How readiness is scored
The ten questions cover two things:
- Need (4 questions): enquiry volume, how repetitive questions are, out-of-hours demand and response-time pressure. Low need means a chatbot adds cost and complexity for little gain.
- Readiness (6 questions): written answers to common questions, accurate up-to-date information, clear “do not answer” topics, a handover route to a person, someone to review conversations, and how sensitive the topics are.
Each scores 0–3 and is converted to a percentage per group. The recommendation comes from both: high need and high readiness means go ahead; high need but low readiness means prepare first; low need means a simpler option (better FAQs, a contact form with instant acknowledgement) is probably better. Sensitive topics (health, legal, financial advice) cap the recommendation at “assistive only”.
What chatbots do well, and badly
- Well: opening hours, service areas, prices from a published list, booking links, simple qualification questions, directing people to the right page.
- Badly: anything not in their source material (they may invent an answer), complaints, nuanced quotes, and situations where the customer wants reassurance from a person.
Keep a visible “talk to a person” option at all times, and review a sample of conversations each week. Read what AI chatbots can and cannot do for customer enquiries, and check safeguards with the Automation Risk Assessment.