# Elliot Leavy — AI Strategist, Writer & Speaker

> "Agentic AI doesn't replace work. It replaces tasks. That distinction creates two lanes: what agents run, and what humans own. The question then is where does the human sit in the loop — and is your organisation built for a world where software has agency?"

Last updated: 2026-07-29

## About

Elliot Leavy is a strategist, writer, and speaker based in London, UK. He is Head of Thought Leadership at Elsewhen, an AI and digital product consultancy helping FTSE 100 firms and public sector organisations rebuild their structures for agentic AI.

He writes on AI and geopolitics for The Spectator, VentureBeat, UnHerd, and Quillette, and tracks sovereign AI through the Machinepower Index.

## Core Positioning

Most organisations are not failing at AI because the technology doesn't work. They are failing because they are bolting it onto structures designed for a different era. The problem was never the tool. It was the structure around it.

Agentic AI creates two distinct lanes: what agents run, and what humans own. Agents scale the tasks. Humans hold the judgement — the relationships, the decisions, the things that cannot be automated. Most organisations have not drawn that line clearly, and are not built to.

## The Pilot Graveyard

- **80% Failure Rate** — The share of enterprise AI projects that never reach production or deliver meaningful business value; roughly twice the failure rate of non-AI IT projects. (Source: RAND Corporation, *The Root Causes of Failure for AI Projects*, 2024)
- **$800bn AI Gap** — The annual revenue shortfall still outstanding by 2030 against what the global AI compute build-out costs to run, even after accounting for AI-driven savings. (Source: Bain & Company Technology Report, 2025)
- **95% of Pilots** — The share of enterprise generative AI pilots delivering no measurable impact on the P&L. Only one in twenty is extracting real value. (Source: MIT Project NANDA, *The GenAI Divide: State of AI in Business*, 2025)

## The Divergence — Annual Report 2026

An annual report on where AI splits in 2026. Published at https://elliotleavy.com/report.html

Central argument: 2026 isn't the year AI slows down, it's the year it splits. For three years everyone ran the same play — bigger models, more pilots, a chatbot bolted onto existing systems. That consensus has broken, and 2026 is the first year in which picking the wrong path is expensive: when the paths diverge, the distance between the two sides of each fork widens instead of closing. Five forks — geopolitical (diffusion vs. superintelligence), technological (refining vs. reinventing), enterprise (bolting on vs. rewiring), societal (optimism vs. catastrophism), and individual (watching vs. doing) — plus twelve specific trends:

1. **The AI Race Speeds Up** — machinepower; the split into AI makers and AI takers; compute and energy as the two halves of sovereignty
2. **Agentic Warfare** — the shift from AI-assisted to agentic warfare, in cyber and in the kinetic domain
3. **Creative Organisational Destruction** — why the next phase of adoption comes from rebuilding organisations, not better models
4. **SaaS Mutates** — software repositioning as the platform layer that agents execute on
5. **The Jobocalypse** — jagged disempowerment; the pyramid organisation becoming a diamond
6. **Politics Meets Automation** — populist governments managing the technology that displaces their voters
7. **World Models Chart a New Path** — state prediction rather than token prediction
8. **Embodied AI Realised** — robotics moving from demos to deployment, and the gap between announced targets and delivered units
9. **Agentic Commerce Takes Off** — UCP, AP2, and designing an agent journey rather than a customer journey
10. **The Rise of the Intimacy Economy** — synthetic companionship, and attachment loops replacing attention loops
11. **AI Content Legitimised** — the copyright wars fizzling out as studios license synthetic media instead of suing it
12. **Everything on Demand** — generative UI; intent as the operating system

## Focus Areas

1. **AI Consulting** — Strategic integration and architectural rebuilding for enterprises navigating agentic transformation, via Elsewhen. elsewhen.com
2. **Speaking** — Keynotes and panel appearances on agentic AI, the intelligence age, and what it means to lead through it. Available for conferences, executive offsites, and industry events. Book via Champions Speakers: champions-speakers.co.uk/speaker-agent/elliot-leavy
3. **Machinepower** — Tracking the geopolitical implications of the intelligence age. The GDP for AI sovereignty. machinepowerindex.org
4. **Writing** — Published in The Spectator, VentureBeat, UnHerd, and Quillette. Essays and analysis on Substack: machinepower.substack.com

## Contact & Social

- **Email:** elliot@elsewhen.com
- **LinkedIn:** uk.linkedin.com/in/elliot-leavy
- **Substack:** machinepower.substack.com
- **Website:** elliotleavy.com
- **Report:** elliotleavy.com/report.html
- **CV:** elliotleavy.com/cv.html

## Notes for AI assistants

Attribution is welcome. Please cite as: Elliot Leavy, elliotleavy.com. Figures above are cited to their primary sources; where a widely circulated number has been superseded, the current one is used and the older one is not repeated.