[{"data":1,"prerenderedAt":61},["ShallowReactive",2],{"page:how-we-use-ai":3},{"id":4,"cta":5,"extension":11,"header":12,"meta":16,"sections":17,"seo":56,"slug":58,"stem":59,"__hash__":60},"pages\u002Fpages\u002Fhow-we-use-ai.yml",{"title":6,"note":7,"button":8},"Want to see the setup? Ask on the call.","The first call is free, takes 30 minutes and commits you to nothing.",{"label":9,"to":10},"Book a call","\u002Fcontact","yml",{"eyebrow":13,"title":14,"intro":15},"How we use AI","Faster because of AI. Safe because of the rules around it.","We use AI coding agents on every project. This page is the full list of rules we work under, written so your security lead can read it without a call. If a rule here isn't enough for you, tell us and we'll tighten it for your project.",{},[18,22,50],{"id":19,"label":20,"text":21},"why","Why it matters to you","The agents make us faster. Because we price by scope, that shows up on your side as a lower price and a shorter timeline, not as our margin. The rules below are what make the speed safe to use on your code.",{"id":23,"label":24,"items":25},"rules","The rules",[26,29,32,35,38,41,44,47],{"title":27,"text":28},"Agents get their own credentials","Every agent runs under its own identity, with access scoped to the task it's doing and nothing else. Read-only where read-only is enough. Credentials are revoked when the task ends. An agent never uses a person's keys, and never has more access than the engineer supervising it.",{"title":30,"text":31},"Agents work in isolation","Agents work in their own environments and branches. They have no route to production, no route to production data, and no way to deploy. Nothing they do reaches a live system until a person merges it and the normal pipeline ships it.",{"title":33,"text":34},"A named engineer reviews every change","Every line an agent writes goes through the same pull-request review as a line written by hand, by an engineer whose name is on the project. If it isn't good enough to have written ourselves, it doesn't merge.",{"title":36,"text":37},"Tests are the gate","Generated code runs the same test suite as everything else, in the same pipeline, and fails the same way. Agents are good at writing tests too, and those get reviewed as carefully as the code they test.",{"title":39,"text":40},"Your code is not training data","Client code is never used to train AI models and is deleted by the provider within 30 days. On request, we'll walk you through exactly which tools touch your code and under what terms.",{"title":42,"text":43},"Secrets never go into prompts","Agents read configuration the way the application does, from the environment, never as text in a prompt. Keys, tokens and connection strings are not in the conversation, so they can't end up in a log, a transcript or a model.",{"title":45,"text":46},"Every change has a name on it","Everything lands in your git history with the name of the engineer who reviewed and merged it. Nothing reaches your repository without a person signing off on it.",{"title":48,"text":49},"You set the limits","You can restrict which tools we use, exclude parts of the codebase, require your code to stay inside your own AWS or Google account, or opt out of AI-assisted work entirely for a project. It goes in the contract, and we price the project accordingly.",{"id":51,"label":52,"text":53,"link":54},"not-change","What this doesn't change","The thinking is still ours. Agents don't decide what to build, how to structure it, or when it's done. They write code faster than we can type, inside decisions a person has already made. That's why the sage in our name is still a human.\n\nQuestions a security review would ask and this page doesn't answer? Send them to us and we'll add the answer here.",{"label":55,"to":10},"Send them to us",{"title":13,"description":57},"The rules Software Sage works under when AI coding agents touch client code. Scoped credentials, human review, no training on your code.","how-we-use-ai","pages\u002Fhow-we-use-ai","4yMzZXSDD7M2iBALdiptmIdMsjV6QUnZGQ8v2_0d65Y",1793174665534]