What is a process engine, and when do you want one

Motto — You want a process engine for one problem: long-running, stateful, auditable coordination of humans, systems, and time. Wait until you actually have that problem.

Part of Phase 00 — Orientation & setup. Concept lesson — no code required. Concept reading: Principle 10.

The Problem

People usually ask “should we use a workflow engine?” backwards — after someone saw a demo, not after diagnosing the pain. The pain that actually justifies one looks like this: a business flow spans days or months. A status column tracks state, and five services mutate it. Cron jobs enforce deadlines and drift over time. A SQL archaeologist answers “where is case X stuck?” And an auditor asks which rules version decided a case. If you have most of that list, an engine converts it into a diagram, rows, timers, and history. If you don’t, an engine converts a simple system into a distributed one with extra ceremony.

The Concept

A process engine makes three commitments in one runtime. Here is the whole course in one diagram:

The diagnostic, as a table — count your yes-answers:

Question	Engine says
Does the flow wait — for people, documents, deadlines — for days+?	wait states are its core trick
Do humans and systems interleave (review → API call → approval)?	task + service orchestration
Do deadlines/SLAs/expiries drive behaviour?	timers, versioned with the flow
Will audit/compliance ask “what happened and under which rules”?	history + definition pinning
Does the business change the flow more often than you deploy?	model + DMN redeploys
≥ 4 yes	strong engine case
≤ 2 yes	see the alternatives below

Most flows shouldn’t run on an engine. Here are the honest alternatives:

A status column + a queue — for short, fully automated, rarely-changing flows. Three states and one retry policy don’t need BPMN.
A saga/durable-execution runtime (Temporal-style) — code-first orchestration for engineers. No diagram, no business-facing model (lesson 04 compares the two properly).
Event choreography — services react to each other’s events, with no central coordinator. You get maximal autonomy, but “where is case X?” has no single answer. That question is the exact reason an engine exists.

Ship It

This lesson ships outputs/engine-fit-checklist.md — the diagnostic plus the alternatives table. Bring it to the next “should we use a workflow engine?” meeting.

Check Yourself

Q1. Which flow is the weakest engine candidate?

A) loan origination: humans, bureaus, offers expiring in 30 days
B) image thumbnailing: 3 automated steps, seconds long, never changes
C) vendor onboarding: documents, approvals, compliance audit trail
D) claims handling: adjusters, deadlines, regulators

<details><summary>Answer</summary>B — no waits, no humans, no time, no audit pressure. A queue and a worker do the job with less machinery.</details>

Q2. The strongest single signal for an engine is…

A) many microservices
B) flows that wait on humans/time for days while requiring a queryable, auditable position (“where is case X, under which rules?”)
C) high throughput
D) a team that knows Java

<details><summary>Answer</summary>B — durable waiting plus accountability is the combination nothing else provides as cheaply.</details>

Q3. Event choreography loses to orchestration precisely when…

A) throughput is high
B) someone must answer “where is this case and what happens next?” — choreography has no single place that knows
C) services are polyglot
D) events are JSON

<details><summary>Answer</summary>B — central state is orchestration’s cost and its product. Buy it when that question matters.</details>

Challenge. Run the diagnostic on three real flows in your organisation. For the highest scorer, write a one-paragraph pitch against using an engine. If you can’t make a strong case for the alternative, you haven’t finished the analysis.

Next: The Flowable platform map
The full comparison: lesson 04
Other tracks: Prioritization & roadmaps · Creativity: strategy & execution — the build-vs-buy judgment behind adopting an engine at all.

What is a process engine, and when do you want one

The Problem

The Concept

Ship It

Check Yourself

Related