Decide What Your Support Bot Should and Shouldn't Handle

You are a support automation strategist. Help me decide which contact reasons our chatbot or AI agent should handle fully, which it should start and then hand off, and which should go straight to a human.

Our contact reasons with rough volumes, from our own data: {{CONTACT_REASONS}}
What the bot can technically do (answer from articles only, look up orders, change settings, issue refunds): {{BOT_CAPABILITIES}}
Our risks and constraints (regulated industry, vulnerable customers, high-value accounts): {{CONSTRAINTS}}

For each contact reason, give:
1. A recommendation: automate fully, bot-assisted with handoff, or human only.
2. The reasoning, against four tests: is the answer the same for everyone, can the bot get the information it needs, what does a wrong answer cost, and does the customer need empathy or judgement?
3. What the bot would need to make it work (content, integration, permissions).
4. The handoff point, if any.

Then provide:
- a table summarising the plan, ordered by volume from my data;
- a pilot plan: the first two or three reasons to automate, and how to tell whether it's helping customers or just deflecting them (repeat contacts, handoff rate, CSAT on bot conversations);
- the contact reasons where automation would likely damage trust.

Rules: use only the volumes I gave you, and don't estimate cost savings.