Escalation Boundary Mapping (Onboarding)

Escalation Boundary Mapping (Onboarding)

1. 01

   Inventory every client inquiry type that will reach the bot before go-live

   Pull 90 days of support tickets, form fills, and chat transcripts, then group them into 8 to 15 recurring intents so the automation scope is grounded in real volume rather than assumptions.

2. 02

   Rank each intent by financial and reputational cost of a wrong answer

   Billing disputes, cancellation requests, and legal or compliance questions sit at the top; hours-of-operation and pricing FAQs sit at the bottom. This ranking decides which intents the bot may close alone.

3. 03

   Define the hard escalation triggers in writing and get client sign-off

   Common triggers include any mention of refund, chargeback, contract terms, or a second consecutive expression of frustration. Chatling and FastBots both expose human handover rules, but the trigger list is a client decision, not a platform default.

4. 04

   Set a response-confidence floor below which the bot defers to a human

   A workable starting point is to route to a person whenever the retrieved answer is weak or the visitor repeats a question. NewOaks AI markets knowledge-grounded responses specifically to reduce hallucinated answers, which is the failure mode this floor guards against.

5. 05

   Name the human owner and the coverage window for each escalation path

   Assign a named person per channel and state the hours they are reachable. An escalation route that terminates in an unmonitored inbox is worse than no bot at all.

6. 06

   Configure the fallback message and capture path for after-hours escalations

   The bot should acknowledge the limit, collect contact details and intent, and promise a specific response window. Social Intents routes conversations into Teams, Slack, or Google Chat, which lets an on-call staffer pick up an after-hours thread without a separate tool.

7. 07

   Run a 20-conversation red-team pass against the escalation rules

   Have someone outside the build try to push the bot past its boundaries using angry tone, ambiguous phrasing, and multi-part questions. Log every miss and tighten the trigger list before launch.

8. 08

   Document the boundary map as a client-facing one-pager and attach it to the retainer scope

   The document states what the bot handles, what it hands off, and what response time the client's team owes. It becomes the reference point when a client later claims the bot was supposed to handle everything.