McKinsey wonders how to sell AI apps with no measurable benefits (www.theregister.com)

Consultancy McKinsey warns SaaS vendors that monetizing AI is hitting a reality check: many products can’t demonstrate measurable customer benefits, yet they risk driving up IT costs. Its report identifies three core roadblocks—weak evidence of savings (only ~30% of vendors publish quantifiable ROI), difficulty scaling adoption, and opaque, unpredictable pricing. McKinsey highlights stark numbers: fully AI‑enabling a typical customer‑service stack could raise prices 60–80%, while an MIT study and government trials show many deployments deliver little or no productivity gains. Buying is also shifting from IT to line‑of‑business leaders who demand outcome‑based value, not features. Technically and commercially this forces major shifts. Vendors must invest heavily in change management (McKinsey suggests ~$3 on change/ops for every $1 on model development), choose pricing units carefully (per‑user caps, metered throughput/tokens, per‑task or per‑outcome), and adopt hybrid subscription/consumption models that are frequently revisited as AI capabilities rapid ly evolve. Offsetting this, inference costs have been falling fast—McKinsey cites >80% annual declines recently—so delivery economics are improving, but vendors still need transparent, measurable value propositions and robust monitoring/training to justify higher prices and drive adoption.