Tell your customers they are talking to an AI, with the facts.
For each assistant and each voice agent: the AI disclosure the vendor provides, whether it is on by default, whether your administrator can turn it off, and its state at your end. With the framework the catalog links to it, and the points to review with your counsel.
The question
“Do our customers know they are talking to an AI?” An assistant answers on your website, a voice agent picks up the phone, a tool analyzes the calls of your contact center. Each one is set in its vendor's console; the list of these settings, assistant by assistant, is nowhere.
Their questions, in their words:
- Is our assistant's AI disclosure on, and who can turn it off?
- Does our call analysis tool infer our customers' emotions? How do they know?
- We built our own assistant: who is its provider?
Ojja does not answer for you. It gives you the facts to answer.
What Ojja shows you
- Each assistant, on its own line. The assistants and voice agents that talk to your customers, whatever the vendor, and those your teams built on an AI vendor's API. Each with its provenance: read in your tools, declared by you, or inferred.
- The AI disclosure, according to the vendor. Provided or not, on by default or not, can be turned off by your administrator or not, with the address of the page read. As long as the vendor says nothing, the fact sheet says “not documented”.
- Its state at your end. When the vendor opens read access, Ojja reads the setting. Otherwise, you declare it, with its date; the line says where the fact comes from.
- Emotion inference. For call and conversation analysis: whether the feature infers emotions according to the vendor, whose (your customers' or your contact center agents'), from what (voice, face or text). “Not documented” as long as the vendor says nothing.
- The state on a given date. On any date since your first readout: what was active, what the vendor said about it, what you had decided.
- The “Customers and public” view. With Ojja Watch, everything that talks to your customers or publishes on your behalf, in a single view, with the state of the AI disclosure on each line.
The texts the catalog links
- The framework the catalog links to assistants, voice agents and call analysis: the European regulation on artificial intelligence, Article 50, applicable since August 2, 2026, with the points to review with your counsel.
- To call analysis, the catalog also links Article 5(1)(f) of the same regulation, on inferring emotions in the workplace.
Fictional example
A company has two assistants. The first, bought from a vendor, answers on its website: its fact sheet says, according to the vendor, AI disclosure provided, on by default, can be turned off by the administrator; the company declares that the disclosure is on at its end. The second, built by its team on an AI vendor's API, answers in its app: its line is declared, the disclosure is not documented there, and the line shows the point to review on its provider. The company reviews it with its counsel.
What Ojja never does
- Ojja reads no conversation, no message, no recording and no transcript.
- Ojja records what the vendor provides and what is set at your end; it does not say whether that is enough.
- Ojja never writes to your software: it turns nothing on, turns nothing off and configures nothing. Your administrator applies; the next snapshot records that it is done.
- Ojja never concludes: no risk level, no role, no assessment of whether a text is respected.
Ojja does not provide legal advice: it records and documents; how the law applies to your situation is for your counsel to determine.