--- title: "AI Stages of Adoption" date: 2024-06-13 description: The adoption model that places organisations by where investment actually flows: money, time, talent, data and process, not aspiration. author: Mario Thomas canonical: https://mariothomas.com/remake/library/ai-stages-of-adoption/ --- ## What it is {#what-it-is} Unlike traditional technology adoption models, AISA deliberately uses **investment** — money, time, talent, data, processes — rather than time as its progression axis. AI adoption does not follow a sequential, organisation-wide path; it happens simultaneously across business functions at varying speeds, and the model is built to read that reality rather than average it away. AISA was introduced in June 2024 in [Understanding the AI Stages of Adoption](/blog/ai-stages-of-adoption/), developed after observing that frameworks built for earlier technology waves — like the Cloud Stages of Adoption — do not capture AI's multi-speed, business-led adoption pattern. From first experiments to ecosystem-wide scale. Progression is measured by what the organisation is actually investing — and it is earned, not scheduled. - **Experimenting** Exploring and testing AI, often without formal oversight — shadow AI emerges as business units reach for consumer tools to solve immediate problems. Ad-hoc initiatives, small budgets, quick wins. - **Adopting** Adoption widens and formalises: AI is integrated into existing processes, initial governance and dedicated budget appear, and the AI Centre of Excellence takes shape. - **Optimising** Use is refined and value maximised: models tuned to the domain, governance and risk frameworks operating, clear ROI measurement, standardised ways of building and deploying. - **Transforming** AI stops assisting the business and starts redesigning it: processes, roles, and business models rebuilt around the capability; new AI-enabled products and revenue streams emerge. - **Scaling** The organisation embraces AI across the enterprise and its ecosystem: platform infrastructure, automated operations, an innovation culture, and advantage that compounds. The model on a page: five stages ordered by what the organisation is actually investing — money, time, talent, data, processes — not by how long it has been trying. Time on the journey proves nothing; investment is the evidence. Organisations typically have multiple AI initiatives at different stages simultaneously: marketing Optimising content generation while finance is Adopting fraud detection; operations Experimenting with predictive maintenance while customer service is Transforming around AI-powered support. Each function progresses at its own rate, set by risk tolerance, data readiness, technical capability, business urgency, and competitive pressure — and success in one area often catalyses adoption in the others. This is why AISA is assessed function by function, never as one averaged score. The Board's job is to govern the multi-speed picture: stage-appropriate governance and capability-building for each function, rather than a single-track programme that over-governs the experiments and under-governs the systems already in production. Within [Remake](/remake/), AISA is declared a **Model** — a way of seeing or structuring something. It is one of the lenses applied inside the ADAPT motion: read at **Remake: Align**, to establish where you truly are before the investment case is made, and again at **Remake: Diagnose**, where the stage evidence separates the real problem from the reported one. It carries both faces the framework requires: the teams doing the remaking use it to locate a function honestly, and the executive owner and board use the same reading to govern the multi-speed picture. One asset, both altitudes. The five stages, why the axis is investment rather than time, and the multi-speed reality — explained on camera in ten minutes, with the full transcript on the watch page. ## Questions {#questions} Look at the evidence, not the ambition. Widespread unofficial use of consumer AI tools and small isolated pilots is Experimenting. A dedicated budget, early policies, and an AI Centre of Excellence taking shape is Adopting. A documented strategy, a standing steering cadence, and multiple production systems is Optimising. AI embedded in core products and services with new revenue attached is Transforming. Enterprise-wide adoption with a self-sustaining ecosystem is Scaling. Assess each function separately — the averaged score hides the picture. Each transition has its own success factors. Experimenting to Adopting turns passive executive awareness into active sponsorship, and discretionary spend into dedicated budget. Adopting to Optimising standardises how models are built, deployed, and measured. Optimising to Transforming is the cultural shift — from AI-assisted to AI-first, with business-model innovation. Transforming to Scaling is standardisation at enterprise scale: platforms, automated operations, and ecosystem orchestration. Five patterns recur: governance paralysis (over-governing too early, under-governing too late), talent gaps for what the next stage requires, technical debt that blocks integration, cultural resistance to AI-driven change, and ROI pressure — expecting returns before the capability that produces them has been built. Progression is earned by removing the blocker, not by scheduling the next stage. No. The stages describe where investment currently sits, not a league table to climb. Some functions may sensibly remain at Adopting or Optimising for years; the failure mode is not a low stage but an ungoverned one — or investment badged as a later stage than the evidence supports. Not IT. The Centre of Excellence coordinates governance, capability, and value across every function, so it reports where accountability lives — typically the CFO or the risk function. Delegating it to IT re-frames an enterprise transformation as a technical programme, which is how adoption goes ungoverned.