
Today we're launching Foresight AI, Gremlin’s agentic resilience product that analyzes and tests your systems for potential failures, fixes them, and verifies reliability at the speed of AI.
I've spent most of my career on call. At Amazon and Netflix, I served as a Call Leader, the person running the bridge when something big broke. Those years taught me the same lesson we founded Gremlin on: the best incident is the one that never happens. Foresight AI applies that lesson to the way teams build software today, enabling them to build resilience across everything they ship and run.
Why we built Foresight AI
Foresight AI is the fulfillment of the vision we’ve had from the beginning: help teams understand how their systems fail, fix the issues, and do it without requiring every engineer to be a distributed-systems expert. We started with powerful fault injection, but the homework was on the user. Now we can put agents to work to analyze the system, recommend and run the right tests, explain what happened, and deliver the remediation. Over time, we can take that even further—while keeping people in control of the level of automation they’re comfortable with.
What makes this urgent today is that AI is rapidly changing how we build software. Organizations are shipping more code than ever, but much of it goes out the door without human review. New failure modes live in that gap, and it’s getting wider with every release. The current tools and workflows are insufficient.
"AI-generated code is bringing architectural decisions that people aren’t really looking too hard at, and they’re catching them in hindsight, which will lead to reliability issues."
Software Engineering Leader at a global CRM leader
We want teams to have their cake and eat it too: keep the speed AI gives them, and put reliability guardrails in place so they stay on the track when something goes wrong.
What Gremlin Foresight AI does
Foresight AI identifies the failure modes in your services before they start alerting, recommends a fix, and validates that the fix works and holds over time.
Gremlin itself has powerful capabilities, but it sometimes requires more time and expertise than many teams have available. Foresight AI picks that up with agentic personas that can get real work done:
- The Analyst gathers details about your services, recommends tests to run, and analyzes test results to better understand your environment.
- The Tester schedules and runs reliability tests on your services.
- The Operator provides actionable conclusions and guided remediations based on testing outcomes.
- The Technical Program Manager (TPM) gives detailed technical advice based on the outputs of the other agents. It tells you what’s been tested so far, where your gaps in testing are, and where your reliability risks are.
But agents are only as strong as the data and context they can reach. Together, Foresight agents have access to custom tools and skills, plus something no other solution provides: the Failure Atlas.
The Failure Atlas is Gremlin’s proprietary record of how systems fail and recover, and it’s the only one of its kind. It draws on millions of fault injection experiments run over the past decade across tens of thousands of real distributed systems. Others have to guess what caused the incident, and by extension, how best to fix it. We can see the cause and effect, so we can understand exactly what went wrong and propose the right fix. Your data stays yours: no system data is sent off to train an LLM.
It proves the fix works
Most AI tooling for reliability works like this: something breaks, you throw a pile of context at an LLM, it tells you the likely cause, and the agent duct-tapes it and keeps moving. But did you fix the root cause, or a symptom?
Because Gremlin runs the tests, Foresight AI can rerun them after a fix. If the test still fails, we haven't found the fix yet, and we keep going until it passes. An alert going quiet doesn't prove anything. A test that passes under load, in realistic conditions, does. That evidence is how you build trust that your system works the way you expect. You stay in control every step of the way: Foresight AI doesn’t run a test or implement a remediation without your approval.
It understands your environment
Foresight AI sees your teams, services, tests, and environment the same way you do. When a reliability test fails or Gremlin detects a risk, Foresight AI starts from facts: how your systems are configured and which of your Health Checks triggered the failure. Using the Failure Atlas, it pinpoints the most likely cause and proposes a long-term fix instead of a temporary bandage.
We don't ask the LLM to guess. Foresight AI makes decisions on real data in the Gremlin platform, and it uses the LLM for what LLMs do best: summarizing, searching, and explaining results in plain English. We run Gremlin on Gremlin, and we hold our platform to five nines of availability. I want the same bar for accuracy.

It keeps your reliability program on track
A lot of companies would like to hand reliability to one team and say, "fix it for everybody." It doesn't work that way. Everyone who writes code, takes a dependency, or runs infrastructure has skin in the game, and there will never be enough SREs to go around.
Every successful reliability program I've been part of had someone driving it. Foresight AI includes an AI-powered Technical Project Manager (TPM) to play that role. The TPM tracks reliability health across your services, writes a weekly narrative report with recommendations, manages commitments, and sends updates to Slack. Every commitment gets a date, because plans without dates don’t get done. We run the TPM on ourselves, and I read our report every Sunday night to see what changed.

It proves the value of your work
Anyone who has run a reliability program eventually gets asked the hardest question in our field: how do you measure the impact of an incident that never happened? If you fix a bunch of issues and never have an outage, did you do anything? You did a lot, but the fire everyone sees is the one that gets attention. Without evidence of progress, teams get complacent. They decide nothing is going wrong, drop the work, and then the outages start again. You need to show the baseline, the score improvements, and the issues you found and fixed.
Foresight AI expands Gremlin's built-in reporting with fully customizable dashboards and reports. Describe what you need in plain language, and Foresight AI builds it for you, complete with charts, tables, tabs, and links you can view and export.

Build reliable systems at the speed of AI
AI lets your teams ship faster, but incidents and outages shouldn't be the price of that speed. Foresight AI gives you the tools to find, fix, and validate reliability risks as fast as you ship.
Foresight AI is available now as an add-on to the Gremlin platform. You can watch our CTO, Sam Rossoff, preview it below, or book a demo and set up a trial to see how it can work for your team.
When we launched Gremlin, our mission was to make the internet more reliable for all of us. Foresight AI now puts world-class reliability in reach for everyone.
Topics
Keep reading
Avoid downtime. Use Gremlin to turn failure into resilience.
Gremlin empowers you to proactively root out failure before it causes downtime. See how you can harness chaos to build resilient systems by requesting a demo of Gremlin.








