[05] Case study

SmartDecision AI

An AI-powered platform that turns plain-language policy into BRMS-compatible decision tables. Built for security-conscious enterprises. Deployable in air-gapped environments.

smartdecision.io
Industry

AI · Enterprise SaaS

Stack

React · Node.js · LLM · Air-gapped

Year

2024

Scope

New product, from zero

[ 02 ] THE CHALLENGE

BRMS, but without the complexity tax

Existing platforms like Drools and IBM Operational Decision Manager are powerful and impossibly complex. We were asked to build a third option.

▸ The problem

Business rules management is a category dominated by tools that are powerful but punishing. Drools, IBM Operational Decision Manager, and their peers ship serious capability behind interfaces a business owner can't use without weeks of training.

The challenge was to design a platform that lets business owners manage complex rule sets directly, without engineering hand-holding. It had to be powerful enough to handle real production rule sets, secure enough for the regulated industries that need BRMS in the first place (which often meant air-gapped deployment), and intelligent enough to help users instead of asking them to translate policy into tables by hand.

And the AI piece had to actually work. We needed to train the model to generate rules that are accurate and cover all possible scenarios, not just look correct on the happy path.

[ 03 ] WHAT WE BUILT

An AI-powered BRMS that doesn't make you regret it

Plain language in, decision tables out. Security, versioning, and a clean integration API built in.

▸ Feature

AI Rule Generation

Trained an AI to convert plain-language policy statements into BRMS-compatible decision tables. Business owners write English, the system writes rules.

▸ Feature

Decision Table Editor

A clean UI for managing complex rule sets without engineering. Business owners create, edit, and version rules themselves.

▸ Feature

Resolve API

A single API endpoint that returns the decision for any input. Easy to integrate from any web application, no BRMS infrastructure required.

▸ Feature

Air-Gapped Deployments

The platform runs without internet access. Sensitive policy data never leaves the customer's network. Security and compliance from day one.

▸ Feature

Versioning & Audit Trail

Every rule change tracked, every decision logged. Roll back, compare versions, satisfy auditors without extra tooling.

▸ Feature

Scenario Coverage Testing

The AI is trained to generate rules that cover edge cases. Built-in testing surfaces gaps before they reach production.

[ 04 ] OUTCOMES

A new way to manage business rules

Security and simplicity, without the BRMS-platform tax.

BRMS, simplified

Drools and IBM ODM complexity replaced with a UX a business owner can use.

Air-gapped

Sensitive policy data stays on the customer's network. No phone-home, no third-party calls.

Faster integration

One Resolve API call where teams used to wire up a BRMS engine.

[ 05 ] TECH STACK

An AI-powered enterprise stack

Modern web tools fronting an LLM-driven backend that ships into security-conscious environments.

React + TypeScript

The decision table editor where business owners spend their time. Built for clarity in a category known for complexity.

Node.js

Rule processing, the Resolve API, and the AI orchestration layer. Same language end to end.

LLM Integration

AI-powered rule generation. Tuned to produce accurate, scenario-covering rules in the BRMS decision-table format.

PostgreSQL

Rule history, versions, and audit logs. Relational data for a domain that lives and dies by traceability.

Air-Gapped Deployment

Self-hosted infrastructure that runs without internet. Customer-controlled, customer-secured, customer-owned.

Docker + Kubernetes

Containerized for predictable deploys into customer environments, from public cloud to fully on-prem.

[ 06 ] MORE WORK

Other case studies

More products we've designed, built, and shipped.