Gandhi Mathi Nathan Kumar is a reliability executive and Senior Staff Incident Commander at Aurora Innovation, specializing in scaling resilient incident operations across autonomous vehicles, cloud communications, and financial services. Industry publications, including the Disaster Recovery Journal, Rootly, and Uptime Labs, frequently highlight his thought leadership on incident management, focusing on where AI should (and shouldn’t) be applied during outages, optimizing the critical first 15 minutes of incident response, and mitigating on-call team burnout.
The conversation surrounding artificial intelligence has evolved rapidly over the past several years. What began as discussions about automation has expanded into broader questions about leadership, governance, responsibility, and the future relationship between technology and human expertise.
For Gandhi Mathi Nathan Kumar, the answer lies not in choosing one over the other, but in recognizing that the strongest operational outcomes emerge when AI and experienced professionals work together.
“Artificial intelligence is changing how organizations operate,” says Gandhi Mathi Nathan Kumar. “Our responsibility is to ensure those changes strengthen human decision making rather than diminish it.”
That perspective reflects a practical approach to AI adoption, particularly in environments where operational decisions carry significant consequences.
The conversation surrounding artificial intelligence has evolved rapidly over the past several years. What began as discussions about automation has expanded into broader questions about leadership, governance, responsibility, and the future relationship between technology and human expertise.
For Gandhi Mathi Nathan Kumar, the answer lies not in choosing one over the other, but in recognizing that the strongest operational outcomes emerge when AI and experienced professionals work together.
“Artificial intelligence is changing how organizations operate,” says Gandhi Mathi Nathan Kumar. “Our responsibility is to ensure those changes strengthen human decision-making rather than diminish it.”
That perspective reflects a practical approach to AI adoption, particularly in environments where operational decisions carry significant consequences.
Technology Continues to Evolve
Artificial intelligence now supports organizations across countless operational activities. It can identify anomalies, analyze trends, organize information, and provide recommendations with remarkable speed. These capabilities create significant opportunities to improve incident management, allowing organizations to respond more efficiently while reducing manual workloads.
Yet Gandhi Mathi Nathan Kumar believes technology should remain focused on enabling better decisions rather than making every decision independently.
“Technology provides remarkable analytical capability,” he explains. “People provide accountability, judgment, and the ability to understand context beyond what data alone can reveal.”
That distinction becomes increasingly important as organizations continue expanding AI into more complex business functions.
The Human Element Remains Essential
Critical incidents rarely unfold according to predictable patterns. Unexpected developments, incomplete information, stakeholder expectations, and operational priorities often require experienced professionals capable of interpreting situations as they develop.
While AI provides valuable analytical support, Gandhi Mathi Nathan Kumar believes human expertise remains essential for determining how organizations should ultimately respond. “AI gives organizations more information,” he says. “Experienced professionals determine how that information should influence action.”
That partnership allows organizations to benefit from technological speed while maintaining thoughtful oversight during important decisions.
From Alert Noise to Actionable Signal
Most incidents begin with a signal: an alert, a customer report, or a metric drifting out of bounds. Modern environments produce a huge number of these signals, and responders are expected to separate what matters from what does not, often in the middle of the night.
AI is well suited to this first stage. It can correlate related alerts, suppress duplicates, group symptoms that share a common cause, and surface the few signals most likely to need attention. When done well, this reduces alert fatigue and gives on-call engineers back time and focus.
“The goal isn’t fewer alerts for the sake of fewer alerts,” Gandhi Mathi Nathan Kumar says. “The goal is that when someone is paged, it truly matters, and they arrive with context instead of starting from zero.”
Deciding what counts as an incident, how severe it is, and who needs to be involved still calls for human judgment. Severity depends on business impact, safety, and customer trust, and those are things an algorithm can estimate but cannot own.
Incident Command in the Age of AI
During a major incident, the incident commander coordinates people, sets priorities, and keeps the response moving. That role depends on clear communication and calm decision-making under pressure.
AI can make this role far more manageable. It can summarize a fast-moving response channel, track open action items, draft timelines in real time, and remind the team of relevant runbooks or similar past incidents. Instead of spending energy on note-taking and status tracking, the commander can focus on leading.
“An incident commander’s most valuable resource is attention,” he explains. “Every task AI can take off their plate is attention they can give back to the people and the decisions in front of them.”
There is a clear line, however. Decisions such as rolling back a release, taking a system offline, or escalating to leadership carry real consequences, and they need a named person who is accountable for making them.
Communicating with Clarity and Trust
Incidents are as much about communication as they are about technical recovery. Executives, customers, partners, and regulators all need accurate, timely updates written for their own needs.
AI can draft stakeholder updates, translate technical detail into plain language, and keep messaging consistent across channels. That speed matters when expectations are high and every minute without an update erodes confidence.
But trust is built by people. What to disclose, how to acknowledge impact, and when to commit to a timeline are judgment calls that shape an organization’s reputation.
“AI can help you write the update faster,” Gandhi Mathi Nathan Kumar says. “Only people can decide what you owe the people reading it.”
Learning After the Incident Ends
The most important work in incident management often happens after service is restored. Retrospectives are how organizations understand what happened, why it happened, and what must change so it does not happen again.
Here AI offers real promise. It can rebuild detailed timelines from chat logs, alerts, and deployment records. It can find recurring patterns across months of incidents and point out systemic weaknesses that no single review would reveal.
Still, a meaningful retrospective requires something no tool provides on its own: a blameless culture where people feel safe sharing what they saw, what they assumed, and where processes let them down.
“Data tells you what happened,” he says. “Conversations tell you why. Lasting improvement needs both.”
Protecting the People Who Respond
Behind every incident is a team of people, many of whom carry on-call responsibilities on top of their regular work. Burnout, fatigue, and the stress of high-stakes decisions are real risks to both individuals and organizations.
Used thoughtfully, AI can reduce that burden by handling repetitive tasks, cutting unnecessary pages, and speeding up diagnosis. Used carelessly, it can add noise, create false confidence, or leave responders unsure whether to trust what they are seeing.
“Responders need to know when to rely on the tool and when to question it,” Gandhi Mathi Nathan Kumar notes. “That trust is earned through transparency, not assumed.”
That is why training, clear guidelines, and honest feedback loops matter as much as the technology itself.
Building the Next Generation of AI
Looking ahead, Gandhi Mathi Nathan Kumar believes the most successful organizations will not necessarily be those with the most sophisticated algorithms. Instead, success will belong to organizations capable of creating productive collaboration between intelligent technology and experienced people.
Responsible governance, transparent implementation, and continued investment in professional expertise will all remain essential as AI continues evolving. In incident management, that means defining clearly where automation can act on its own, where it must ask for approval, and where people must always stay in control.
“Innovation succeeds when it strengthens people,” he concludes. “That is how organizations create technology that delivers lasting value.”
Artificial intelligence will continue reshaping incident management for years to come. Yet as Gandhi Mathi Nathan Kumar sees it, the future will not be defined by replacing human expertise with automation. It will be defined by building systems where technology enhances the judgment, experience, and leadership that remain at the heart of every important decision.
