Will JonesGandhi Mathi Nathan Kumar is a senior reliability and operations leader with over 16 years of experience driving incident command, service reliability, and crisis management across major technology and autonomous driving companies like Twilio and Aurora Innovation. Throughout his career in high-stakes environments, he has specialized in transforming incident management through AI integration, building automated response workflows, and scaling cross-functional operational frameworks to minimize downtime and elevate system resilience. Beyond his operational roles, he actively contributes to the field as a writer, speaker, and consultant on AI-assisted incident management and service dependability.
Artificial intelligence is becoming increasingly embedded in how organizations detect problems, interpret information, and respond to operational incidents. Yet the growing capabilities of AI have also sharpened a practical question for technology professionals: which decisions can machines support, and which still require experienced human judgment?
For Gandhi Mathi Nathan Kumar, that question is closely connected to his career experience and the lessons he has drawn from working around technology and incident management.
Rather than viewing AI primarily through the lens of automation, Kumar sees it as a decision-support capability. His perspective centers on combining the speed and analytical capacity of AI with the context, accountability, and judgment that professionals bring to high-pressure situations.
In this Q&A, Kumar discusses the professional thinking behind that position, why incident response makes the limits of automation particularly visible, and what he believes organizations should consider as AI assumes a larger role in operational decision-making.
Q: How has your career experience shaped the way you think about artificial intelligence?
Gandhi Mathi Nathan Kumar: My perspective comes from looking at technology in terms of what it actually enables people to do. New technology can give professionals access to information faster, reduce repetitive work, and help identify things that might otherwise take much longer to find. But experience also teaches you that having more information doesn’t automatically mean you have the right answer.
That distinction has influenced how I think about AI. I don’t see the most interesting question as whether AI can perform a particular task. I think the better question is how that capability improves the decisions people ultimately have to make.
Q: Why is that distinction particularly important in incident response?
Kumar: Incident response puts decision-making under pressure. You may have incomplete information, circumstances can change quickly, and there can be several competing priorities at the same time.
AI can help organize information, recognize patterns, or surface something that deserves attention. But the person responding still has to understand the context. They have to consider what the information means for the organization and what the consequences of a particular response might be. That’s where human judgment remains very important.
Q: Where do you think AI can provide the most value during an incident?
Kumar: One of the strongest applications is reducing the amount of manual information processing that professionals have to do before they can start making decisions.
If AI can help identify relevant signals, correlate information, summarize what is happening, or bring an unusual pattern to someone’s attention, the professional can spend more time evaluating the situation. The objective shouldn’t simply be to automate more steps. It should be to improve the quality and speed of the overall response.
Q: Does that mean there are decisions you believe should remain with people?
Kumar: Yes, particularly when a decision carries significant consequences or requires context that isn’t contained in the data itself.
There is an important difference between recommending an action and being accountable for that action. AI can support the first. Organizations still need to think very carefully about the second. Human oversight shouldn’t be something added after an AI system has been deployed. It should be considered when the process itself is designed.
Q: How has professional experience affected your view of that balance?
Kumar: Experience makes you more aware of exceptions. A process may work very well under normal circumstances, but incidents are often defined by something unusual happening.
That is why I’m interested in the relationship between AI and experienced professionals. Technology can process information at a scale and speed that people cannot realistically match. People, however, can bring context and judgment to situations that don’t fit neatly into an established pattern. The two capabilities are different, and I think that difference is useful.
Q: There is considerable discussion about AI replacing human roles. Is that how you see the technology developing?
Kumar: I think replacement is too narrow a way to frame the discussion. In many operational environments, the more useful question is how AI changes the work people do.
If a system can handle repetitive analysis or help a professional identify what deserves attention, that person can concentrate on the parts of the job that require judgment, communication, and accountability. To me, that’s a much more practical way to think about AI.
Q: What do organizations sometimes misunderstand when they introduce AI into operational processes?
Kumar: There can be a tendency to focus heavily on what the technology can do and not enough on how people will actually use it.
An AI capability can be technically impressive and still create problems if employees don’t understand its role, don’t trust its recommendations, or aren’t clear about when they should intervene. Implementation therefore, isn’t only a technology question. It’s also a process and people question.
Q: What does responsible human oversight look like in practice?
Kumar: It means being deliberate about where people remain involved. Organizations should understand which decisions can be automated, which decisions require review, and who is ultimately responsible when an automated recommendation influences an important outcome.
Human oversight also needs to be meaningful. If someone is technically responsible for approving a decision but doesn’t have enough information or time to challenge the system, that isn’t necessarily effective oversight.
Q: How do you expect the relationship between AI and incident-response professionals to change?
Kumar: AI will become more capable, and professionals will become more accustomed to working with it. I think that makes the design of the relationship between the two even more important.
The goal should be to use AI where it has an advantage while preserving human involvement where judgment, accountability, and context matter most. As the technology improves, organizations will have more choices about what they automate. Having more choices makes it even more important to decide carefully which ones they should automate.
Q: What principle ultimately guides your view of AI and incident management?
Kumar: Technology should help people make better decisions. That’s the principle I keep coming back to.
AI can provide speed, scale, and analytical support. Human professionals contribute experience, context, and responsibility. Incident response is a good example of why those strengths should not be treated as competing with one another. The strongest model is one where the technology improves what experienced people are able to see and do while keeping meaningful decisions accountable to people.
That view places Kumar’s career perspective within a broader shift in thinking about AI. The question is moving beyond what machines can automate toward how organizations can structure collaboration between automated systems and the professionals accountable for their use.
For incident response, where decisions can be time-sensitive and consequences significant, Kumar’s position is straightforward: greater AI capability does not eliminate the need for human judgment. It makes deciding where and how that judgment should be exercised more important.