Sun Country Airlines' initial push-based agent assist deployment—which automatically surfaced pre-defined information based on detected topics—achieved only 1% adoption and agent feedback indicated it was intrusive and unhelpful, prompting a shift to a pull-based model where agents initiate a generative AI chatbot connected to the full knowledge base. Industry consensus confirms that agent assist effectiveness depends critically on design, delivery, and knowledge management; the pull approach allows agents to request help on their own terms whilst the system reasons through scenarios rather than relying on keyword matching, making it non-intrusive and more relevant to actual agent needs.
Agent assist can work, but design, delivery and how you manage knowledge make a huge difference — here’s one contact center leader’s experience.
Contact center experiments with push vs pull for agent assistance No Jitter