strategic rigor. To replace such defaults, the column recommends supplying models with live signals: sales call transcripts, win‑loss data, product usage patterns, customer objections, competitor movement, and market shifts. Larson — Kelley cites synthetic audience modeling tools like Mavera as early attempts to tie AI outputs to real-world inputs.
The problem, she explains, is technical and predictable: large language models are trained to predict the most plausible language, not to infer a company’s unique market position. When prompted without evidence, models return the statistically likeliest articulation of product marketing — the average version — rather than an accurate, differentiated story. That statistical bias produces smooth, persuasive copy that can nonetheless be misleading.
The consequences are practical for both marketers and builders. Outputs that read finished may never have been strategically vetted; claims can lack proof; and teams risk shipping messaging that won’t hold up under sales scrutiny or competitive pressure. Larson — Kelley argues AI should not be used to automate guesswork: if a generated claim cannot be tied to specific evidence, it should not drive positioning.
For practitioners she offers workflow changes that fit existing processes. Before prompting, clarify who the buyer is, what choices they face, and what recently changed to make the product relevant. Pull a few real examples — direct customer quotes, summaries of lost deals, usage anomalies — and surface the patterns you see. Build prompts as briefs that include context, constraints, audience tension, and market inputs rather than one‑line requests, and use AI to draft and synthesize, not to invent the underlying proof.
Sources
Replies (0)
No replies in this topic yet.
