I'm currently developing our own in-house AI agents now with exactly this in mind.
These are the type of customizations that most out of the box AI solutions don't currently allow and why I'm building our own:
- Knowledge Base - Train the AI on every detail of your business - hours of operation, FAQs, approved word tracks, objection handling, etc.
- Guardrails - Clearly define what the AI should never say or offer (e.g., no discounts, no free vehicle delivery).
- Messaging Style - Specify the desired response style (formal, casual, under 160 characters, etc.).
- Agents - Create specialized agents for different use cases, lead sources, campaigns, etc (e.g., lease turn-ins, vehicle acquisition, warranty recalls, etc.).
- Follow-up Cadence - Set a follow-up schedule (e.g., send SMS after 4 hours, initiate an outbound AI voice call after 1 day).
- Information Gathering - Specify data points that you'd like the AI agent to try to collect from the customer (e.g., employer, current vehicle, etc)
- External API Calls - Allow the AI to access your vehicle inventory, schedule appointments, and more.
- Human Handoff - Set conditions for transferring a conversation to a human agent or creating a CRM lead.
- Context - Update the AI with any new activity in the CRM (e.g., email replies) so it knows the customer’s current status.
- Enable / Disable - Give human agents the ability to turn AI agents on or off, and set rules for when a customer should exit the AI workflow.
AI chatbots are relatively easy to create because they rely on "live" conversations.
However, things become significantly more complex when you’re dealing with the more common async communication across multiple channels (text, email, and phone) where the customer isn’t actively responding in real time.
