---
title: Whitepaper — Workflow Intelligence as a Service (WIaaS)
description: The Cromus WIaaS whitepaper. A technical and strategic treatment of the Workflow Intelligence as a Service category, the Croms™ metric, the Open Spec Stack, and the pre-execution governance layer for AI workflows.
canonical: https://cromus.ai/whitepaper
source_html: https://cromus.ai/whitepaper
---

# Whitepaper: Workflow Intelligence as a Service

> A technical and strategic treatment of the WIaaS category.

---

## Abstract

AI workflows are being built and operated without a governance layer that precedes execution. Orchestration frameworks (LangGraph, CrewAI, AutoGen) run workflows. Observability platforms (LangSmith, Helicone, Langfuse) measure them after the fact. Runtime agent stacks (Anthropic Claude Managed Agents, Google Agent Platform, AWS Bedrock Agents) govern and execute deployed agents.

None of these tools answer the question that matters most before spending: *should this workflow run, in its current shape, on its proposed model, and at what cost?*

This whitepaper defines **Workflow Intelligence as a Service (WIaaS)** — the category that answers this question deterministically, before execution, with zero LLM calls.

---

## 1. The problem: governance after the fact

AI workflow cost, quality, and failure are observable only after execution. Teams that discover cost spikes, model mismatches, or structural inefficiencies post-deployment face re-architecture under live conditions.

The industry has built excellent tools for observing AI workflow behavior. It has not built a tool for governing AI workflow design.

---

## 2. The WIaaS category

WIaaS operates in the gap between workflow design and workflow execution. It provides:

- **Structural scoring** — is the workflow well-formed and AI-ready?
- **Cost simulation** — what will this cost, across model options, before any token is spent?
- **Waste quantification** — how much of the projected cost is preventable?
- **Governance validation** — does the workflow comply with declared behavioral policies?
- **Portable compilation** — can this workflow be exported to any execution framework?

---

## 3. The Open Spec Stack

Cromus defines and maintains three open specifications:

**SKILL.md** — A portable capability definition file. Captures execution profiles, model requirements, token estimates, IO contracts, and governance constraints. Validated at `/validator/skill`.

**ETHOS.md** — A behavioral governance file. Defines ethical constraints, escalation policies, fallback behaviors, and transparency requirements for AI agent behavior. Validated at `/validator/ethos`.

**MEMORY.md** — A portable memory specification. Defines what an agent retains across sessions, how it compresses context when limits are reached, and how memory transfers between agents. Validated at `/validator/memory`.

All three specifications are open and free to adopt. Cromus validates them for any team, regardless of whether they use the Cromus platform.

---

## 4. The Croms™ metric

A Crom is one unit of preventable AI workflow waste. Croms are computed deterministically across four dimensions:

- **Cost waste** — tokens spent on tasks a cheaper model handles equally well
- **Latency waste** — time lost to serialized steps that could run in parallel
- **Failure risk** — expected cost of retries and error correction from flaky prompts
- **Structural inefficiency** — tokens spent on redundant context, dead branches, and duplicated prompts

Croms give teams a single, actionable number. The same workflow always produces the same Croms score — no randomness, no model-dependent variance.

---

## 5. The 4-step measurement framework

| Step | Produces |
|------|---------|
| 0 — Workflow Classification | Model tier map for every AI touchpoint |
| 1 — Baseline Cost per Workflow | Fully-loaded per-run cost |
| 2 — Croms Analysis | Preventable waste score and optimization actions |
| 3 — TCWO | True per-outcome cost over the workflow's operational lifetime |

---

## 6. Conclusion

WIaaS is not an alternative to orchestration, observability, or runtime governance. It is the upstream layer that produces the governed, costed, validated inputs those layers execute and observe.

Organizations that deploy AI workflows without a WIaaS layer are making architecture and cost decisions blindly. Cromus changes that.

---

## Related pages

- [Workflow Intelligence as a Service →](/workflow-intelligence-as-a-service)
- [Pre-Execution Governance →](/pre-execution-governance)
- [Croms™ →](/croms)
- [Open Workflow Ecosystem →](/open-workflow-ecosystem)