Last updated: September 2026. Reviewed: every six months.
1. Why we have this policy
We use AI to answer support questions and to help us build and improve the platform. Our shops work in a food movement that values honesty about how things are made, and we think the same standard should apply to how we run our business. This policy says where we use AI, what rules we hold ourselves to, and what our shops can expect from us.
2. What AI makes possible for your shop
Using AI lets a small team do things for you that we couldn't otherwise:
Answer faster. You get a reply straight away, at any hour, rather than waiting for someone to be at their desk.
Get you to the right help sooner. The assistant points you to the relevant support article as soon as you ask, so standard questions are often sorted in minutes.
Put people where they're needed most. With routine questions handled, the team has more time for the complex, urgent or unusual problems that need a human.
Spot problems and patterns earlier. Support conversations show us where shops are getting stuck, which features cause confusion and where something isn't working. That shapes what we fix, what we build and which help articles we write next.
3. Our principles
We tell you when something came from AI.
A person is accountable for everything we send you, and for everything we ship.
You can always bypass the AI and reach a person.
We use AI to free up our time for the platform, not to avoid talking to you.
We monitor what the AI says and keep improving it.
4. Where we use AI
Standard Support. An AI assistant answers questions from shops in chat and by email. It currently resolves around two-thirds of tickets at any hour.
Technical Support. AI tools may be used under technical guidance to debug complex or specific technical queries, especially where diagnosing the issue requires tracing a shop's or customer's recorded activity on the platform. Before data is accessed by AI tools, we remove direct contact details such as email addresses and phone numbers. Names and order history remain, in order to respond to the support requests accurately.
Building and improving the Ooooby tools and features. Our developers use AI tools to help write, review and test code; to investigate bugs; and to analyse how the platform is being used so we can decide what to fix and build next. This is a significant part of why a small team can ship as much as it does. Every change is reviewed by a developer and tested before it goes live. AI doesn't deploy anything.
Drafting. We sometimes use AI to draft replies, help articles and marketing copy.
Product. If we ever build features that use AI directly in the platform, where you or your customers are interacting with AI, we'll say so clearly in the interface and in the release notes, and we'll update this policy first.
5. Labelling
AI-generated responses are labelled as AI-generated. We don't give the AI a human name or let it sign off as a member of staff. Where we've used AI to draft something a person then reviewed and sent, that's human communication and is not labelled as AI.
6. Human oversight and reaching a person
You can always bypass the AI. Every AI response tells you how to reach a person, and asking for one is enough without having to explain why. A member of the team picks it up during normal working hours. This applies to every channel, every time.
The AI assistant also escalates on its own when it can't answer.
Nothing AI drafts for a person to send goes out unread. We check it and correct it.
No AI-written code reaches your shop without a developer reviewing and testing it. The accountability for what we ship is unchanged.
Mistakes are ours. If the AI gets something wrong, we own it and fix it, the same as if a person had.
7. Accuracy
We monitor the AI's responses and the quality of the answers it gives. When it gets something wrong, we identify it, correct it and retrain the assistant on our own help content, which we keep current as the platform changes. Improving it is ongoing work, not a one-off setup. It gets better because we keep working on it.
8. Your data
We only give AI tools the information needed to answer the question or solve the problem in front of them
We don't hand over your customer lists or your customers' personal data to AI tools for training
When we use AI in development, we work with test or anonymised data
We use business AI services under terms that don't allow our data or our code to be used to train public models, and we check this when we adopt a new tool
When we look at support conversations for trends, we look at patterns across shops, not at individual shops' business data.
9. What we won't use AI for
AI-generated imagery in our marketing. The food, farms and people we show you are real.
Pretending to be a human.
Generating reviews, testimonials or endorsements.
Shipping changes to your live shop without a human reviewing them.
10. Environmental impact
We'd rather say something honest and limited here than nothing at all.
AI uses energy and, through data centre cooling, water. The impact of a single question is small; the impact at industry scale is not, and it is growing. We don't think it's credible to claim AI use is environmentally neutral.
What we do instead:
Proportionality. We use AI where it removes real, repeated work – answering the same platform questions hundreds of times a month, or the routine parts of writing and testing software – rather than applying it to everything because it's available.
Smallest tool that works. We use efficient models sized to the task rather than the largest available by default.
Displacement, not addition. Automated support and AI-assisted development replace work that otherwise takes staff time and duplicated effort; they aren't layered on top.
No false precision. We can't tell you the footprint of a single support reply, because the companies providing these tools don't publish figures at that level.
We're not claiming to be at the cutting edge of this. We're claiming to be paying attention to it and to be willing to say plainly what the trade-off is: a modest ongoing environmental cost against a small team being able to support hundreds of small-scale food businesses.
11. Standing preference
If you'd rather not deal with the AI at all, tell us and we'll set your account so your support enquiries always go straight to a person. Expect slower replies as a result. We ask that you consult our support articles before contacting human support for standard platform questions.
This applies to standard support. AI-assisted development is part of how the platform itself gets built, so there's no version of Ooooby that's been built without it. Technical support may use AI tools to debug your specific support queries, but a human will reply to you.
12. Review
We'll review this policy every six months, or sooner if we change how we use AI.