TechTalk · AI Readiness Assessment

AI Readiness Assessment

Project: ai-readiness-assessment · 2026-06-03
AI Readiness — Habitat Maturity: Level 3 (Regulating)
Next Step / Gap: +1.2 to Level 4 (Orchestrating)
1Dictating
2Commanding
3Regulating← you are here
4Orchestrating→ target
5Supervising
Habitat Maturity (model)Level 3 Cognitive read (Sovereign Engineer)Level 3 — Habitat design Habitat Build Gap +0.2 · Coherent

A strong, persistent habitat — synced context, a running reflection loop, policy-as-code CI, a published plugin — with the specifications layer (L3→L4) as the one missing rung.

This is the plugin assessed against its own repository — a dogfooding example of the report /ai-readiness-assess produces.

Habitat Maturity Profile

All fourteen dimensions of the Agentic Experience 5-Level Habitat Maturity Model, placed L1–L5.

Habitat Maturity Level: L3 (mean L2.8) — held back by L2 Agent input, Agent composition, Testing, and Observability.

Discipline Maturity

Context Engineering
4 / 5
Architectural Constraints
3 / 5
Guardrail Design
3 / 5

AI Readiness Score — five readiness dimensions

A view over the same evidence; the headline level is the 14-dimension Habitat Maturity.

Context
L4
Conventions
L4
Architectural guidance
L3
Guardrails
L3
Agent readiness
L2

Operational Axes (Part D)

The four discipline-aligned headline dimensions (a focused view; the gap uses all fourteen).

Composition
L2 — single + saved patterns
Testing
L2 — Asserting
Observability
L2 — Captured
Governance
L3 — Constitutional

Habitat Build Gap

Habitat Maturity Level (model):  L3  (14-dim mean L2.79)
Cognitive read (Parts A–C):      L3
Habitat Build Gap:               +0.21   (cognitive − 14-dim mean)
Interpretation:                  Coherent

Team and habitat are in step. The single dimension most worth lifting is Agent input (toward specs) — it is also the L3→L4 ceiling for the cognitive read.

Strengths & Gaps

Strengths

  1. The reflection→promotion loop runs end to end (verified Promoted lines into AGENTS.md / HARNESS.md).
  2. Policy-as-code governance: two required CI checks on a protected main + a changelog-driven release pipeline.
  3. Rich, synchronised context (HARNESS / AGENTS / ONBOARDING + Cursor/Copilot/Windsurf) and a 22-page docs site.

Gaps

  1. No specifications layer (no specs/, no spec-first ordering, no adversarial spec review) — the L3→L4 ceiling.
  2. Verification is shallow: structural assertions against committed samples; behavioural/semantic checks are manual.
  3. No agent composition or observability for development (single agent; CI logs only).

Prioritised Improvement Plan

Team develops = practice the team builds; Org provides = enablement the organisation supplies.

  1. Introduce a lightweight specs layer — a one-paragraph spec atop substantive PRs, graduating into a specs/ directory, with an adversarial “what could go wrong” review before implementation. Closes the L3→L4 ceiling.
  2. Deepen verification — golden-output diffs or an LLM-judge harness for the B/C tiers, moving Testing from Asserting to Verifying.
  3. Add a read-only critic step and basic observability — lifting Composition and Observability off L2.

Reading Path

Assessed at L3 — Habitat design. The next read in The Sovereign Engineer is the Level 4 (specifications) chapter — make intent first-class, separating “what” from “how”.

Read on Leanpub

Next Steps

Your habitat is healthy at L3, and the L3→L4 jump is about making specifications first-class. TechTalk can support a specification-first engagement — a specs/ layer wired into your existing HARNESS and CI, an adversarial-review touchpoint at plan approval, and orchestration patterns so agents act on specs rather than ad-hoc instructions.

Book a call with TechTalk
thomas.stangl@techtalk.at