Every intelligent system needs a core brain. Especially in data and analytics automation, that brain is the Business Rules Engine — the layer that encodes how calculations, validations, and transformations should happen.
Without it, no automation truly works: not regulatory reporting, not risk models, not financial close, not even basic revenue analytics.
The crumbling monolith
Traditional rules engines are collapsing under AI-era workloads. They couldn't adapt to changing times.
- Rising infra costs from SQL-heavy databases
- Broken orchestration with fragmented tools
- Opaque workflows, poor lineage, zero transparency
- Choked pipelines with large, granular datasets
- Steep learning curves for Excel-native business teams
- Heavy vendor lock-in for every small change
Instead of an unstable monolith, imagine a federation of agents — each mastering one task, handing over seamlessly, and together solving challenges that traditional systems simply can't. This is our innovation: the six Agents of Orchestration, choosing cooperation over conquest and specialization over domination.
Innovation on business rules — the six agents
Agent 1: The Conductor (Master Orchestrator)
- Centralized orchestration brain
- Function library — Excel-like, but in Python
- Proprietary formula parser: Excel functions → Python scripts
- Sequencing engine (order of execution) ensures every agent plays their part at precisely the right moment
The Conductor translates business needs into coordinated action across all other agents.
Agent 2: The Gatekeeper
- Receives instructions from the Conductor
- Builds on-the-fly SQL in DuckDB
- Processes only localized in-memory subsets
- Prevents choking of pipelines
The agent allocates exactly what's needed, when it's needed — preventing resource wastage.
Agent 3: The Specialist
- Executes business rules based on the sequencing engine, on Pandas
- Accelerates file operations (read & write) with Rust
- Delivers precision and speed
This agent embodies efficient processing: careful planning combined with rapid, flawless execution.
Agent 4: The Connector
- Streams data between the Gatekeeper and the Specialist using PyArrow
- Reduces memory footprint
- Maintains seamless, high-performance communication
In the monolith, systems hoarded information and memory. This agent ensures seamless communication — in-memory compression without information loss.
Agent 5: The Archivist
- Stores all business data in Parquet
- Builds a cost-efficient data lake
- Immutable, versioned, compressed
Agent 6: The Chronicle
- Captures complete execution metadata in Postgres
- Full lineage + audit trail
- Transparency and trust by design
The greatest weakness of traditional rules engines was opacity — no information of who made what decisions or why. Our greatest strength is transparency: every action tracked, every decision auditable.
The results
- Familiarity = Adoption: Excel-like rules keep business teams in control.
- Separation of Compute & Storage: 80% lower storage costs.
- Optimized Memory Flow: 80% less memory footprint, even with text-heavy datasets.
- Relevance-First Execution: 2× faster processing — focused efficiency vs. trying to do everything in one shot.
This isn't just technical elegance on business rules — it's business resilience at scale. A new way for enterprises to automate without vendor lock-in, runaway infra costs, or breaking the trust that business teams place in Excel-like rules.
