TechTalk · AI Readiness Portfolio

Separate harness repo, partially bound

payments-tribe · 2026-08-14
Coverage4 of 4 subjects assessed; 0 stale; 0 unreachable.
Gap spread−0.29 (orders-api) to +1.50 (legacy-batch) — range 1.79
Synthetic example — constructed to illustrate the report shape. These figures are invented to show the report shape. They are not a reading of any real estate.

Coverage ledger

SubjectStatusReport dateAgeNote
orders-apiassessed2026-08-0113d
billingassessed2026-07-2916d
reportingassessed2026-08-068dharness pinned 412d
legacy-batchassessed2026-08-0212dharness declared, not bound

4 of 4 subjects assessed; 0 stale; 0 unreachable.

Dimension matrix

SubjectABAIWFOMTPOROAHRWPACADTeObGv
orders-api33434333333434
billing33434333323424
reporting32333333223323
legacy-batch21121222212111

Level in cell, coloured L1–L5. inherited · inherited-unbound · unmarked is local. Red outline marks the weakest cell in the column among active subjects. Non-active rows are de-emphasised and do not set the ceiling.
AB Agent behaviour · AI Agent input · WF Workflow · OM Operating model · TP Teams provide · OR Output role · OA Output artefact · HR Humans review · WP Work patterns · AC Agent composition · AD Agents… · Te Testing · Ob Observability · Gv Governance

Gap table

SubjectHabitat maturity meanCognitive level GapRegime
orders-api3.293-0.29Coherent
billing3.143-0.14Coherent
reporting2.713+0.29Coherent
legacy-batch1.503+1.50Ambition outpaces enablement

Spread

orders-api -0.29
billing -0.14
reporting +0.29
legacy-batch +1.50
-0.50gap = cognitive read − 14-dimension habitat mean+1.71

One team, one way of working, four habitats. The same engineers who sit comfortably Coherent in orders-api are working a full level and a half ahead of their environment in legacy-batch. That is not a capability difference — it is an environment difference, and a single-repository assessment of orders-api would have reported a healthy team and missed it.

Portfolio regime

Portfolio regime

Distributed

Split ceiling

Common weak — enablement backlog

Observability, Agent composition — weak in three of four, including subjects where the harness does bind. The shared harness does not supply these, so binding harder will not fix them.

Locally weak — team backlogs

legacy-batch: Workflow, Teams provide, Testing, Governance — all four inherited-unbound. A binding failure, not a team failure. It belongs to whoever owns the harness.

What this estate should do next

Bind the harness in legacy-batch. The cheapest uplift in the estate, and not a build — the testing policy and governance model already exist. Four dimensions move the day it lands.

Refresh the reporting pin. It is bound, which is the hard part. It is 412 days behind, which is the easy part.

Then lift observability and agent composition in the harness itself — after the binding work, not before, or you improve a harness half the estate cannot see.

The instinct on seeing this matrix is to treat legacy-batch as a struggling team. The provenance marks say otherwise: the capability exists twenty metres away.