Twenty years inside the gap. The work addresses it.

Chuks Anochie · Abu Dhabi · UK · GCC

0

Years inside the gap

Five optimisation waves across the UK and Gulf

£0M+

Portfolio value owned, in-house and in consulting

AiZen · Infosys Consulting · Capgemini

£0M+

Pre-sales and business development generated

Consulting only: Capgemini, Infosys Consulting, AiZen

$0.0M

Most recent AI portfolio

Two agentic AI platforms and a change framework at a Gulf national oil company

0%

Proposal win rate

Consulting only: Capgemini, Infosys Consulting, AiZen

At industrial-scale AI, the gap between investment and outcome is an accountability structure problem, not a technology problem. It surfaces most sharply here, across every domain the initiative touches, because AI moves fastest, costs most, and touches most.

The fifth wave is the wave the two disciplines were built to answer. The academic argument is in progress at Edinburgh Business School. Perspectives on the symptoms of the gap are published via The Transformation Generalist.

The situation

Performance deviations across eight Group Companies caught reactively, in unstructured operational logs, with no automated mitigation and no reuse of historical incident knowledge. The business, the central digital function and the prime supplier each held legitimate authority over AI behaviour.

The approach

Reframed the mandate from deploying AI to orchestrating decisions with AI support. Governance came before build. Decision rights sat on one page across three authorities, with veto rights held inside each domain only, and nothing went live until every cell was filled. Control was sized to what a wrong answer costs at each layer: a light correction on language, a confidence gate on classification, a named-owner gate on mitigation.

What was built

A three-layer AI core and an agentic orchestration layer inside the client’s private Azure environment. An NLP engine converting unstructured operational notes into structured root cause metadata. A confidence-scored classification model mapping near-real-time deviation signals to historical categories. A constraint-aware mitigation engine generating recommendations adjusted against live operational constraints across eight Group Companies. A named person accepts, rejects or ignores every recommendation and logs a reason. Degradation alerts reach the central digital function before they reach operations. All mitigation outputs HOTL+HITL-gated via ServiceNow with named decision owners at each tier.

What it produced

Live five months early across the first three Group Companies, inside a four-phase sequence covering eight, with zero governance breaches across the portfolio. 90% of decisions closed inside three to five days and 80% of escalations resolved below the steering committee. Governance benchmarks were set as reportable targets from day one: 80% response time reduction, 70% classification accuracy, 65% mitigation precision

What it showed

The build was the smaller half. My part was the governance model, the decision rights and the data ruling. Aligning both governance bodies, establishing decision rights before go-live, and keeping the $600M value line visible under delivery pressure, that coordination consumed the initiative. Designing that governance layer is what Value Governance Architecture is built on. Running the initiative through it in real time is what AI Value Orchestration runs on.