DecisionManager

Enterprise BRMS for regulated decisions

Credit, claims, and eligibility — decided the same way every time.

The decision automation platform banks, insurers, and healthcare teams use to author, govern, and execute policy — underwriting, claims, formulary and eligibility — without redeploying code. Every decision is bounded by a deterministic step budget, so the same input reaches the same verdict under load and when a regulator replays it years later.

  • Deterministic step budget
  • Edge .dmb bundles
  • Decision warehouse traces
  • AI authoring · human review
  • ODM migration assessment
  • Seeded trial workspace

No credit card · Does not auto-convert · Seeded workspace ready in minutes

Or run the live demo — no signup required

  • No credit card to start
  • Does not auto-convert to a paid plan
  • Book an executive demo via contact
  • EU data residency on hosted plans
hash-verified decision bundle
947 B
cold start to first decision
~20 ms
WebAssembly engine build
1.5 MB
2 MB request, 500 candidates
~3.4 ms

Measured end to end on the edge binary against a definition taken from a running database — with Postgres stopped. Throughput is a full execute including parsing the request. Cold start from a debug build, so a release build is the floor. Not customer-reported averages.

Migration proof

Two real ODM rulebases, imported and executing — including what did not translate.

These are not estimates. Each number below is asserted by a test that runs in CI against a ODM export, and the two caveats are reported by the translator itself, not discovered later in production.

BRL rules
26 / 26
Sample refund rulebase — all translated
Ruleflow leaves
12
One node, structure preserved
Reported caveats
2
Flagged before commit, not after
Import to first decision
minutes
Upload, analyse, commit, execute
What translated from the two ODM sample exports, and what did not.
ODM constructResultDetail
.brl action rulesMappedAll 26 rules of the Sample refund rulebase translate to platform rules with conditions and actions intact.
RuleflowMappedTask structure and ordering are preserved; the largest node keeps all 12 leaves.
Decision tablesMappedRows, conditions and actions carry across; row identity is retained so a gap can name the row that moved.
BOM / B2XMappedObject model and its virtual-to-execution bindings are read and used to resolve terms.
Templated messagesCaveatA message body built with an ODM function such as Max(850000, 'High Balance Limit') is imported as a template placeholder and flagged for a human to confirm. Three such caveats in the eligibility rulebase.
Unresolvable vocabularyCaveatA term with no match in the exported BOM is reported as "not resolvable against the vocabulary" rather than guessed. A silently wrong output field in a regulated rulebase is worse than a reported gap.
Two real ODM rulebases, imported and executing — including what did not translate.
Illustrative roadmap

A realistic migration spine — not a contractual SLA

Days are indicative for a scoped pilot ruleset with human review. Multi-service estates take longer; we will not invent a calendar for yours.

Migration roadmap from inventory to first production decision
  1. 1Day 0

    Private inventory

    Browser scan of your export — size the estate without uploading.

  2. 2Days 1–3

    Analyse & map

    Trial workspace: translator + gap report; caveats named before commit.

  3. 3Days 4–10

    Human review

    Policy owners confirm templates, vocabulary, and exception paths.

  4. 4Days 11+

    Publish & execute

    Governed lifecycle, deploy to an environment, warehouse evidence.

Source: backend/crates/odm — sample_refund tests and eligibility fixture tests. The counts above are the assertions those tests make.

Assess your own exportRead the migration APIThe archive is parsed in your browser. Nothing is uploaded.

Frequently Asked Questions

Everything You Need to Know

Got questions about migration, compliance, performance, or deployment? We've got answers.

How does DecisionManager replace IBM ODM (Operational Decision Manager)?
DecisionManager imports legacy IBM ODM RuleApps, BRL action rules, BOM/XOM verbalizations, decision tables, and ruleflows into native BAL 1.0 with AI-assisted conversion and required human review before publish. Parallel execution replay helps validate behaviour. The free browser /migrate tool inventories an export only — it does not upload rule content, and we do not claim one-click or 100% automated migration.
What is the decision execution latency compared to Java rule engines?
The core engine is compiled Rust with multi-key indexed paths and no GC pauses. The figure we publish is measured end-to-end: about 3.4 ms scoring a 2 MB payload with 500 candidates (including parse). Cold start to first decision is about 20 ms on the edge path; the hash-verified decision bundle is 947 bytes. We do not publish a freestanding “sub-2 ms average” without method.
Where can DecisionManager be hosted or deployed?
Three product shapes, same images and features: hosted plans run in EU regions (Frankfurt by default) — we do not operate a multi-region hosted SaaS mesh listing us-east-1 or similar as our regions. Business and Custom run the same containers in your own AWS, Azure, or GCP account in any region you choose (no provider-specific code). Custom also supports on-premise or air-gapped Kubernetes. A hash-verified edge bundle can execute with no database and no outbound network; when you self-host, the wasm/native engine can also run on Lambda, Cloudflare Workers, or Cloud Run under your account — portable edge execution, not a DecisionManager multi-region network.
Does DecisionManager support compliance standards like SOC2, HIPAA, and GDPR?
We hold no SOC 2, ISO 27001, HIPAA, or equivalent certification today, and we will not imply one. What the platform does provide: an audit trail, least-privilege RBAC, draft → review → approved → published governance, workspace isolation, EU residency on hosted plans (Frankfurt default), and an export/erasure-friendly data model with a DPA on Starter and above. Operational practices are SOC 2-aligned; alignment is not certification. Formal certification is on the roadmap.
How does AI-Assisted Rule Authoring work?
Multi-provider AI authoring (Anthropic, OpenAI, Gemini, Bedrock-compatible, and self-hosted endpoints) drafts rules, tests, and rule applications from a natural-language prompt, with ordered provider fallback and a content-hash cache. Output is always reviewed by a person and never auto-persisted as the sole authority.
What currencies and billing options are available for enterprise customers?
DecisionManager offers Free trial, Starter, Business and Custom plans. Commercial terms are quote-based (no public list prices). Hosted self-serve evaluation is as-is unless your agreement includes a written SLA; Custom includes an SLA when contracted.
Run the live demo