As enterprises continue their digital transformation journeys, the challenge of building truly reliable systems at scale has never been more critical. While Chaos Engineering emerged as a powerful approach to understanding and strengthening systems�pioneered by companies like Netflix�many organizations have struggled to move beyond basic implementations to comprehensive reliability programs that deliver measurable business value. Kolton Andrus, CEO and Co-Founder of Gremlin, has been at the forefront of this evolution, helping Fortune 100 companies build successful reliability programs while observing firsthand the gap between having the right tools and knowing how to use them effectively.
In this exclusive VMblog interview, Andrus discusses the current state of Chaos Engineering adoption across enterprises, reveals Gremlin’s latest innovation in Reliability Intelligence that leverages AI and a decade of chaos experiment data, and shares his perspective on the emerging trend of “AI SRE” solutions and whether artificial intelligence can truly handle the accountability demands of production operations.
++
VMblog: First of all, let me start by saying our readers will be very happy to hear from Gremlin. Before we get into the announcement, can you provide us with an overview of what Gremlin has been up to the past few years?
Kolton Andrus: We’ve been hard at work in the lab, cookin’. We found that while Chaos Engineering was a necessary component to building reliable systems, our customers needed more. They need a comprehensive approach to a complex problem that provides them the tools to be successful at scale. Leadership needed visibility and accountability into the action taken (and an opportunity to reward the right behavior!). Engineers and SREs needed clearer guidance on what to test (e.g. Test Suites, better insight into the topology and behavior of their systems (e.g. Dependency Discovery, Intelligent Health Checks, and Monitoring Integrations). And our customers want to partner with us, to hear about our experience and expertise in this space – we’ve built that into Gremlin wherever possible.
VMblog: What is the state of Chaos Engineering today? Is it still a practice primarily for the cool kids at Netflix or has it seen adoption across the enterprise?
Andrus: Chaos Engineering is and always has been a valuable engineering approach to understanding and strengthening systems. What we have learned is what many software industries have learned – the hard part is getting the people to do the right thing. We have the best tools, we have extended support into the vast majority of systems and platforms in use today. But none of that matters if the team doesn’t know what to do. This can be that the team lacks confidence in what to test, their ability to safely execute the test, or how to interpret the results. One sad story we’ve heard over and over again, if Leadership doesn’t have visibility into the activities and results of the team’s reliability efforts, then future budgets may be cut or eliminated altogether.
So I would say we’ve seen many of the Fortune 100 be quite successful with their Reliability programs, with Gremlin being a core component of that. But we see many companies flounder with a one-off, “let the SRE’s handle it”, no visibility approach that is basic Chaos Engineering.
VMblog: Today, Gremlin is announcing Reliability Intelligence. Can you explain to our readers what it is and how it works?
Andrus: At Gremlin, we’ve always focused on how to make it easy to do the right thing. Running Chaos Engineering experiments, interpreting the results and turning them into actionable outcomes is difficult. It requires experience in how the system behaves, a variety of tools, and knowledge of the underworking code. Wouldn’t it be great if we could leverage the expertise of great engineers, the outcomes of millions of chaos experiments and the rise of AI to help us more quickly find what’s wrong and recommend a fix?
That’s exactly what we’re doing with Reliability Intelligence. We’ve taken all the data we’ve collected at Gremlin over the past decade and synthesized it with the expertise of our team and the industry experts to build a system that is able to execute tests, analyze results and recommend remediations.
We also know the power of being able to leverage and explore your own data. So we’re launching our MCP server as part of our Reliability Intelligence offering.
The best part of this announcement? We’ve already been finding a ton of value internally with these new tools. We use Gremlin as a regular part of our internal reliability program. We’ve uncovered bugs that would have been real problems in production before they were ever released. While beta testing the MCP server, we found multiple opportunities for improvement while analyzing our own data. So we’re confident that these new tools are going to be impactful out of the gate in making our customers more effective and more reliable.
VMblog: There’s a trend in the market where solutions are positioning themselves as an “AI SRE” — what do you think of this trend and can AI really do the jobs of folks on-call?
Andrus: I think it will be some time before we’re able to delegate mission critical operations in production to an AI. This reminds me of an old IBM quote: “A computer can never be held accountable – therefore a computer must never make a management decision.” This accountability for actions that can make or break production is not something I see delegated in the near future.
I think building tools that enable SREs to be more productive, to more quickly analyze the system and understand potential issues, will continue to be important.
##
