TechTalk · AI Readiness Assessment

AI Readiness Assessment — re-assessment

Project: ai-readiness-assessment · 2026-08-16
AI Readiness — Habitat Maturity: Level 4 (Orchestrating)
Next Step / Gap: +1.4 to Level 5 (Supervising)
1Dictating
2Commanding
3Regulating
4Orchestrating← you are here
5Supervising→ target
Habitat Maturity (model)Level 4 Cognitive readLevel 4 — Specification-led Habitat/Workflow Gap +0.36 · Coherent

Progression. The baseline (view) read a coherent L3 (gap +0.2); the second reading flipped to +1.07 — ambition outpaces enablement when spec-first lifted the cognitive read and the habitat did not follow. It has now followed: the mean rose L2.93 → L3.64 and the gap closed to +0.36 — Coherent. Both reads sit at L4.

A re-assessment of the plugin against its own repository, 74 days after the previous reading and immediately after the v1.0.0 release. Read how it closed: of the three dimensions the last report named as the ceiling, only Testing moved. Agent composition and Observability are still L2 and are still the ceiling — coherence was restored by everything else rising.

Habitat Maturity Profile

Habitat Maturity Level: L4 (mean L3.64) — held back by L2 Agent composition and L2 Observability.

Discipline Maturity

Context Engineering
4 / 5
Architectural Constraints
5 / 5 · was 4
Guardrail Design
4 / 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
L4
Guardrails
L3
Agent readiness
L3 · was L2

Operational Axes (Part D)

Composition
L2 · unmoved
Testing
L3 · was L2
Observability
L2 · unmoved
Governance
L4

Habitat/Workflow Gap

Habitat Maturity Level (model):  L4  (14-dim mean L3.64)
Cognitive read (Parts A–C):      L4
Habitat/Workflow Gap:            +0.36   (cognitive − 14-dim mean)
Interpretation:                  Coherent

The gap has closed from +1.07 to +0.36 — back to coherent, and at a higher level on both reads. But of the three dimensions the last report named as the ceiling, only Testing moved (L2 → L3). Agent composition and Observability are unchanged. Coherence was restored by everything else rising: the behavioural read climbed six levels across five dimensions as practice shifted to supervising whole slices, and Agent input and Output artefact each gained a level from the specs work. A real result — but not the same as fixing what was named.

Strengths & Gaps

Strengths

  1. Specs genuinely drive the work — 11 specs, adversarial review in every one, and spec 0011 refined across six slices with its acceptance ticked as each landed.
  2. The suite grew 54% (72 → 111 assertions), every one written red-first — and caught real defects, including surface drift and a broken link before it deployed.
  3. Multi-repo scope shipped at v1.0.0, with a real portfolio example published exactly as it came out.

Gaps

  1. Agent composition is L2 and has not moved since the baseline — no critic, no orchestrator. Every check here is a script or a human.
  2. Observability is L2 and is the least-evidenced placement in the report. Nothing captures agent activity, tokens or cost; the HARNESS Observability section is four “Not yet configured” placeholders.
  3. The governance loop has stopped closing — the audit is 74 days stale, and the harness drifted twice in its own favour without either document being updated.

Prioritised Improvement Plan

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

The gap has closed, so the constraint is no longer coherence but altitude — the L4→L5 platform layer:

  1. Install a governance audit cadence (org provides) — the cheapest L5 marker. The machinery exists; the Status block is 74 days old.
  2. Add a read-only critic agent (team develops) — the one move that lifts Agent composition off L2, where it has sat across two assessments.
  3. Capture agent activity (team develops) — token, latency and cost per session, and an observability snapshot at a cadence.

Next steps

The gap is now +0.36 — Coherent: the habitat has caught up with the thinking, and both reads sit at L4, so nothing here is out of step with itself. The constraint is no longer coherence but altitude — the L4→L5 platform layer, where three of six markers are present (a published plugin, fitness functions, portfolio artefacts) and three are missing (governance audit cadence, decision archaeology, cost and model-routing discipline). TechTalk can support a platform-engineering engagement on those three, together with the two dimensions that have now been the ceiling across two consecutive assessments — Agent composition and Observability — so the next level is reached by lifting what was named rather than by everything else rising around it.

Your assessment shows where your repository is today. A review helps you understand why these findings matter, which improvements will have the biggest impact, and what should come next.

Not sure what these results mean?

Your assessment tells you where you are.
A review helps you understand why you received these scores and what to do next.

Review your results with Russ Miles

Creator of the AI Readiness Assessment | Author of The Sovereign Engineer

For one hour, you’ll review your report together, discuss the reasoning behind the findings, and identify the highest-leverage improvements for your organisation.

After 60 minutes you’ll leave knowing
  • which findings actually matter
  • where to invest first
  • what can wait
  • how to explain the results internally
Prefer exploring on your own first? Read the chapter matched to your assessment
Level 5 (systems and orchestration)

Reading Path

Prefer to go deeper on your own first? Here is your matched read.

Assessed at L4 — Specification-led. Your weakest disciplines are Guardrail Design and Context Engineering, tied at 4 — but the binding constraint is altitude, not a single discipline. Next read: the Level 5 (systems and orchestration) chapter, alongside the Level 3 chapter on harness engineering for the observability and critic-agent work.

Read on Leanpub