Conversation Escalation Routing Map (Delivery)

Conversation Escalation Routing Map (Delivery)

1. 01

   Inventory every question the client's team already answers by hand each week

   Pull 30 days of email, contact-form, and phone notes, then tag each inquiry by intent so the routing map reflects real traffic rather than a hypothetical FAQ list.

2. 02

   Split intents into three lanes: bot-resolved, bot-assisted, and human-only

   Pricing exceptions, legal terms, and account-specific status belong in the human-only lane; product specs and hours belong in bot-resolved. A tool like SiteGPT can escalate to a live agent when a thread crosses the line, but the lane definitions must exist before configuration starts.

3. 03

   Define the handoff trigger conditions in writing before touching the widget

   Name the exact signals (visitor types a refund request, sentiment drops, two failed bot answers in a row) and the destination for each: shared inbox, Slack channel, or on-call phone.

4. 04

   Assign named owners and response-time targets per lane

   A retainer client with no staffed inbox will lose leads the bot captures. Confirm who replies within business hours and what happens on nights and weekends.

5. 05

   Configure the widget to collect context before escalation fires

   Pass visitor name, page path, and the last three messages into the handoff so the human agent opens with context. Platforms such as Havlo push a notification with full thread history, which removes the 'can you repeat that' opening.

6. 06

   Test each escalation path with a live visitor session, not a preview mode

   Send a real message from an incognito browser and confirm the notification lands, the transcript is complete, and the reply reaches the visitor in the widget.

7. 07

   Document the routing map as a one-page client artifact and store it with the delivery folder

   The map becomes the reference when the client adds a product line or changes staffing, and it protects the agency when scope questions arise later.