How We Use AI Responsibly
AI is the most useful technology we've worked with in twenty years. It also raises fair questions about jobs, accuracy, privacy, and energy. Here's how we answer them, and we expect you to hold us to it.
People firstA person in chargeAlways labeled as AI
Nine promises we hold ourselves to.
They answer the questions people ask most about AI: jobs, judgment, quality, honesty, data, energy, and fairness.
People first
AI should make people better at their work, not replace them.
We build AI to take on the repetitive work and give people time back for judgment, relationships, and new ideas.
A person stays in charge
When a decision carries real weight, a person reviews it first.
AI can draft, sort, suggest, and act. That isn't doubt about the technology. It's respect for what's at stake.
We test it before you trust it
Nothing we build goes live, or stays live, unmeasured.
AI makes mistakes, and it can be confidently wrong. So we agree targets with you before launch, and we keep checking after.
You'll always know when it's AI
AI is always labeled, and it never pretends to be a person.
We'd rather show you what AI can do than tell you. This website, our videos, our AI agents, and most of our articles are made with AI, and a person approves all of it.
We take care with your data
We never sell your data or use it to train AI models.
We use only the information a job needs. We never sell it or share it for advertising. The service providers named in our Privacy Policy handle it only to run our services for us.
We use AI's resources with care
AI runs on real electricity and water, so we don't waste it.
We don't run the data centers behind AI, but we do choose how much we ask of them.
AI should treat people fairly
When AI affects people, we check that it treats them fairly.
Customers, staff, and job applicants should get the same fair treatment, whoever they are.
We answer for these promises
We review them at least every six months, and a real person reads every message you send us.
You can always reach a person
Whenever you're working with our AI, you can ask for a person, and you'll get one.
What each promise means in practice.
Choose a commitment to see how it shows up in our work.
AI should make people better at their work, not replace them.
- We measure success in hours given back and what your team does with them.
- We don't pitch cutting jobs as the goal of a project.
- We involve the people whose work will change from the start, and we train them to use what we build.
When a decision carries real weight, a person reviews it first.
- Every system we build has a named person who owns its results.
- We agree with you where people review, approve, or step in, based on what's at stake.
- AI agents work with limited permissions, keep a record of what they do, and stop for approval where it matters.
- People can always reach a person, and they can override what the AI suggests.
- We don't build AI that makes final decisions about people's jobs, pay, or access to services on its own.
Nothing we build goes live, or stays live, unmeasured.
- We set targets with you up front: how accurate it needs to be, how fast, and where a person reviews the result.
- We test each system against those targets before anyone relies on it, and keep tracking it once it's live.
- We tell you plainly what a system can't do.
- If a system isn't good enough, we fix it or switch it off.
AI is always labeled, and it never pretends to be a person.
- Our AI agents are labeled as AI everywhere they appear.
- Videos made with AI twins or AI presenters are labeled.
- Our AI agents' faces are AI-generated and aren't based on real people. We only use a real person's likeness, like Chris Ryan's AI twin, with their consent.
- When AI helped write one of our articles, the byline says so.
- We only make claims about results that we can show evidence for.
We never sell your data or use it to train AI models.
- We use only the information a job needs.
- We never sell your data or use it to train AI models.
- Only the service providers named in our Privacy Policy handle your data for us, and the policy explains your choices.
- Our Security & Trust page explains how we protect access to your data.
AI runs on real electricity and water, so we don't waste it.
- We size the model to the job, and check that choice when we scope each project.
- We don't use AI where a simpler tool does the job.
- We test with small, low-cost runs before large ones, and where we can, we fix work instead of generating it again from scratch.
- When two providers can do the job equally well, we choose the one that's more open about its energy and water use.
When AI affects people, we check that it treats them fairly.
- Before launch, we test each system with sample cases that differ only in details like name, age, or location, and check that the results don't change unfairly.
- If the results do change, we fix the system before anyone relies on it.
- We keep checking after launch.
We review them at least every six months, and a real person reads every message you send us.
Whenever you're working with our AI, you can ask for a person, and you'll get one.
What we won't build.
These are the lines we hold on every project, so you always know what to expect from us.
- AI that secretly monitors employees
- Chatbots or agents that pretend to be human
- AI likenesses of real people made without their consent
- Systems that make high-stakes decisions about people with no person involved
- AI designed to manipulate or deceive people
We can build tools that track productivity and capacity, as long as the people they cover know about them.
Who's accountable.
If you think we've fallen short, or you have a question, tell us. A real person reads every message.
Owners
Chris Ryan and James Ashton
Reviews
At least every six months
Questions
Frameworks we use as guides
Guides onlyWe use them to shape how we work, not as a certification.
Last reviewed November 12, 2026
People deserve AI that's done well
If you see AI the same way, book a call with our team or start with the free AI Assessment.
- Your data stays yours, always
- Never used to train AI models
- Never sold or shared