01 - The shift
From averages to context
Standardisation was historically needed as the rational response to cost. When measurement was limited, systems had to be built around averages. Production variation was expensive, and repetition was the most efficient operating model. With scarce data, products had to be designed for broad groups. The result was an economy built around generalisation, the average patient, the standard material, the fixed production line.
That model is not disappearing, but it is no longer the only scalable one. Systems can now be measured more continuously, designed more computationally, produced more flexibly and adapted more dynamically. The common thread is not personalisation in only a consumer sense. It is specificity: the ability of a system to respond to the particular conditions of a patient, product, environment or use case.
Specificity was valuable, but expensive. That cost is now falling.
02 - The evidence
It is already visible in regulated and commercial settings
The shift is no longer theoretical. In each domain below, though it is worth separating demonstrated capability from commercial maturity, the direction is consistent.
Of newly approved therapeutic molecular entities at the FDA in 2024 were personalised medicines, 18 in total. Personalised medicines have been at least a quarter of new approvals every year for a decade, up from under 10% just over a decade ago.12
Newly discovered stable materials identified by DeepMind's GNoME, from 2.2 million predicted stable crystal structures. Berkeley Lab's autonomous A-Lab then realised 36 of 57 targets over 17 days, the loop from prediction to synthesis beginning to close.45
Cost per part in a powder-bed-fusion example that held flat across 50 design variations. In high-mix, low-volume contexts, a 2026 review argues, the cost penalty for customisation can fall effectively to zero.6
The demand side is moving in the same direction. McKinsey finds that 71% of consumers now expect personalised interactions and 76% are frustrated when they do not receive them, while faster-growing companies generate 40% more of their revenue from personalisation than slower-growing peers.9 Consumer personalisation is not the same as personalised medicine or application-specific materials, but the expectation that systems respond to specific context is spreading across enterprise software, healthcare, finance and industrial workflows.
A current limitation
The gap between a laboratory result and a deployed system is large, and crossing it is where most of the cost and time of deep technology is spent. This leaves timing as a constraint, but the direction is consistent.
03 - Why it matters
Specificity makes complexity operational
A more specific world does not completely remove complexity. Specificity requires systems to measure more, adapt more, integrate more and perform under more variable conditions. In every domain above, the visible tailored product sits on top of a stack that has to work for the product to exist at all.
Patient-specific care depends on diagnostics, monitoring, data infrastructure and reimbursement. Application-specific materials depend on computational discovery, synthesis, testing and qualification. Context-specific AI depends on memory, security, orchestration, integration and governance. Configurable manufacturing depends on design automation, inspection, process control and certification. The more specific the system, the more it leans on these enabling layers.
Specificity creates new bottlenecks which creates new opportunities.
04 - Where the value concentrates
The value moves down the stack
This reframes where durable value is likely to sit. Many "personalised" products will be features, services or fragmented markets, easily replicated and competed away. The more durable position tends to be the infrastructure that lets specificity scale: the layers that are hard to build, hard to qualify, and shared across many tailored end products. In each domain, the enabling layer is needed regardless of which tailored product ultimately wins.
A second pattern is where the difficulty now lies. In most of these domains the scientific insight is increasingly demonstrated; the harder and slower work is engineering, integration, qualification and scale-up, turning a capability into something deployable, certified and economic. In regulated and safety-critical fields, adoption is gated as much by trust and qualification as by raw performance, and the layers that supply that trust are correspondingly difficult to displace.
The pattern is less a single sector than a recurrence across them: as specificity becomes viable, complexity becomes operational, and the enabling layers that carry it become the point and foundation at which value concentrates.
The cost of building systems around context, rather than averages, is falling and the enduring value will accrue to the infrastructure that makes context operational at scale.