Cast AI - EKS fully automated cost optimization and monitoring

What do you like best about the product?

I use CAST AI to optimize Kubernetes infrastructure cost and resource utilization across both production and non-production environments. CAST AI helps us automate node scaling, workload right-sizing, and instance type selection, which is a huge help. I also appreciate the dashboard view, which makes it easy to see all namespace workloads and identify usage patterns to reduce resources. CAST AI effectively solves challenges related to Kubernetes cost management, scalability, and operations across multiple cloud provider clusters. The cost management is fantastic, as CAST AI chooses the best instance types based on workloads and automatically provisions the right size of nodes, CPUs, and memory, which saves costs. The initial setup was quite good, and it helped us learn more about cost optimization. Overall, I would definitely recommend CAST AI.

What do you dislike about the product?

I would say CAST AI can improve by automatically picking the pod usage and changing the resources without any downtime of production workloads. Basically, right sizing is shown in the dashboard but we have to do it manually. It would be better if it could solve this automatically to spin up new pods and reduce the workloads.

What problems is the product solving and how is that benefiting you?

I use CAST AI to optimize Kubernetes infrastructure costs and resource utilization. It automates node scaling, workload right-sizing, and selects cost-effective instance types, helping us manage multiple cloud provider clusters efficiently.

What do you like best about the product?

What I value most is the granular, real-time visibility and the "app-aware" engine that scales resources based on actual workload DNA rather than just generic metrics. The seamless integration with our existing CI/CD pipelines meant we saw performance improvements and massive cost reductions within hours of deployment. It has effectively bridged the gap between our DevOps and FinOps goals through one unified, automated control plane

What do you dislike about the product?

While the automation is powerful, the "black box" nature of the decision logic can initially make it difficult to trust the system with mission-critical production workloads without extensive testing of the guardrails. We also found that the coordination between the Workload Autoscaler and Node Autoscaler could be tighter, as they sometimes operate independently rather than planning for future node utilization in perfect tandem. Additionally, the "percentage of savings" pricing model can feel like a "savings tax" as you scale, making it harder to predict long-term tool costs compared to a flat-tier subscription

What problems is the product solving and how is that benefiting you?

Cast AI solves the persistent "Kubernetes waste" problem by automating rightsizing, bin-packing, and spot instance orchestration that are traditionally too complex to manage manually at scale. For me, this has replaced hours of tedious YAML tuning and "firefighting" during traffic spikes with a reliable, autonomous engine that keeps our clusters lean and high-performing. The biggest benefit is the reclaimed time; I can finally focus on high-impact architectural work instead of constantly babysitting node groups and cloud bills.

What do you like best about the product?

The UI and dashboards are great to go through. It’s also really helpful to add cast.ai to a cluster so you can see the potential savings before implementing it.

What do you dislike about the product?

Upgrading to newer versions of Kubernetes is difficult, and the Terraform modules feel lacking in support and features.

What problems is the product solving and how is that benefiting you?

The main reason I started using cast.ai was for cost savings. The UI highlights a lot of potential savings, but in actual use I’ve seen resources scale up, and the savings end up being a lot less than what the UI suggests.

What do you like best about the product?

The automation is genuinely impressive - once Cast AI is connected to our clusters, it handles the scaling decisions that used to eat up hours of our engineers' time each week. The cost savings kicked in pretty quickly after setup, and the visibility into where our cloud spend is going has been really useful. We had a multi-cluster setup and Cast AI handled it better than I expected. The recommendations are solid and the UI makes it easy to see what's happening without digging through logs.

What do you dislike about the product?

The initial setup and onboarding documentation could be a bit clearer - there were a few gotchas around IAM permissions that took us longer to figure out than it should have. The alerting options feel a bit limited compared to what we're used to with other tools. Nothing that's been a dealbreaker, but there's room to improve on those fronts.

What problems is the product solving and how is that benefiting you?

We were over-provisioning across our Kubernetes clusters and had no real visibility into where the waste was coming from. Cast AI helped us right-size workloads automatically and brought down our cloud bill noticeably within the first month. The auto-scaling also means our team isn't getting paged for manual interventions nearly as often, which has been a big quality-of-life improvement for the on-call engineers.

What do you like best about the product?

The automated rebalancing and Spot Instance management are game-changers. Unlike other FinOps tools that just give you a list of suggestions to fix manually, CAST AI actually executes the changes in real-time. The Autoscaler is incredibly aggressive (in a good way) at bin-packing pods, which allowed us to shrink our cluster footprint significantly without any downtime. Also, their Spot fallback mechanism gives us the confidence to run production workloads on Spot instances because we know it will move them to On-Demand instantly if capacity drops.

What do you dislike about the product?

While the onboarding is fast, there is a slight learning curve when it comes to fine-tuning policies for very complex stateful workloads. I also noticed that the Workload and Node autoscalers sometimes feel like they are operating on two different planes—it would be great to see even tighter coordination between the two so that resource requests and node provisioning are perfectly synced 100% of the time. Lastly, the pricing can feel a bit steep for very small, static clusters where there isn't much to optimize.

What problems is the product solving and how is that benefiting you?

We were facing massive cloud waste (roughly 40%) due to over-provisioning and 'shadow' Kubernetes spending. CAST AI solved this by automating our rightsizing.

Benefit 1: We reduced our AWS/GCP bill by nearly 50% within the first two months.

Benefit 2: Our DevOps team no longer spends hours every week 'hand-tuning' instance types or manually handling Spot interruptions. It has effectively shifted our team from 'infrastructure babysitting' to actual feature development.