CASE STUDY 14 / 16TESTING & QATEAM LEADERSHIPENGINEERING LEADERSHIP

Specification & QA Governance.

An independent source of truth for the platform behind BMI Global Ed's 60 to 80 international education fairs a year: a repository of specifications and tests, kept apart from the application code, that connects what the platform must do to proof that it does it.

MY PART
Directed the QA team and instituted its spec-driven, test-driven way of working.
PROBLEM
External vendor engineers and AI coding agents kept drifting from what the platform was meant to do, and each dispute over what was right ran to several email threads.
SOLUTION
Written specifications for all seven parts of the platform, 96 in all, each checked automatically on every change, with the exhibitor CRM's matched one-for-one to its tests.
RESULT
Vendor disputes that once took several email threads now take none: the specification and its matching tests settle them.
FIG.01 · SPEC TO TEST, ILLUSTRATIVE
Several → 0Email threads per vendor dispute: the specification and its tests settle them now
96Specifications under governance
7Parts of the platform, each with its own specifications
39/39/39/39Specifications, test cases, integration tests and unit tests matched one-for-one for the exhibitor CRM
01WHAT IT DOES

A source of truth that outlives any one repo

External vendor engineers and AI coding agents both change the platform, and every change risks drifting from what it is meant to do. The platform takes student registrations and exhibitor bookings and payments for every fair, so that drift reaches real students and universities. This repository holds specifications and tests but no application code. It is the independent record that settles the question, and it outlives any single application repository, vendor or engineer's tenure.

Specification-to-test traceability

Every specification, test case, integration test and unit test matched one-for-one for the exhibitor CRM, checked automatically on every change.

Cross-repository sync

Specification and test content kept consistent automatically across the repositories it governs, whenever any one of them changes.

Guarded AI authoring

An AI agent runs a structured interview on each new requirement, and a separate challenge step questions it, before the specification, test cases and automated tests are drafted. Hard guardrails keep it away from application source code.

Drift triage

A documented, recurring process that catches specification/code mismatches and converts each one into a prioritised action item for backend, frontend or QA.

A hierarchy of truth

The specifications are the reference everyone works from. When code and specification disagree, a documented rule applies: the code wins, and the specification is corrected so the reference stays true.

Governed issue intake

No engineering work accepted without an explicit specification reference, a named owner, and linked test-coverage evidence.

02HOW IT WORKS

From requirement to verified proof

The repository runs no service of its own. Its structure is the contract: every specification (SPEC) carries matching identifiers for its test case (TC), integration test (IT) and unit test (UT). That structure is checked automatically on every change and used by both the QA team and the AI agents that draft new specifications.

FIG.02 · SPEC-TO-TEST FLOW, SIMPLIFIED
03MY ROLE

What I directed

I directed BMI's QA team and instituted the spec-driven, test-driven practice this repository supports, entirely through verbal direction and written specifications rather than through code.

  • Set the direction for a QA/specification-governance system covering 96 specifications across seven directories that mirror the platform's seven modules one-for-one, with a complete, CI-enforced traceability chain, specification to test case to integration test to unit test, verified matched one-for-one for the exhibitor CRM.
  • Set the direction for a bidirectional documentation-sync architecture keeping specification and test content consistent across the repositories it governs, triggered automatically whenever any of them changes.
  • Directed the QA team's work through verbal direction and written specifications, and instituted the spec-driven, test-driven engineering practice these specifications support: developers feed them to AI coding agents to drive implementation.
  • Instituted hard guardrails preventing AI coding agents from touching application source code, within a governed authoring pipeline that generates specifications, test cases and automated tests from a structured requirement interview.
  • Established a spec-vs-implementation triage process that catches and resolves drift on a documented cadence, converting each mismatch into a prioritised action item.
  • Improved the experience of the vendor engineering team building on the platform by directing specification governance with a CI-checked chain from specification to test, so vendor disputes that once took multiple email threads now take none.
04TECH STACK

Tech stack

[a] LANGUAGE & RUNTIME
JavaScript/TypeScriptNode.js
[b] TESTING
JestSupertestPlaywright
[c] GOVERNANCE & CI
GitHub ActionsHasura GraphQL
[d] AI AUTHORING
GitHub Copilot custom agents

Say hello.

LINKEDINlinkedin.com/in/danielgorgonia

Hidden from bots until you click.

LOCATIONLondon, United Kingdom
LOCAL TIME--:--