liveBook · Manning

Automation

As explained in Build High Value Software Systems

The use of software or machines to perform tasks with reduced manual intervention.

Automation uses software to perform repeatable checks and delivery steps with less manual intervention. Its value is faster, trustworthy feedback: a linter catches a typo quickly, tests check a business rule on every commit, and a pipeline moves verified changes toward production. Automating more work is not the goal; preventing defects and waiting is. section 5.6 section 5.6.2

Before automating a check, ask what it protects and what it costs:

  • Will it catch a consequential defect earlier, remove repeated manual setup, and fail only when a meaningful contract breaks? That is a strong candidate. section 5.6.1
  • Will it fail on cosmetic changes, depend on replaceable implementation details, duplicate cheaper protection, or require constant maintenance? It may create more waste than it removes. section 5.6.1
  • Will it lengthen the path to feedback for every small change? Keep necessary deep validation, but consider layered testing, parallelization, or post-deploy validation rather than blocking every commit. section 5.6.2 section 6.5.2

Protect the payment rule, not the gateway choice

A checkout service must reject zero or negative payment amounts and must not reserve inventory when a card is declined. Fast, deterministic tests of those outcomes are worth running on every commit: they can catch a misplaced validation before it reaches staging. An assertion that the service uses a particular gateway client protects an implementation choice instead. Changing providers could break that assertion while leaving the required behavior intact. section 5.6.1

Listing 5.7 Justified automation of validation rules
def test_payment_rejects_invalid_amounts():
    result = process_payment(amount=-10, card="4111111111111111")
    assert result.status == "rejected"
    assert result.error == "Invalid payment amount"
 
def test_declined_payment_does_not_reserve_inventory():
    mock_gateway.configure_to_decline()
    result = process_payment(amount=100, card="4111111111111111")
 
    assert result.status == "failed"
    assert inventory.reserved_count() == 0

The two short tests show the boundary worth automating: observable payment and inventory outcomes. section 5.6.1

Treat CI/CD as an end-to-end flow, not a collection of individually useful checks. In one case, added integration tests and scans stretched a build from 12 to 42 minutes. Commits queued, developers switched tasks, and feedback slowed. Shorter coding time could not fix a validation queue; the pipeline had become the constraint. section 6.5.2 section 6.5

For a recurring manual step, start with a checklist that names the check and how to perform it. Mark candidates for automation, then move stable checks into CI and remove the corresponding manual items. Measure and share the result: even a tested deploy script with documented time savings can go unused if teammates do not adopt it. Keep the check that protects an outcome; reconsider one that merely adds activity. section 14.2 section 15.0

How Build High Value Software Systems covers it

How deep it goes
Works through contract-versus-implementation tests and pipeline bottlenecks, but stops short of a full automation implementation.
Examples
Contrasts payment-rule tests with a gateway-client assertion and a build that grows from 12 to 42 minutes.
What it assumes
Assumes familiarity with commits, tests, and deployments while spelling out how to judge an automated check’s value and cost.
What you can do after
Practice/perform it by selecting checks that protect behavior and moving repeatable checklist steps into CI.
How it connects
Links test automation to Lean waste and CI/CD value streams.

Also explained in

Compare how 23 books explain automation →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Build High Value Software Systems. The section links point to the parts of the book it draws on.