AWS re:Invent 2025 - Human-Centric AI: Avoiding the High Cost of Getting It Wrong in IT (AIM205)

Human-Centric AI: Avoiding the High Cost of Getting It Wrong in IT

Overview of Ninja One

• Ninja One is a company that provides tools for IT teams to manage devices across various operating systems (Windows, macOS, Linux, iOS, iPadOS, Android)
• Businesses now rely more on endpoints than ever before, including laptops, servers, manufacturing equipment, ticket kiosks, digital billboards, and even movie theater projectors
• These devices need to be secured and supported throughout their lifecycle

Challenges in Modern IT Operations

• The last 5 years have seen a significant shift to remote and hybrid work, requiring IT teams to support users across different geographic locations
• IT teams are expected to do more with less, supporting more users and devices of varying types
• This is compounded by factors like mergers, acquisitions, changing compliance requirements, and auditing standards

The Need for Efficiency and Automation

• IT teams often have to "overproduce" to meet their goals, likened to "roasting a marshmallow 50 ft above a campfire"
• The solution is to automate easy issues and streamline more complex, time-consuming troubleshooting processes
• This is where AI can play a role in enhancing endpoint management

Applying AI in Endpoint Management

• Early AI applications in Ninja One include:
  • Scraping public data to assess sentiment around Windows patches and provide summaries to help decide on patch deployments
  • Automating the first touch response to common issues before escalating to a human technician
• The ambition is to have autonomous endpoint management where the AI can take direct action without human involvement
• However, there are significant risks with AI that need to be addressed:
  • AI is only as good as the data it's trained on, which can be incomplete, outdated, or inaccurate
  • Correcting misinformation spread by AI-driven decisions can be very difficult
  • Applying AI to powerful endpoint management tools could lead to undesirable outcomes, such as an AI deleting data to free up disk space

Categorizing AI Use Cases in Endpoint Management

1. Generative AI: Using language models to generate responses, scripts, or solutions
2. Agentic AI: Automating the first touch response to issues before escalating to a human
3. Predictive AI: Analyzing data to predict future device health and prevent problems

Balancing AI and Human Expertise

• The narrative of AI replacing people is not necessarily accurate, as past technological advancements have not eliminated the need for human involvement
• AI should be a reliable tool that augments and enhances human knowledge and judgment, not a replacement
• Successful integration of AI requires:
  • Transparency and auditability of AI decision-making
  • Continuous training and skill development for IT staff to understand, interpret, and question AI outputs
  • A "human in the loop" model where AI augments but does not replace human expertise

Conclusion

• Integrating new technologies like AI is challenging, but the long-term benefits come from aligning human expertise with the capabilities of these tools
• Ninja One's approach is to enable autonomous endpoint management, reducing the need for human involvement in common tasks while providing guardrails to prevent dangerous decisions
• Ongoing training, knowledge sharing, and a focus on human-centric AI are key to successful implementation and adoption