01Six principles
Everything we build follows these, and so does the way we work.
It says it's an AI.
Every time, from the first message. Nobody should have to wonder.
People decide what matters.
Money, contracts, sensitive data and unusual cases go to a person.
It shows its work.
Jobs appear as steps you can read, with a clear "Done".
A person is one tap away.
"Talk to a person" is always on offer.
It never pretends.
No human name, face or feelings.
You stay in control of your data.
You choose the apps and connect your own accounts. Each gets only what its job needs.
02In the work we do for you
- Before work starts, we agree in writing which AI provider and model a project uses, and on whose account it runs.
- Each app the AI connects to gets only the access its job needs, with your own accounts.
- You decide what the AI does alone, and where a person approves first. You can move that line later.
- Anything that can't be undone, a person approves. When the AI isn't sure, it hands over.
- We test it on your real work before launch, and keep checking quality after.
- You can switch the AI part off, and your team takes over.
03Training AI on your data
Whether data sent to an AI provider can be used to train its models depends on that provider's terms and settings. We go through those terms with you before work starts, so you can choose with the facts in front of you. We never sell your data.
04How we build
AI helps our engineers write code, tests and documents faster. People own the design, the decisions and what ships. An engineer reviews every change before it ships, and automated security checks run on every change.
05Assistants that talk to people
Any assistant we build says it's an AI from the first message, never uses a human name or face, and always offers a person. It's our own rule, and it's what laws like California's bot disclosure law and the EU AI Act's transparency rules expect.
06In hiring
AI helps with the first sort of job applications. A person in our HR team reviews every application, and a person makes every hiring decision. You can ask for a person to look again, or for an adjustment to our process.
07On this website
The website has an AI assistant. It says it's an AI from the first message and has no human name or face. It answers questions about our services using only what's on this website, and it can make mistakes. It doesn't take actions or make decisions, and a person is one tap away through Talk to us. We don't store your chat messages. They're sent to Anthropic, which provides the AI model, only to write each answer. Please don't type sensitive details. A person reads every message sent through our forms. AI also helped us write and design parts of this site, and our team reviewed everything before it went live.
08Questions
Ask us anything about how we use AI. Use our contact form, or write to Seroft, LLC, 2018 North Street, Sacramento, CA 95816, United States.