Approval first
Nothing important goes out until a person has said so, unless an owner has chosen to let it.
Responsible AI
Responsible AI at Dvilite
People decide what our AI may do, and can switch it off.
- Approval first: nothing goes out until a person says so
- Handover to a person when the AI is unsure or a customer is upset
- Owners choose which actions are automatic and can switch them off
- Every action is logged so it can be checked
Our approach in one paragraph
AI is useful and it can be wrong. We build it so that it drafts and suggests, a person approves, every action is recorded, and anyone responsible can switch it off. The same rules apply to the AI we use in our own work and to the AI we build for clients and in Bizilite.
Our four principles
A person can take over
When the AI is unsure or a customer is upset, the conversation passes to a person.
Everything is logged
Actions are recorded so they can be checked, explained and corrected.
Your data stays yours
You decide what the AI may use. You can export or delete your data.
How we use AI in delivery
Our AI drafts and suggests. Engineers and owners review and approve what ships. AI does not release code, change client systems or contact clients on its own.
How AI works inside Bizilite
Bizilite drafts replies and follow-ups. Owners decide when it may act on its own, and a person can take over at any time.
Human oversight in practice
- Approval first: nothing goes out until a person says so
- Handover to a person when the AI is unsure or a customer is upset
- Owners choose which actions are automatic and can switch them off
- Every action is logged so it can be checked
Data handling
- Your data stays yours, and you can export or delete it
- We do not sell customer data
- The AI uses only the information you approve for it
- Access is limited to the people who need it
Limits: what our AI does not do
AI can be wrong. It can misunderstand a message, repeat a mistake in the information it was given, or sound more certain than it is. That is why people review important steps, and why we log what the AI does so you can check it. We tell you what a system cannot do before you rely on it.
Quick answers
Frequently asked questions
Does Dvilite's AI act without human approval?
Only where an owner has chosen to allow it. By default it drafts and suggests, and a person approves.
Can a person take over from the AI?
Yes, at any time. The AI also hands over when it is unsure.
Is my data used for anything else?
Your data is used for your assistant. We do not sell customer data. You can export or delete it.
How do you check what the AI did?
Actions are logged, so you can see what happened and why.
Can AI make mistakes?
Yes. That is why important steps need a person, and why we state the limits of every system.
Who do I ask about our AI?
Contact us. A person will answer.