Frequently asked questions
What's the difference between an AI agent and a chatbot?
A chatbot responds inside a conversation. An agent acts on your tools and data, calling APIs, writing to systems, and executing multi-step tasks. The chatbot is contained; the agent operates on real systems.
How do you measure if an AI agent is working?
Same way you measure a junior employee: output quality, error rate, escalation rate to humans, and cycle time on the task it owns. Tasks completed is a vanity metric without quality measurement.
What tasks should an AI agent NOT do?
Anything that requires reading a customer's emotional state, building trust, or making a judgment call with no clear precedent. Also anything regulatory (compliance, legal review) where being wrong is more expensive than being slow.
How much oversight does a deployed AI agent need?
Heavy oversight for the first 30 to 60 days, sampled review thereafter. Build in human-in-the-loop checkpoints at high-stakes steps (sending a customer email, updating a closed-won record).
How does Landbase use AI agents?
Internally, agents enrich accounts, score signals, and prepare SDR queues. For customers, the Landbase CLI lets teams script agentic workflows using Landbase data, while the web app handles the human-supervised loop.
.png)
.png)