If you're running BigQuery at scale, you've probably felt this before: your reserved slots are sitting underutilized, your autoscale bill is climbing, a critical dashboard is missing its SLO, and somewhere an engineer is spending their afternoon manually tweaking reservation configurations yet again.
It's a frustrating pattern, and it's more common than most teams want to admit.
The Slot Management Problem
BigQuery's pricing model is built around slots — units of compute that determine how fast your queries run. You can commit to a baseline number of slots upfront (cheaper, but fixed), or let BigQuery autoscale on demand (flexible, but expensive). In theory, you balance the two to get performance at a reasonable cost.
In practice, it rarely works out that cleanly.
Most organizations end up over-provisioning baseline slots they don't fully use, while simultaneously overspending on autoscale to cover gaps they shouldn't have. High-priority workloads compete with low-priority batch jobs for the same pool of slots. Engineers set reservations based on gut feel and historical snapshots, then react when something breaks.
The root of the problem isn't visibility — Google has made real strides there. The deeper issue is organizational: FinOps teams and engineering teams are often working toward the same goal from different angles, without a shared framework for making and enforcing slot allocation decisions. At scale, with dozens of jobs, multiple projects, and mixed real-time and batch workloads, the complexity quickly outpaces what any team can manage by hand.
How Sedai Approaches the Problem
Today we're launching autonomous BigQuery optimization. The key word here is autonomous.
Sedai connects to your GCP project and continuously analyzes slot consumption across all of your workloads. It learns how your jobs actually behave: when they run, how many slots they need, and how they interact with each other. Then it acts, making ongoing adjustments to slot allocation and reservations to keep your most important workloads running smoothly, without unnecessary autoscale spend.
This isn't another dashboard. It's not a report that tells you what to fix and leaves the rest to you. Sedai closes the loop, so that action directly follows from insights.
The new capabilities we’re introducing today include:
Slot usage analysis. Sedai surfaces which workloads are consuming slots efficiently, which are over-allocated, and where autoscale is filling gaps that baseline slots should be covering. You get a clear picture of where your slot investment is going, and where it's being wasted.
Workload classification and separation: Sedai detects whether workloads running in a reservation are batch (e.g. scheduled pipelines, overnight data jobs) or interactive (e.g. dashboards, analyst queries, executive reports). When the two types share a reservation, they compete for the same slots, and time-sensitive queries often lose. Sedai identifies these mixed reservations and recommends separating them where it makes sense, so critical workloads always have the capacity they need.
Reservation right-sizing. By analyzing actual demand patterns by workload and time of day, Sedai recommends reservation adjustments that protect high-priority jobs without over-provisioning. Low-priority batch workloads stop crowding out time-sensitive queries.
Autoscale max optimization: Setting your autoscale ceiling too high isn't just a cost risk — it's an SLO risk. A high autoscale max enables BigQuery to borrow idle slots from neighboring reservations, which looks efficient until those reservations need their capacity back and your workload gets interrupted mid-run. It also means queries can consume far more slots than necessary simply because the capacity is there, running faster but at significantly higher cost. Sedai finds the right autoscale ceiling for each reservation, with enough headroom to handle real spikes, sized tightly enough to protect performance and control spend.
Autonomous execution. Unlike native GCP tools and FinOps platforms that stop at recommendations, Sedai closes the loop, applying changes with clear before/after cost and performance projections, and the ability to validate or roll back any action.
Who This Is For
If you're in platform or data engineering, this means less time hand-tuning reservations and more confidence that your slot allocation is actually aligned with workload priorities.
If you're in FinOps, this means a clearer picture of where your BigQuery spend is going, along with concrete, automated actions that bring it down without requiring constant back-and-forth with engineering.
Both teams get a shared, governed view of slot usage and optimization actions. That alone removes a lot of organizational friction.
Ready to optimize your BigQuery Slots?
Book a Sedai demo to speak with a technical expert.
Get Started
Sedai BigQuery optimization is generally available today. If you're ready to stop managing slots by hand, stop by the Sedai booth at Google Cloud Next, or schedule a demo and we'll show you what it looks like in your environment.