All case studies

Finance · A leading financial-services firm

Finance: an analyst copilot the whole firm could afford to use

The situation

The firm runs AI on the work analysts do all day: summarizing filings and research, drafting client-ready notes, and answering questions across long internal documents. The value was obvious from the first pilot — and so was the problem. Every seat wanted the copilot, and every additional seat multiplied a premium-model bill.

Cost had quietly become the rollout gate. Usage was rationed to a fraction of the analysts who wanted it, which meant most of the firm got none of the benefit.

What moved to Run BiOS

On Run BiOS, the workload split two ways. Bulk summarization and question-answering go through BiOS Adaptive, which routes each request to the lowest-cost model that handles it well — the long-tail of everyday analyst work moved off the premium model entirely.

For drafting, the firm fine-tuned a model on its own research style and terminology, running on dedicated GPUs and billed per second. The resulting weights belong to the firm — a detail its compliance team cared about as much as its engineers did.

BiOS AdaptiveManaged fine-tuningDedicated GPUsCustomer-owned weights

The outcome

  • AI spend came down enough to turn the copilot from a rationed pilot into a firm-wide tool
  • Drafting quality improved noticeably — the fine-tuned model writes in the firm’s own voice
  • Weights are owned outright, so the model investment compounds instead of renting it

Before

Copilot rationed to a pilot group on a premium model

With Run BiOS

Firm-wide rollout, routed and fine-tuned

Illustrative, not measured.

Facing the same bill?

Estimate your workload on the rate card, or talk to us about the deployment pattern that fits.

More case studies