When the Sagiss survey asked workers which situations most likely lead to mistakes, 55% pointed to rushing between tasks or meetings. Another 48% cited multitasking. High message volume, constant context switching, and the pressure to stay responsive across multiple channels have created conditions where careful verification is difficult to practice, even for people who know exactly what they should be doing.

With AI, attackers have introduced a weapon into that environment that is purpose-built to exploit overworked, distracted employees.

Seventy-two percent of workers in the Sagiss survey said phishing attempts are more convincing than a year ago because of AI-written language. Nearly 65% said an AI-generated message could likely impersonate someone they work with. More than half (57%) said AI makes phishing harder to spot because it feels more professional. This is the reality many workers face in their inboxes.

The change in attack quality has been rapid. Large language models can generate polished, contextually accurate messages at scale. They can draw on publicly available information, such as LinkedIn profiles, company websites, and prior data leaks, to produce content that mirrors a colleague’s tone or references a real project. Thirty-three percent of survey respondents said they’ve noticed better grammar and writing in suspicious messages over the past year. Twenty-seven percent pointed to greater personalization. Twenty-six percent said the tone now feels more natural and human.

People spent years learning to spot awkward phrasing, generic greetings, and mismatched urgency. Now, those red flags are precisely the things AI removes first.