This research analyses how early detection and mitigation responses can reduce the impact of a deepfake-based cyber-attack, using the real-world case of KnowBe4 as inspiration. In that case, an attacker used deepfake technology to impersonate a legitimate job applicant and infiltrate the company. Based on this scenario, we developed a dynamical system model to test various combinations of detection thresholds and mitigation speeds, focusing on two primary outcomes: data access and ransom threat. We ran 25 simulation scenarios to explore how these variables affect system behaviour over time. The results showed that when detection was weak, even with fast mitigation, both outcomes increased significantly. In contrast, when detection was strong, the system stayed more stable, and threats were less likely to escalate. Our What-If analysis confirmed that the most severe outcomes occurred under low detection conditions, showing that early detection acts as the first and most effective barrier against deepfake-based attacks, while mitigation is more effective when used as a supporting response rather than a standalone solution. Based on these findings, we recommend that organizations prioritize investments in advanced detection tools, especially those capable of identifying deepfake content or synthetic identities early in the process.