#automation

Marketers obtain stronger outputs from large language models when they treat the systems as sparring partners instead of answer machines, according to MarTech.

Tech columnist Christopher Mims described a test using prediction scenarios from Polymarket. Pure AI models outperformed human-only groups. Hybrid teams that accepted AI answers without challenge performed worse than AI alone.

Marketers obtain better results from large language models when they challenge outputs instead of accepting them directly. According to MarTech, an experiment described in The Wall Street Journal tested AI alone, humans alone, and hybrid teams on one-hour predictions of real-world events drawn from Polymarket scenarios. Pure AI models outperformed human-only groups but trailed prediction market benchmarks.

HubSpot released 15 updates in March 2026 that focus on increasing flexibility in automation, workflows, and CRM management, addressing previous limitations for managers and admins. These changes allow teams to adapt HubSpot more precisely to their business operations, particularly in areas like timing logic and data structure, according to [MarTech](https://martech.org/15-hubspot-updates-from-march-2026-managers-and-admins-need-to-know/).