---
title: "Five Pillars of AI Capability"
date: 2025-03-02
description: Five capability pillars that answer the question following any stage placement: how an organisation knows it is truly there.
author: Mario Thomas
canonical: https://mariothomas.com/remake/library/five-pillars/
---

## What it is {#what-it-is}

The AI Stages of Adoption answers where you are; the Five Pillars answer the question that always follows — "how do we know when we're truly ready to move from one stage to the next?" Unlike technology-focused maturity models, the pillars recognise that AI success depends on balanced capability across all five domains: strength in one cannot compensate for weakness in another.

Balanced development across all five determines whether stage transitions succeed — weakness in one undermines progress in all.

- **Governance & Accountability** 
AI operating within safe and legal boundaries, with clear human oversight: who decides, and who is responsible — questions that sharpen as AI takes on more autonomous decision-making.

- **Technical Infrastructure** 
The foundational technology under the ambition: platforms, data architecture, and compute that enable development and deployment at scale.

- **Operational Excellence** 
The processes and practices for running AI reliably in production — consistently, efficiently, and beyond the pilot.

- **Value Realisation & Lifecycle** 
Capturing measurable value across the full lifecycle: use-case selection, vendor relationships, intellectual property, and honest ROI.

- **People, Culture & Adoption** 
The human dimension: change management, skills, and cultural readiness — whether sophisticated systems get used effectively or actively circumvented.

## What it examines {#what-it-examines}

A function's underlying capability across the five domains — the evidence beneath the AISA stage it claims. The inputs are concrete: governance artefacts in actual use, the state of the platform and data estate, production operating practices, value measurement discipline, and the honest temperature of adoption on the ground.

## The readings {#the-readings}

A per-pillar strength reading, and the pattern across them: balanced capability that can carry the next stage transition; an imbalance that will block it (the classic — technical infrastructure racing ahead of governance and adoption); or capability badged at a later stage than the evidence supports. Each transition on the AISA curve has its own pillar signature, so the reading tells you not just whether you are ready, but which pillar to invest in first.

## Running it {#running-it}

Function by function, alongside an AISA reading — the stage placement and the capability evidence belong in the same conversation, with the function owner in the room. The pillar-by-pillar discussion works best as structured self-assessment challenged against artefacts: not "do we have governance?" but "show me where it operated last month."

Within [Remake](/remake/), the Five Pillars are declared a **Diagnostic** — read at **Remake: Diagnose**, where the capability evidence separates the real constraint from the reported one, and again at **Remake: Plan**, where capability building is designed alongside delivery so the next stage transition is earned rather than scheduled.

## Questions {#questions}

Because pilots prove a use case, not a capability base. The transition from Experimenting to Adopting fails on missing governance and budget discipline; Adopting to Optimising fails on operational practice; beyond that, on culture and value measurement. The pillars name the missing foundation before the transition exposes it.

The weakest one. The model's core claim is that the pillars compound: weakness in any domain undermines progress in all of them. A portfolio-grade technical estate under weak governance is a liability, not an asset — and world-class governance over infrastructure that cannot deliver is theatre.