Sep 30, 2026
AI SRE Buyer's Guide
Learn how to evaluate and adopt AI SRE in production.
Learn how engineering teams are transforming software engineering with agentic AI.
Coinbase is one of the world’s largest crypto platforms, powering trading, payments, and custody for customers across more than 100 countries. The company serves over 120 million users who trade more than 40,000 crypto assets. As of late 2025, Coinbase manages over $500 billion in assets on the platform and processes close to $300 billion in quarterly trading volume, supported by more than 4,700 employees globally.
Behind the scenes, this scale runs on a large microservices architecture on Kubernetes, backed by many specialized data stores and services. Reliability is business-critical, and incidents can have immediate financial and regulatory impact, so SREs and engineers need fast, trustworthy answers whenever something looks off in production.
On a crypto platform with billions in daily trading volume, even small pockets of downtime or increased latency can stall deposits and withdrawals, cause trades to miss price windows, and trigger spikes in support tickets. For Coinbase’s central engineering team, slow or noisy incident response means more revenue at risk in the moment and less time to invest in the shared platform that keeps dozens of product teams moving.
Before Resolve AI, on-call engineers had to manually stitch together context from deploy pipelines, Terraform events, Datadog, and custom dashboards to understand what changed and whether an alert was real.
This created friction across both incidents and day-to-day production work:
Coinbase adopted Resolve AI as both an AI SRE on the incident bridge and a way for engineers to code with production context. During incidents, Resolve AI is their always-on AI SRE, auto triaging and running investigations across infrastructure, deploy data, and telemetry, and answering “what changed” in real time. Outside of incidents, engineers use Resolve to bring live production signals into their day-to-day work so design, reviews, and code changes are grounded in how services actually behave in production, not just what static dashboards or docs suggest.
On the incident side:
For everyday production work:
Coinbase needed an AI system that could understand their production environment and give engineers instant, accurate context, not a generic chatbot.
All of this is currently done without exposing source code or proprietary runbooks, which keeps the security and governance model simple while still giving Resolve AI enough context to be useful. As code access comes online, Coinbase expects to expand what it automates with Resolve AI.
Resolve AI compresses Coinbase’s incident response and day-to-day production work. It cuts investigation time, brings median first response measured in seconds, and gets to likely root cause in minutes for critical incidents. That shortens the window where customers might see failed orders, delayed transfers, or slow trading experiences, while giving engineers the production context they need to ship changes with confidence.
Coinbase benefits with Resolve AI:
The result is an AI layer that supports everyday delivery and operational decisions at Coinbase, not only emergency response, so reliability and velocity move in the same direction instead of competing.
As Coinbase continues to scale its global crypto platform, Resolve AI is becoming a core part of the production backbone, serving as its AI for prod and woven into every on-call rotation and reliability decision. By pairing always on incident investigations with the ability to design and ship changes using real production context, Coinbase is building a future where reliability and shipping velocity reinforce each other so the core flows of global crypto trading, funding, and custody stay fast and available as the platform grows.
Machines on-call for humans
