Skip to content
Contacts & leads

Understand automatically discovered contact details

See how names, emails and phone numbers mentioned naturally in a conversation become contact context.

Where to find itConversation → contact/context panel
In this article

Before you begin

Open Contacts or the contact/context panel beside a conversation. Check the linked conversation and the provenance of each value before using it.

Example workspace · Demo data. Inspect the original message and recognised details before relying on them. Select the image to enlarge it.
Contact context beside an example conversation showing unverified details

Example workspace · Demo data. Inspect the original message and recognised details before relying on them. Select the image to enlarge it.

Step by step

  1. 1

    Enable discovery intentionally

    In the agent’s contact discovery settings, choose whether relevant details may be extracted and whether a natural name question is appropriate.

  2. 2

    Let the conversation supply context

    A visitor can introduce themselves or provide contact details while asking for help. The agent should avoid repeatedly asking for information already supplied.

  3. 3

    Inspect the provenance

    Open the contact panel and check the detail, its source message and the subtle unverified styling. A quoted third-party email should not automatically be treated as the visitor’s identity.

  4. 4

    Correct uncertain details

    Use the available review/correction controls and respect a visitor’s request not to collect a detail. Investigate conflicting evidence before using it.

Collect useful context without turning support into an interrogation

Use volunteered details when they clearly describe the visitor. A sentence such as “I’m Jamie and my email is jamie@example.test” has a different meaning from a visitor quoting a supplier’s email or describing another person. Inspect the original message when the association is uncertain.

An optional natural name question should help the interaction, not block the answer. Avoid asking again when the person already introduced themselves, declined or supplied the relevant information through a form. Review overlapping prompts and workflows if the same detail is requested repeatedly.

Keep confidence visible. Discovered information is unverified; a staff review confirms the interpretation, not control of an address or phone. A valid email format and a submitted contact form do not independently verify identity. Use the supported verification or trusted account process before sensitive actions.

When a visitor corrects a detail or asks not to remember it, review the current record and follow the supported correction/opt-out controls. Do not copy an incorrect value into another source or workflow just to keep it available.

Good to know

Discovery is an inference, not identity verification. Italic/unverified presentation helps staff distinguish discovered values from verified evidence.

Do not use an email mentioned in chat as proof of ownership of an order or account. Follow your business verification process.

Check the result

Confirm that the value belongs to the right conversation and that its verification/provenance label is accurate. A name or email is not proof of identity.

Questions & troubleshooting

The extracted name is wrong

Inspect the exact source message, reject or correct the value and check whether it referred to someone else.

The agent keeps asking for a name

Review ask-name settings, prompt and overlapping workflow forms. Ask only when useful and avoid duplicate collection paths.

Was this article helpful?

Rate this guide on this device. Ratings are not sent to Corthex.