The verifier pattern: why we ship two models per workflow
How we got first-pass accuracy from 91% to 98.4% by adding an independent verifier — without doubling cost.
Read →
Journal
Notes from the agentic frontier.
Essays, benchmarks, and field notes from our engineers, designers and forward-deployed team.
We migrated a $180M company off NetSuite in 9 weeks
What it actually took: a Migration Agent, a Saturday cutover, and a single controller in the loop.
Read →Bizos-Bench: a public benchmark for finance agents
Introducing the first reproducible eval suite for agentic back-office workflows. All datasets MIT-licensed.
Read →Designing for agents as users
Forms were built for humans. Our UI now treats agents as first-class operators. Here is what changed.
Read →Why we built our own ledger
Conversational accounting needs a data model legacy ERPs cannot give us. So we wrote one in Postgres + Temporal.
Read →Series B: $80M to fund the agent operating system
We are doubling forward-deployed engineering and expanding to the UK + EU this year.
Read →The CFO first day with Bizos
A walkthrough of onboarding from the perspective of a finance leader at a 280-person manufacturer.
Read →Tool-use under load: 200 calls/sec without breaking idempotency
Our orchestration learnings from running thousands of agent runs daily.
Read →Sandboxing tools without sacrificing latency
How we keep the executor agent fast while keeping every API call safe.
Read →