We believe in AI. That means taking its impact seriously.

There is a version of the AI future we would love to help build. A small business owner gets through the paperwork and closes the laptop. Someone learning a difficult subject finds another way into it. A team spends less of its day chasing information and more of it doing work that needs their experience.

Those are the possibilities that interest us at Fox by the Lake. Useful, ordinary improvements to a working day. More confidence. A little breathing room. Technology earns its place when it helps with something that matters to the person using it.

But it would be far too easy to tell that story and stop there. The same technology can put jobs under pressure, consume substantial resources and invite people to place trust where it has not been earned. Anyone choosing to build with AI has to take that seriously. We do, because the people on the receiving end of our decisions matter more than our enthusiasm for a new tool.

What happens to the time we save?

Imagine a team getting back an hour that used to disappear into administration. What happens next? They could spend it helping a customer properly, learning something useful or finishing at a reasonable time. They could also be handed another hour’s work. The software alone does not decide which future arrives.

That is why we want conversations about AI to start with the people doing the work. What gets in their way? Which tasks are draining time and attention? What would they like to do better? Reducing repetitive administration, making information easier to find and supporting learning are worthwhile ambitions. Measuring success only by how many roles disappear would miss what drew us to this work.

There is no honest promise that every job will be safe. The International Labour Organization’s 2025 research identifies transformation as the most likely overall effect of generative AI, with exposure varying substantially between occupations. That is an assessment of how work might change, not a guarantee against job losses. A person worried about their livelihood deserves a more serious answer than ‘AI will free you up’.

Our proposed approach is to involve affected people, discuss training and redeployment early, and challenge projects whose main purpose is removing people. We will not market mass workforce replacement as the purpose of our services. We cannot control every employment decision a client makes, but we can choose the advice we give and the work we accept. Time saved should create room for better work, without automatically demanding longer hours, faster replies or permanent availability.

A conversation is not the same as care

Learning, wellbeing and nature sit together in the wider vision for Fox by the Lake. We care about people having room for life beyond a screen. That makes the way conversational AI is designed particularly important to us. A system that sounds patient and understanding can become easy to trust, especially when someone is struggling.

Clearer information, manageable steps and accessible learning are things we want technology to help with. They are design goals. They are not evidence that our software treats anxiety, prevents burnout or improves a clinical condition. An assistant can sound reassuring and still be wrong.

The World Health Organization’s guidance recognises potential benefits from AI in health while warning about inaccurate information, bias, privacy risks and excessive trust in automated outputs. That is a reason to take the boundaries seriously, not an endorsement of any particular product.

Recent accounts of chatbot-associated harm are troubling. They deserve careful attention, including care with the evidence itself. A September 2026 preprint by Morrin and colleagues examines retrospective, self-selected, unverified reports. It cannot establish how common these harms are or prove that chatbot use caused them. We should take warning signs seriously without presenting early research as a settled account of what happens to everyone.

For us, a general-purpose AI assistant must not be presented as a therapist, a crisis service or a replacement for human relationships. Our proposed approach requires clear limits, consideration of foreseeable misuse and appropriate routes towards human support. Work involving children or vulnerable people needs additional safeguarding and specialist input. We want to help someone find their next useful step, including a step away from the technology.

There is a physical world behind the screen

The word ‘cloud’ is wonderfully convenient. It also makes it easy to forget the buildings, equipment, electricity and water behind a response. Those costs belong in this conversation, even when they are difficult to see from a laptop.

The International Energy Agency’s 2026 analysis projects that global data-centre electricity consumption will roughly double between 2025 and 2030. That figure covers more than AI, although AI-focused data centres are expected to grow faster. The report also recognises substantial efficiency improvements. Individual tasks can get more efficient while growing demand and more energy-intensive applications push total consumption upwards.

Water is more complicated than a neat figure beside each prompt. Research by Li and colleagues examines AI’s water footprint, while UNEP describes environmental impacts across the technology’s lifecycle. Depending on the infrastructure, water is used in cooling, electricity generation and equipment manufacture. The impact varies with location, weather, technology and local water availability. There is no single water-use figure that fairly describes every AI interaction.

We cannot turn that complexity into a claim that our AI is impact-free. We can ask better questions about what we build. Does this job need AI at all? Could an approved document answer the question? Do we need fifty generated options when three useful ones would do? Does a process need to keep checking every minute, or could it wait until something changes?

Those questions will shape our proposed design approach: use a simpler solution where it works, choose an appropriate model, avoid repeated processing, reuse suitable results and limit unnecessary background activity. Efficiency has to be judged against the completed job. A cheap response that leaves someone fixing mistakes is not the whole calculation.

We will also ask suppliers what their environmental figures cover, where processing happens when that information is available, and how they address energy and water use. Some answers may remain unavailable. A supplier’s ambition cannot be treated as an achieved result, and gaps in the information need to stay visible.

Giving back should mean something

We intend to establish an environmental contribution budget alongside work to reduce avoidable resource use. Water conservation, wetland or catchment restoration and additional climate action are the proposed areas of interest. Arrangements are not confirmed. When they are, we intend to publish the organisations supported, the amounts contributed and the outcomes they report.

We would want those contributions to be worthwhile in their own right. A donation does not prove that an AI interaction has become environmentally neutral. Water impacts are local: support for a project in one place does not automatically relieve pressure on a water supply elsewhere. Carbon accounting and water stewardship deal with different impacts. Neither gives us permission to stop improving how we operate.

The choices we want to be judged on

It is easy to write that people come first. The test is what happens when that principle makes a decision less convenient: when a launch needs more review, a client wants more automation than is appropriate, or a system needs to pause.

Our proposed commitments require named ownership, clear limits on what AI systems can access and do, and meaningful human review where decisions have significant consequences. They include assessment before deployment, further review when models or capabilities change, a route for concerns and the ability to pause a system. These are proposed arrangements that need approval and operational work, not a claim that every safeguard is already in place.

We also intend to report progress, keeping completed work separate from plans and measurements separate from estimates. Some approaches will need to change as the evidence develops. We should be able to explain those changes openly.

We still feel the excitement of what useful AI could make possible. We want to hold on to it, while being thoughtful about the work it asks people to trust us with. A business with more time for its customers. A learner who feels able to try again. More room for a life outside work. That is a future worth putting effort into.

Our Responsibility: People, Planet & AI page sets out the full proposed commitments, their qualifications and how we intend to report progress.