Several employees call from the same company switchboard. An AI intake workflow treats that number as one person and writes each conversation onto the first matching HubSpot contact.
A shared line can identify a possible organisation without identifying the individual caller or the relevant business record.
Model organisation recognition separately from personal identity
Decide whether the workflow needs a company association, a specific contact, or only a new enquiry awaiting review. The answer depends on the action, not on which record is easiest to find.
| Evidence | What it may support |
|---|---|
| Shared company number | A possible company relationship |
| Caller-supplied name | A claim requiring the appropriate matching policy |
| Verified reference | A more specific permitted workflow lookup |
| Multiple plausible contacts | Clarification or staff review |
Do not automatically merge employees because the phone field matches. The same number may be legitimate on several records.
Ask for the minimum useful clarification
For a fictional supplier account, the receptionist might ask who is calling and what the enquiry concerns. It should not list private contact names or deal details to help an unverified caller select an identity.
If the person cannot be resolved under the approved policy, capture a company-level request or use a review queue. Avoid fabricating a personal email merely to satisfy a required contact field.
Keep the selected context attached to the write
Store whether the call was matched to a company, verified contact, or unresolved enquiry. A summary should preserve that uncertainty rather than invent a confident personal association.
If staff later resolve the identity, update the association with a traceable reason. Do not create another copy of the call simply to move it to the correct record.
The deal-and-ticket guide covers another decision that remains after contact identity is known.
Test legitimate repeats without rewarding duplicates
Use fictional calls from two employees on one number, the same employee twice, a transferred call with changed caller ID, and an unknown caller claiming a known company.
Inspect contact count, company associations, call notes, and any information exposed in the conversation. The right outcome may intentionally contain fewer personal details than a model could infer.
The phone-normalisation guide addresses formatting. Normalising a shared number does not make it unique to one person.
Shared-line questions
Should a shared number be removed from duplicate contacts?
Review the data model and business use before treating it as an error.
Is company recognition permission to disclose account details?
No. Apply the relevant authorisation policy for the requested information.
What should the pilot include?
Test several callers on one line and an unresolved identity in the CRM acceptance cases.