The Interface That Excludes Everyone Who Needs It Most

Let me start with a number that should stop every technology leader cold. Goldman Sachs says AI could add $7 trillion to global GDP. But the same Goldman Sachs, in early 2026, confirmed that AI investment contributed "basically zero" to measured economic growth last year. The reason? The bottleneck is not the technology. The bottleneck is adoption.

And adoption has a human face. It's the face of the 1.4 billion people who'll be over 60 by 2030. The billion people living with a disability right now. The non-technical professional staring at an interface that assumes she already thinks like an engineer. These aren't edge cases — they're most of the people on earth.

This is not a UX article. It is an economics argument. And it is backed by converging research across four domains that I want to walk you through.

Before we even talk about AI interfaces, consider the baseline they're being built on. The 2025 WebAIM Million Report analysed the top one million websites and found that 94.8% fail to meet basic WCAG accessibility standards. The average homepage contains 51 accessibility errors. And six years of effort have produced just 3.1 percentage points of improvement — while homepage complexity grew 61% in the same period.

Let that sink in. We are making digital surfaces harder to use three times faster than we're making them more accessible.

Now add AI interfaces on top of that broken foundation. Conversational AI introduces probabilistic outputs, multi-turn context, hallucination risk, and agentic actions — none of which has an accessibility standard equivalent to WCAG. We are building the most cognitively demanding interfaces in the history of computing on a base layer that excludes the majority of the people who most need decision support.

What makes this more than a design problem is the way it feeds on itself. Each stage of exclusion sets up the next, and the spiral gets harder to climb out of the longer it runs. Here's how it works:

Here's where the argument sharpens. AGI doesn't just sit on top of the accessibility problem we already have — it makes it worse, in four ways that set it apart from every technology wave before it.

The capability stratification is already here. In 2026, a documented "capability gap" separates power users — engineers and early adopters building autonomous agentic systems — from general consumers interacting with outdated, throttled interfaces. This isn't projected. It's happening now. The gap between those building AI systems and those navigating them is already producing two different economic realities.

The adoption divide is widening globally. Microsoft's Global AI Adoption data shows generative AI reached 16.3% of the world's population by end of 2025 — but the Global North grew nearly twice as fast as the Global South, widening the gap from 9.8 to 10.6 percentage points in a single year. The trajectory is divergence, not convergence.

Frictionless AI induces cognitive surrender. This one is counterintuitive. Interfaces designed for zero friction — delivering confident, fluent, complete answers — exploit the brain's tendency to conflate fluency with accuracy. For non-technical users, this produces automation bias: accepting AI outputs without the calibrated judgment that good decisions require. The interface that feels easiest to use may be the most cognitively harmful for those who most need support.

Complexity compounds on complexity. Web pages grew 61% more complex in six years with only 3.1% accessibility improvement. AI interfaces layer on top of that. There is no WCAG equivalent for probabilistic outputs, multi-turn reasoning, or agentic consequence. We are in a design vacuum.

I want to bring an unexpected lens to the leadership dimension of this problem. Ray Dalio's principles — developed at Bridgewater through decades of navigating complex, probabilistic, data-dense environments — map with striking precision onto the AGI-HCI challenge. Not metaphorically. Functionally.

Because here is what Dalio said in 2025: "The days of people making decisions in their own heads are ending. The future lies in brilliant people working with brilliant AI." But he also built his entire philosophy on the idea that success is founded on meaningful work and meaningful relationships — and that money, or technology, is merely the instrument toward those ends.

That tension is exactly the problem this series is investigating. Here is how his eight principles translate into organisational imperatives for inclusive AGI:

The research converges on four design imperatives. These aren't wishlist items. They're the conditions under which inclusive AGI-HCI becomes structurally possible.

If the moral argument doesn't move your organisation, the economic one should. This isn't a charitable concern.

People with disabilities and their families represent $13 trillion in annual spending power globally — more than China and Japan's GDP combined. 69% of disabled consumers leave inaccessible websites. UK businesses lose £17.1 billion annually by ignoring disabled customers. The disability and ageing population is the fastest-growing consumer demographic on earth, and companies building for them are outperforming those that don't — 1.6 times more revenue, 2.6 times more net income.

This is, in Dalio's framing, a failure of hyper-realism. A strategic misallocation hiding in plain sight, rationalised as a compliance problem rather than a market opportunity and a growth necessity.

AGI will be transformative, and the economic projections are real. But none of that value shows up without adoption — and adoption depends on interfaces that work across the full range of human cognition, age, and ability, not just for the fluent early adopter.

We are currently failing that test by a wide margin. 94.8% of digital surfaces fail basic accessibility standards. AI interfaces are more complex, not less. The people who most need decision support — older adults navigating healthcare, non-technical workers adapting to AI-transformed jobs, disabled people who stand to benefit most from AI augmentation — are the least represented in AI product development.

The gap will not close by itself. It will compound. Each stage of exclusion feeds the next, until the people excluded from AI interfaces are excluded from the economic and social systems AI mediates.

Leadership has to make this a design imperative, not an afterthought. And the Dalio framework tells us exactly how: embrace the reality of the gap, treat every adoption failure as a diagnostic signal, build transparency as the foundation not the decoration, and systematically map the boundary between human and AI judgment so that boundary is always visible to the user.

The path to AGI's promised value runs directly through accessible interface design. That is not an ethical claim. It is an economic one.