Why Not Us? Why Can't We? The Barrier Drops from Millions to Thousands | Geometry of Trust | The Map Back to You - Part 2
This is the second post in the Geometry of Trust future series. Part 1 argued that communities can own and verify their own AI locally. This post asks what changes economically when they do.
The majority of knowledge work is busy work
A solicitor spends most of their time on legal research, document review, and case preparation. The actual legal reasoning — the part that requires judgement, experience, and understanding of the client — is a fraction of the working day. The rest is searching, cross-referencing, formatting, chasing.
The same pattern applies across every profession. An accountant spends most of their time on data entry and compliance checks, not financial strategy. A teacher spends most of their time on material prep and marking, not teaching. A doctor spends most of their time on cross-referencing and admin, not clinical judgement.
AI automates the busy work. That’s not new. What’s new is the ability to run that automation locally, on verified models, with data that stays in the practice, the school, the surgery — rather than flowing to a platform company.
What changes when the automation is local
Knowledge work
A local solicitor’s practice runs its own legal research AI. Trained on UK case law, statute, and regulatory guidance. Scoped to legal.research. Verified against legal reasoning values — precedent, procedural fairness, accuracy of citation. The solicitor’s judgement drives the strategy. AI does the searching.
A local accountancy firm runs its own financial AI. The expertise stays in the community. The accountant’s relationships with their clients, their understanding of local business conditions — that stays human. The cross-referencing and compliance checking becomes automated, verified, inspectable.
Creative industries
This is where the economic transformation gets interesting. Film production, music production, documentary making, graphic design — all of these currently have barriers to entry that concentrate them in a handful of cities. London, Los Angeles, a few others.
A filmmaker in Hull doesn’t need a London studio budget to make a documentary. Script development, storyboarding, editing, music, translation — AI tools running locally handle the production work. The filmmaker’s taste, their story sense, their connection to the subject — that stays human. AI removes the production barrier.
The same applies to music. A producer running Suno-class models locally doesn’t need a studio booking. Every bedroom becomes a production studio. Every city becomes a creative hub.
Education
AI tutoring tailored to the local curriculum. Verified against educational values — is the child learning? Not engagement metrics — is the child clicking? The difference matters, and it’s a governance decision the school makes, not the platform.
The model runs in the school. Data doesn’t leave the building. Teachers are augmented, not replaced. The teacher’s relationship with the class — knowing which kid is struggling silently, which kid needs challenge not support — stays human.
Tourism
AI-powered interactive city guides. Multilingual translation running locally. Accessibility tools — audio description, sign language generation. Cultural heritage presented through AI storytelling. Every city becomes a destination, not just London and Edinburgh.
Media and journalism
Local news augmented by AI research and data analysis. Investigative journalism with AI pattern recognition. Community radio and podcasts with AI production tools. Local voices amplified, not replaced by national chains.
The pattern
Every industry that currently depends on expensive expertise or distant platforms can be localised. Small verified AI makes the expertise local. The value stays in the community.
The barrier to entry drops from millions to thousands. Not because the AI is free — hardware costs money, training costs money, governance costs time. But because the economics of a 500M-parameter model on a single GPU are fundamentally different from the economics of a 70B-parameter model in a data centre.
The shift isn’t from expensive to cheap. It’s from rented to owned. From value flowing upward to value staying local.
Next in the future series: if communities can own their own AI and the economics work, what does collaboration look like? The answer involves sharing geometry rather than data — and it changes the relationship between communities from customer-vendor to peer-peer.