If you answer to auditors, boards or clients, you need an agent that can say it is not in the material. This one reads the same vault as you, and every answer arrives with the address it came from.
Ask the way you speak
A question in plain English, an answer with its address
Ask as you would ask a colleague: which slides mention gross margin, what the contract says about termination. The agent answers from what the engine read on entry, and brings the address: the slide, the note, the section, the sheet. Nothing it answers is built from another company’s data.
A chatbot gives you an answer; an agent with source gives you a claim you can check: the click opens slide 12 with its notes around it.
The conversation is with the AI
The AI talks; a deterministic engine touches the file
The AI understands what you want and chooses the operation. A deterministic engine executes it over the file’s known structure. Where the model’s own text goes into a file, as in translation, what comes back is checked, not assumed. Details in the conversation is with the AI.
Interfaces
The vault screen, your AI client, or your team's API
The same conversation happens in three places, over one vault. The AI-client channel is an open standard, so your team can build dashboards and automation on it. More in interfaces.
Capabilities
What it does, and what it refuses
The agent answers only from what the engine recorded, and it is built to say so when the record does not hold the answer. It is a language model: the address it gives you is how you check it. What the agent can do is a subset of what the API can do, by contract. The full list is in capabilities.
When a question becomes a request, the agent triggers it. The first: translate, with a receipt.