AI: Tool or transformation?
The dominant story says AI adoption is urgent, inevitable, and mostly a question of which tools to choose and how fast to move. This Zone of Attention Scan of small and medium enterprises, consultancies, sole traders and not-for-profits found five patterns that framing leaves out, and a set of tensions that cannot be resolved by better planning.
What the scan found: five patterns
-
1. The failure rate no one mentions
Between 42% and 95% of AI projects fail to deliver measurable return, depending on the study; an MIT study found 95% of enterprise AI pilots fail to reach production or generate revenue impact. This finding circulates in specialist publications but rarely reaches adoption guidance for small businesses or non-profits. At the same time, "AI washing" is emerging: companies using AI adoption as cover for cost-cutting decisions that have little to do with automation. The question is not whether AI can deliver results, in some contexts it clearly does. It is whether the organisations succeeding are navigating trade-offs that most adoption guidance never mentions.
-
2. Efficiency gains have hidden costs
The promised savings are often achieved by shifting costs elsewhere: onto workers, through intensified labour, unpaid learning and normalised surveillance; onto communities, through job displacement or service degradation; onto the environment, through energy and water consumption that is rarely tracked; and onto the future, through technical debt and long-term dependency on platforms whose interests are not the same as yours. The question is not whether there are costs, there always are. It is whether their distribution is visible, and whether it is defensible.
-
3. The people most affected have the least say
In most cases the decision to adopt AI is made by leadership, investors or external consultants. Frontline workers, the people whose roles will be most affected, are consulted after the decision, not during it. They shape how the tool is implemented, but not whether it should be introduced at all, or what problem it is actually solving. This is not only a fairness question. It is a strategic one: the people closest to the work often see problems that leadership misses, and their silence, when it comes, is rarely neutral.
-
4. Augmentation often precedes displacement
"AI augments human work, it doesn't replace it" is true in many contexts, but augmentation and displacement are not opposites; they are often stages. A landmark 2025 Stanford study analysing payroll records from millions of US workers found a 13% relative decline in employment among early-career workers in high-AI-exposure occupations, while most experienced workers in the same roles saw employment stable or growing. The line between augmentation and displacement is not fixed; it shifts based on economic pressure, vendor incentives, and the choices organisations make, or avoid making.
-
5. Something is ending, whether you name it or not
AI introduction involves real losses that are rarely acknowledged: certain kinds of craft and mastery; slower, more relational ways of working; tacit knowledge that was never formalised but held the organisation together; autonomy, as work becomes more algorithmic and more monitored. When these losses are named, they are often dismissed as resistance to change. Yet grief is not the same as refusal; it is a signal that something meaningful is ending. Grief that is not given space goes underground, where it becomes disengagement, resentment, or quiet withdrawal from the work that matters most.
The tensions the scan holds
Anyone considering AI for their team is not facing a simple technology decision but a set of tensions that cannot be resolved by choosing one side. Efficiency versus meaning: AI can make work faster, and it can make it shallower; both are true. Augmentation versus displacement: it can extend what people are capable of, and replace what made their work meaningful; both are possible. Democratisation versus concentration: it can give smaller organisations capabilities they couldn't afford before, and concentrate power in the hands of platform providers; both are happening. The organisations that navigate AI introduction well are not the ones that move fastest. They move with clarity about what is being gained and what is being lost, and with care for the people whose work will change.
The full scan
This page carries the summary scan (March 2026), offered free, in a spirit of generosity. The full scan is a paid companion that does the deeper work: the complete movement cycle with more than 50 sources, nine contextual perspectives across communities and sectors, and a complete analysis of the structural and mythological drivers shaping AI adoption.
Summary scan: 13 pages, free, from The Seed Bank.
Full scan: 52 pages, A$49, from The Seed Bank.