Scenario 0: How AI Could Reshape Ordinary Life

Present Day

Who’s Rachel, and why are we starting with her?

A lot of conversations about the future of AI start with grand predictions about capability, compute, and capital: how many billions are being spent, how fast models are improving, which world power is ahead of the others. What gets overlooked is the average person actually living through these changes. Someone commuting and cooking dinner, worrying about their child’s screen time, wondering whether their job will exist in five years.

We think that this omission matters. To kick off our analysis of the future of AI, we’re starting with Rachel: a character we’ve built to stand in for the ordinary person navigating this decade of change. The scenarios that follow are built on extensive research and workshops with experts, but forecasts about labour markets, energy grids, and geopolitical risk can make it easy to treat actual people as footnotes to a race between corporations and countries. Starting from an ordinary day, lived twice, is how we hope to drive home that AI’s success has to be measured by how it affects people’s lives.

This narrative sits alongside the systemic risk-specific scenario briefs in this series, translating their findings into a day-in-the-life of an average person, ten years from now.

Mid 2026

Where She’s At

35 Rachel, age 35, Outer London

Work Claims handler at a mid-sized insurer

Family A five-year-old daughter called Abby, who Rachel is raising as a single parent

Nothing about Rachel’s actual day-to-day life has changed too much yet. She still commutes to work every day, manages the same caseload, cooks dinner for her daughter. What’s changed, however, is the background noise, and the worries keeping her up at night.

Her company has just rolled out a new AI assistant agent, and a target adoption rate on everyone’s quarterly review, with increased targets employees have to hit for claims processed. Her daughter’s new primary school keeps implementing new forms of technology in the classroom, supposedly to boost tech literacy, but she’s worried about her daughter’s soaring screen-time despite her efforts to keep screens out of the home. The news, when she skims it over breakfast, only makes her worries worse: heatwaves against the backdrop of an enormous data centre boom that’s eating up more and more energy and water; blockbuster IPOs concentrating enormous wealth in the hands of tech entrepreneurs while hunger and disease kill ever more people in the Global South; all while geopolitical conflicts worsen all over the world.

Where is this going to go in the next decade?

Ten Years Later

A Day in 2035

The same day in Rachel’s life, ten years from now, read across the two ends of the range the other five scenarios describe. Each moment is given twice: once in the worst case, once in the best.

Ten Years Later

06:40

Rachel wakes up to news from the war

Best case

Finally, after almost six years of conflict, a complete ceasefire between the US, Israel, and Iran has been reached. This step has been taken as part of the international push towards a new AI-era security architecture. In the same announcement, the US and China have also ratified a binding agreement restricting autonomous weapons and AI-enabled bioweapons.¹

Worst case

Notifications and emergency alerts about an Iranian strike killing several civilians outside of a UK air base — the latest escalation in the US’s ‘forever war’. Autonomous weapons systems have massively driven up casualties despite their promise of higher precision. Along with the proliferation of (potentially fake) explicit footage which both sides weaponise for propaganda, the war shows no signs of abating.²

07:15

Passing by the fridge

Best case

Energy bills are down. The local grid has been updated over the past few years to dynamically schedule loads based on demand forecasts; this has helped to even out the big fluctuations in energy demand driven by AI training. Data centres are now more water and energy efficient, as they are often fully powered by local renewables.³

Overall, global emissions have started coming down; the large frontier AI models have mostly been replaced by smaller, specific, and also more secure models that are adapted to specific business and personal tasks.⁴ AI has also made other parts of the energy system more efficient: HVAC optimisation in buildings is increasingly AI-enabled, predictive maintenance powered by AI has reduced downtime and costs, grid integration for renewables has accelerated because of improved forecasting and dynamic balancing capabilities — all leading to significant energy savings across geographies.⁵

Worst case

There’s a record-breaking heatwave ripping across Europe. The office air conditioner has been providing some relief over recent weeks, so Rachel is rushing to leave the house early. Passing the fridge on her way out, the energy bill is higher than it’s ever been. The surge in demand for energy, driven by both data centre and AC buildout, has sent consumer bills soaring.⁶

08:20

Abby leaves for school

Best case

Abby wakes up ready and excited for school. Her school has implemented AI technologies for specific use cases, which has greatly increased the amount of time her daughter spends with her teacher in small groups and one-on-one. This is because many of Abby’s teacher’s administrative tasks have been augmented by AI, freeing up more time for students, and also because human teaching is supplemented with an AI tutor that teaches Socratically (a chatbot that encourages critical thinking rather than handing over answers, and is upfront about what it doesn’t know and when students should defer to an in-person educator).

Abby is doing well at school and her social skills are well developed. Parent and teacher satisfaction have been boosted since these changes have been implemented.⁷

Worst case

Rachel’s daughter, Abby, is glued to an AI chatbot ‘friend’ that’s coached Abby on how to fake being sick to get out of school. Abby has attempted this repeatedly over the last weeks. Over breakfast, there’s a fight over the tablet that Rachel barely wins.

Rachel sends Abby off to a school hollowed out by AI-driven edtech. Her school, which has decided to cut costs, has automated many teachers — Abby spends most of her school day interfacing with AI bots on her tablet, and most feedback on her work also comes from her AI teacher. Academically, this seems to have had a negative effect — Abby isn’t doing well on her standardised tests. Socially, she also seems to be struggling with making friends.⁸

09:30

Rachel arrives at her job

Best case

Augmentation has been the priority for AI deployment at her firm.⁹ Leadership asked staff what was and wasn’t working before rolling anything out, and developed tools after extensively consulting with employees.¹⁰ All employees have also received training on how to evaluate AI outputs, with particular emphasis on mitigating automation bias. The main tool that has stuck is a small proprietary model trained on the firm’s own claims data, which is built-to-fit, narrow, legible, and upholds data privacy.

Rachel’s day is now spent focused on more complex, and high-level tasks: edge cases and judgement calls. She’s also mentoring two junior colleagues from a previous customer assistant job, who have joined the claims team via a retraining scheme. She herself is midway through a firm-sponsored upskilling track toward underwriting. She’s feeling fulfilled and optimistic about her career, and leading a growing and increasingly skilful team.

Worst case

The insurer where Rachel leads a team of claim handlers has recently been acquired by a PE-backed ‘AI transformation’ rollup. Her firm’s main claims processing tool is built off of a frontier large language model, and the uneven nature of its training data and its opacity means that the model often makes wrong or biased recommendations. So far, Rachel’s junior direct reports mostly catch these mistakes in their manual checks.

However, Rachel will have to reduce her team size drastically soon, which makes her worried about the reliability of her team’s work: the model provides no reasoning and is programmed to flag the smallest number of borderline cases possible; the fewer juniors trained ‘the traditional way’, the more her team’s judgement might worsen. But all of this cost cutting is going ahead, despite the onslaught of consumer complaints that have already led the regulator to open an investigation into the firm’s use of automated decision-making.¹¹

As a result, Rachel fears that her own capacity will be incredibly strained going forward: she will have to attempt to evaluate all of her team’s work and also support the regulator’s probe. Will she ultimately have to cover up what she knows are badly automated decisions? All of this is happening in a context where everyone’s computer use is monitored to improve the model that’s meant to eventually fully replace them. Morale is at an all-time-low, but in the macroeconomic environment, no one can afford to quit.

18:05

On the train home

Best case

It’s a smooth journey home, due to fewer delays from AI-enabled predictive maintenance tools that catch wear in equipment before it causes a breakdown. More generally, there’s been a growing pattern of public services working a little better. The deployment of custom-built AI across the UK government, trained on public services data, has helped to overcome some financial bottlenecks: shorter NHS waits because triage and admin have been AI-augmented, more targeted preventative health measures enabled by AI prediction systems (saving costs), fewer accidents and traffic jams due to predictive maintenance leaving fewer potholes unfixed.¹²

Worst case

Trains have been massively delayed, due to the heat waves rolling over Europe. On her arduous journey home, an FT story breaks: the pension fund her employer auto-enrolled her into has a meaningful position in a fund backing an autonomous weapons company, that is now implicated in the strike — the very one from the morning’s headlines. Rachel reckons with the realisation that she’s an unwitting investor in the war that has been terrifying her for years.¹³

20:30

Dinner and winding down

Best case

After Abby returns from her after-school club, Rachel and Abby have dinner together. Abby is excited about all of the friends that she has made over the past year since she’s started at her new school. Rachel listens, happy to see her daughter happy.

Nothing has gone wrong today. If Rachel really thinks about it, everything has gone right, even. And there’s no reason to think it won’t keep going that way.

Worst case

At home, Rachel’s daughter is sullen and spends most of the evening after dinner staying in her room, talking to her chatbot. Rachel is worried about the nature of this relationship, about whether the conversations occurring are inappropriate, and whether the tech companies behind the chatbot are taking her daughter’s data. But she has no way to control or monitor the situation, as these chatbots don’t have parental control switches yet.

Lying in bed at night, unable to sleep, Rachel worries about her daughter’s future, and her own financial situation, as she wonders how much longer she’ll be able to hold down her job.

Where each thread comes from

Every element of Rachel’s two days is drawn from one of the five systemic scenarios in this series. Each one sets out the evidence, the branching point, and what investors can do about it.

Scenario I: AI, Energy & Water — the energy bill, the heatwave, and the data-centre build-out.
Scenario II: AI, Jobs & Productivity — Rachel’s claims team, and whether the tools augment or replace it.
Scenario III: AI, Dual-Use & Defence — the strike in the morning headlines, and the pension fund behind it.
Scenario IV: AI, Children, and Learning — Abby’s school, her tablet, and her chatbot.
Scenario V: AI, Geopolitics & Power — the security architecture that either holds or fails.

References

The most recent historical parallel is the Nuclear Non-Proliferation Treaty, entered into force in 1970, which established a multilateral framework for containing nuclear weapons in the aftermath of the Cold War arms race. Similarly, the Biological Weapons Convention (1975) banned an entire category of biological weapons outright.
Marijan (2022): ‘Autonomous Weapons: The False Promise of Civilian Protection’ — pushes back on claims that autonomous weapons could reduce civilian harm, outlining instead how the unreliability and fragility of AI systems could make their battlefield deployment dangerous to civilians.
Ceres (2026): ‘Sustainable Solutions for Data Centers’ — outlines efficiency opportunities like advanced cooling and demand flexibility, noting data centers cutting demand for just ~40 hours/year could avoid ~100GW of new capacity, saving $100–150bn.
HBR (2025): ‘The Case for Using Small Language Models’ — argues SLMs fine-tuned for specific domains offer greater accuracy, control, and efficiency than general-purpose LLMs, making them well-suited to specialised, knowledge-intensive tasks in fields like healthcare, finance, and agriculture.
Algburi et al. (2025): ‘The Role of Artificial Intelligence in Accelerating Renewable Energy Adoption for Global Energy Transformation’ — this literature review highlights the many opportunities that AI poses for accelerating renewable energy adoption, by improving forecasting accuracy, resource allocation, and grid stability.
Bloomberg (2025): ‘AI Data Centers Are Sending Power Bills Soaring’, found that electricity now costs as much as 267% more for a single month than it did five years ago in areas of the US located near significant data center activity.
Neugnot-Cerioli & Muss Laurenty (2024): ‘The Future of Child Development in the AI Era’ — underscores that AI’s main opportunities in education lie in its role as an assistive technology that augments rather than replaces teachers, and highlights Socratic questioning as an effective model for AI-supported learning.
Jared Cooney Horvath (2024): ‘The EdTech Revolution Has Failed’ — cites OECD and J-PAL research finding increased screentime in classrooms has worsened learning outcomes and failed to close achievement gaps.
HBR (2026): ‘Why Companies That Choose AI Augmentation Over Automation May Win in the Long Run’ — argues how leaders face a strategic choice between automation (cutting headcount for near-term cost savings) and augmentation (investing in tools that grow the business in new ways). While automation shows faster early gains, they argue augmentation wins out over time, since long-term performance depends on how employees feel about their work, whether they meaningfully engage with new tools, and whether top talent stays.
Participatory AI (2025): ‘A Scandinavian Approach to Human-Centered AI’ — shows how involving frontline workers in the design process leads to AI tools that better fit real workplace needs, through a case study of LEGO workers who took part in field studies, workshops, and prototyping sessions to help shape an AI-enhanced augmented reality system for their measurement tasks.
The use of AI to automate decisions in high risk sectors like healthcare and insurance is likely to face growing regulation and scrutiny. Guardian (2025): ‘New AI tool counters health insurance denials…’ — outlines how UnitedHealth’s nH Predict algorithm pressured case managers to cut Medicare patients’ care short, with a 90% error rate on appeal; a class-action lawsuit is ongoing, alleging that UnitedHealth exploited the fact that few patients knew to appeal.
OECD (2025): ‘Governing with Artificial Intelligence: Are Governments Ready?’ — identifies four broad opportunity areas for AI in government: automated, streamlined and tailored processes and services; better decision-making, sense-making and forecasting; enhanced accountability and anomaly detection; and unlocking opportunities for external stakeholders.
This is a situation that many pension fund enrollers are already in. Novara Media (2025): ‘Britain’s Biggest Pension Fund Invests in Israeli Weapons Firm Elbit Systems’ — found that Nest, the UK’s largest auto-enrolment pension scheme, holds a position in Israeli weapons firm Elbit Systems, with many members unaware.