$28,492.61 USD.
That was the AWS spend in a single month. I looked at the figure in the Cost Explorer dashboard and refreshed the page expecting an error or an untagged resource. It wasn't.
The figure was real and rising rapidly.
Our client, a B2B SaaS platform with heavy data consumption, had doubled its user base in six months. The product was growing, the team was shipping features quickly, but the infrastructure was drowning and the cloud bill had become a runaway train.
In theory, cloud costs should scale with usage. In practice, they were scaling due to inefficiency.
The architecture, based on traditional EC2 instances, RDS clusters, and continuously running workers, was designed for availability, not efficiency. At two in the morning, compute and storage services were still active, consuming budget as if it were peak hour.
As experienced AWS consultants, we detected that it wasn't just a cost problem; it was a design problem. What followed was rethinking everything from scratch. Below we explain exactly what we did and how.
PHASE 1 Understand the problem
Before fixing anything, we had to identify what was driving the spend. We analyzed AWS billing reports, usage dashboards, and CloudWatch logs to see where the money was going and which components consumed the most resources.
We detected clear patterns: EC2 instances running 24/7 even without traffic, RDS databases overprovisioned for peaks that rarely occurred, background jobs running continuously, and a lack of autoscaling in many services. Many tasks were triggered by cron rather than actual user activity, so the system was always on.
It became clear that the problem wasn't just the bill; it was the way the system was designed to operate. Optimizing costs wasn't enough; a new architecture was needed.
PHASE 2 Serverless migration planning
With the diagnosis ready, we planned the migration without rushing. The system was in production and any error could affect the business, so we opted for an incremental approach with a clear goal: first replace the most expensive and inefficient parts.
We decided to start with background jobs that don't require continuous execution, scheduled data processing tasks, and internal flows that aren't part of the critical path for the user. These were low-risk changes with high cost impact.
Chosen tools
We opted for a serverless architecture that only consumes when used. We used Lambda for serverless code, API Gateway for the entry layer, Amazon SQS for queueing tasks, Step Functions for orchestration, EventBridge for events, and DynamoDB as scalable, cost-effective storage. With this stack, we paid for usage and avoided idle resources.
Migration approach
Strict steps followed: start with non-critical tasks, test in staging, deploy with feature flags for quick rollback, and closely monitor performance and cost. We didn't just change technology; we changed how the system operates.
PHASE 3 Execution
We started with a nightly job that processed user reports. It was simple, isolated, and consumed many resources. We rewrote that job as a Lambda function, connected it to EventBridge and SQS for triggers on new data and queueing, configured CloudWatch to monitor executions and errors, tested in staging, and gradually rolled out with feature flags starting with 10 percent of users.
We also built a CI/CD pipeline with GitHub Actions and AWS CodePipeline for automated, secure deployments. That pattern was replicated week after week, transforming functions and processes one by one into serverless, event-driven components.
We faced challenges along the way: some Lambdas were slow to start, which we solved by reducing packages and using Provisioned Concurrency where needed. Migrating from RDS to DynamoDB required rethinking data models, but it gained in performance and cost. Debugging with microservices was more complex, and we resolved it with centralized CloudWatch logs, AWS X-Ray, and Grafana dashboards.
Little by little, we replaced the monolithic system with event-driven components. Performance improved, latency dropped, and most importantly, costs began to decline.
PHASE 4 Results
The transition to serverless significantly reduced the cloud bill. In this case, we achieved a nearly immediate reduction of 60 percent, and in optimized scenarios with greater refactoring, reductions close to 90 percent are possible depending on the previous load and architecture.
The system became faster and more stable under traffic spikes. Security and monitoring improved thanks to more restrictive access controls and centralized logging. The development team gained speed by working with small functions, automated deployments, and simple rollbacks.
In the end, it wasn't just about saving money. It was about building better, moving faster, and regaining control over infrastructure and cloud costs.
About Q2BSTUDIO
At Q2BSTUDIO, we are a software development company specialized in custom applications and custom software. We offer AWS and Azure cloud services, artificial intelligence solutions and AI for businesses, AI agent implementation, business intelligence and visualization services with Power BI. Additionally, we provide cybersecurity expertise to protect data and operations in cloud and on-premises environments.
We work with B2B clients to design efficient and scalable architectures that combine custom applications, artificial intelligence, and business intelligence services. Our goal is to reduce operational costs, improve performance, and accelerate value delivery through practices such as serverless adoption, CI/CD automation, and comprehensive observability.
If you need to optimize your cloud costs, migrate workloads to serverless solutions on AWS or Azure, integrate AI agents, or develop custom software with security and scalability, Q2BSTUDIO can help. We combine technical expertise in cybersecurity, artificial intelligence, and Power BI to deliver complete solutions that improve both technical efficiency and return on investment.
Keywords
custom applications custom software artificial intelligence cybersecurity AWS and Azure cloud services business intelligence services AI for businesses AI agents Power BI
