Machines on-call for humans

Instead of building and innovating, engineers spend multiple hours productizing for every hour of coding–spinning up and optimizing new infrastructure or services, updating runbooks and dashboards, etc.. Heavy on-call and operational work lead to burnout and the tools are not sufficient–when things go wrong, developers and SREs have to jump between tools and manually connect the dots. At Splunk, where I led the Observability business, 90% of our SRE team resigned within six months due to burnout from stressful on-call duties, and almost all customer escalations I handled were triggered by reliability issues and outages, not by missing features. Complexity grew faster than tool capabilities. With AI accelerating software development, we anticipate even more software complexity and potential for outages.

At Resolve AI, we set out to rethink software engineering tools from first principles for an AI-first world. Our initial focus is on on-call, incident investigation and remediation—the most stressful operational tasks that, when mishandled, lead to outages and poor customer experiences. We are building AI that knows how to use tools and cloud services as an experienced developer. Agentic workflows are the right way but these won’t be implemented as simple wrappers on models, we need to develop deep comprehension of production systems, code and all the tools. We are also rethinking the UX because neither chat interface nor dashboards are appropriate for AI systems that do work.

But this is just the beginning. In real-world software systems, engineers spend most of their time understanding existing code and infrastructure and making changes. AI tools for software engineers must comprehend, end-to-end, both source code and production environments to be effective. Our vision is to continuously improve the depth (reasoning, training, accuracy) and breadth (tool integrations and data types) of our platform in close collaboration with our customers. This will enable us to assist humans more effectively across more tasks—from incident management to proactively improving reliability, security, testing, bug fixing and beyond.