# Miji

> Miji is an enterprise decision automation platform. It turns business policy into governed workflows: versioned rules and decision tables, exact decimal calculations, AI steps where judgment helps, human approval where it matters, and a hash-chained audit trail that records which rule version made each decision. You describe an operation and Miji's AI drafts a runnable workflow for you to review. It is live today in lending, sold first to manufactured-home lenders. Miji is a product of Canopy Mind, Inc.

## Platform

- [Decision automation](https://mijios.ai/decision-automation): what Miji is, the problems it solves, how it automates a decision, what you can use today, and how it compares with rules engines, workflow tools, AI agents and enterprise suites.
- [Business rules and workflow automation](https://mijios.ai/business-rules-workflow-automation): turning a policy into a running workflow — rules as conditions or decision tables, field-to-field conditions, effective-dated lookup grids, exact calculations, approval gates, test and production environments.
- [Enterprise AI workflows](https://mijios.ai/enterprise-ai-workflows): where AI runs in Miji (drafting workflows; bounded judgment steps with your own model and key), why the policy decision stays on rules and people, and the governance controls.
- [Policy testing](https://mijios.ai/policy-testing): testing a rule or policy change before it decides anything — replay one recorded case on a draft and compare it with the original, batch-replay thousands of past cases and see every decision that would change, shadow a challenger beside live traffic without it acting, scenario sets of cases with expected outcomes, and A/B tests on a capped share of live traffic with guardrails and a second approver. Replays answer every external call from the original run's record.
- [Case lifecycle management](https://mijios.ai/case-lifecycle-management): lifecycles of stages above workflows, gates on the arrows between stages that hold a case until its conditions (requirements) are cleared or waived, versioned condition templates, evidence intake from portal, email, text, staff and connectors, checks, letters made from a gate's list, and audited waivers and exceptions. In lending: conditions at prior-to-docs (PTD) and prior-to-funding (PTF).
- [Governance and audit trail](https://mijios.ai/governance): the hash-chained audit trail with on-demand verification, the rule version recorded with every decision, human approval gates, role-based access, OIDC single sign-on, test and production environments, AI steps with your own model and key, and managed or self-hosted deployment.
- [Use cases](https://mijios.ai/use-cases): operations outside lending that the same engine builds (claims triage, prior authorisation, onboarding, compliance review, invoice exceptions).
- [What's new](https://mijios.ai/whats-new): dated release notes, newest first. Each entry links to the section of Policy testing or Case lifecycle management that explains it: replay, batch replay and scenario sets (30 September 2026), shadow runs (1 October 2026), A/B tests with guardrails (4 October 2026), gates and condition templates (9 October 2026), and working the conditions on a case (10 October 2026).

## Lending

- [Lending decision automation](https://mijios.ai/solutions/lending): eligibility, overlays, pricing grids, exact DTI/LTV/payment calculations, Equifax credit pulls, underwriter routing and the decision record. Runs alongside a lender's LOS and AUS (DU, LPA); does not replace them.
- [Miji for Lenders](https://mijios.ai/lenders): prequalification for manufactured-home lenders — chattel and land-home overlays, retailer applications, home-dependent limits. Mortgage and consumer lenders run on the same engine.
- [Lender blog](https://mijios.ai/lenders/blog): decision records, verification policy, exception queues, and the Fannie Mae / Freddie Mac AI governance requirements.

## Comparisons

Facts about other products are taken from their own public pages, cited on each page.

- [Compare decision automation platforms](https://mijios.ai/compare)
- [Miji vs FICO Platform](https://mijios.ai/compare/fico-platform)
- [Miji vs Equifax Decision Hub](https://mijios.ai/compare/equifax-decision-hub)
- [Miji vs Drools](https://mijios.ai/compare/drools)
- [Miji vs Pega](https://mijios.ai/compare/pega)

## Learn

- [What is decision automation?](https://mijios.ai/learn/what-is-decision-automation)
- [Decision automation vs business rules engines](https://mijios.ai/learn/decision-automation-vs-business-rules-engines)
- [Decision automation vs workflow automation](https://mijios.ai/learn/decision-automation-vs-workflow-automation)
- [How to automate policy-based decisions](https://mijios.ai/learn/how-to-automate-policy-based-decisions)
- [How AI changes enterprise decision automation](https://mijios.ai/learn/ai-in-enterprise-decision-automation)
- [Blog](https://mijios.ai/blog): notes on governed operations.
  - [The difference between an AI agent and an AI worker](https://mijios.ai/blog/ai-agent-vs-ai-worker)
  - [What a work trail actually needs to contain](https://mijios.ai/blog/work-trail)
  - [Human-in-the-loop is a feature, not a limitation](https://mijios.ai/blog/human-in-the-loop)
  - [Deterministic workflows and AI reasoning: when to use each](https://mijios.ai/blog/deterministic-vs-ai)
- [Roadmap](https://mijios.ai/roadmap): what is shipped, what is being built, and what is next.
- [Vision](https://mijios.ai/vision): where Miji is going (goal-first operation with AI managers). This is the direction, not what ships today.

## What is and is not built

Shipped: AI-assisted workflow authoring; workflow engine (conditions, loops, batches, parallel branches, sub-workflows, files); AI steps with your own model and key; versioned, fingerprinted rules and decision tables with field and expression conditions and a rule impact check; effective-dated lookup grids; exact decimal arithmetic with finance functions; approval gates (any one, all, majority) and work queues; about 30 connector templates, custom HTTP endpoints, workflows as web services and an MCP server; test and production environments with pinned deployments; hash-chained audit trail with on-demand verification; the Lab (replay, batch replay, shadow runs, scenario sets, A/B tests with guardrails); lifecycles with gates, condition templates, evidence intake, checks and letters; managed or self-hosted on PostgreSQL in GCP, AWS or Azure; OIDC single sign-on (Okta, Entra ID, Keycloak) and role-based access. Lending: prequalification, verification policy, Equifax OneView credit, document custody.

Being built: draft-approve-publish for policy changes, decision notices, compliance pack (AI usage register, examiner export), SQL Server for self-hosting.

Next: pre-built manufactured-home package (configured per lender today), servicing, fair-lending exports, goal-first AI managers.

Not offered: predictive model or scorecard building; DMN import or export; replacement of a loan origination system or automated underwriting system.

## Contact

Book a demo at https://mijios.ai/demo or email hello@mijios.ai. Under the hero of https://mijios.ai and https://mijios.ai/lenders, "Tell us your operation" drafts how Miji would run an operation you pick or describe; from the draft you can open it in a free workspace, book a demo with it, or have it emailed to you.