Our Take
Resolve AI is naming a real problem (deployment systems are still manual) but the article doesn't show evidence the problem is being solved.
Why it matters
Deployment and production monitoring remain largely manual work at most organizations, creating both reliability gaps and velocity drains. If AI can credibly automate these layers, it reshapes where engineering teams spend their cycles.
Do this week
Platform leaders: audit your current deployment and monitoring toolchain this week to identify which steps still require human approval or manual intervention, so you can benchmark where AI agents could reduce toil.
The deployment automation gap
Resolve AI's CEO Spiros Xanthos has identified software deployment, production monitoring, and workload management as areas where automation has lagged, according to a McKinsey Insights interview. Unlike code generation, which has seen rapid AI adoption, the systems that push code to production and keep it running remain largely manual and human-gated.
Xanthos frames the opportunity around three problem areas: reliability (systems fail when humans miss signals), pace (manual gates slow release cycles), and workload efficiency (humans still make decisions about resource allocation that AI could predict). The gap is not technical hubris—it is an observed reality that these layers have resisted the same automation wave that touched testing, CI, and infrastructure-as-code.
Why this matters now
Deployment and production systems are the last mile of the software factory. Code quality and velocity mean little if moving code to users requires human sign-off, manual monitoring, and reactive incident response. Organizations that run lean platform teams feel this acutely: they cannot hire enough humans to keep up with deployment and monitoring complexity.
AI agents—systems that can observe production state, make decisions, and take action without human intervention each cycle—could compress these bottlenecks. The catch: this is speculative. Xanthos is describing a problem and a direction, not a shipping solution with independent benchmarks or customer deployment data. McKinsey is platforming the CEO's vision, not reporting a capability that exists in production.
What to do this week
If you own a deployment or observability platform, or you lead a platform team, inventory your current toolchain. Map which decisions still require human judgment: approval gates, alert response, resource scaling, rollback decisions. Separate the ones that are gated for compliance or safety from the ones that are gated because your tools cannot act autonomously. That gap is where AI agents will land first, and where your roadmap should look.
Do not wait for Resolve AI or any vendor to ship. Identify your own highest-friction manual step and prototype an AI agent (using an off-the-shelf LLM API and your own production logs) to see where the real friction is: latency, reliability, or governance.