# AI agents, beyond the demo.

Source: https://www.getwilco.ai/guides

How to deploy agents, measure their value and handle the failures that a demo won’t reveal.

- [Deploy the workflow, not just the agent.](https://www.getwilco.ai/guides/build-vs-buy-healthcare-ai-agents): Where agent deployments create value, why they break and how to design the operating system around the model.

- [Before buying more leads, price the return visit.](https://www.getwilco.ai/guides/patient-reactivation-economics): How to compare patient reactivation with acquisition using incremental attended visits, full campaign costs and a credible baseline.

- [WhatsApp is a patient’s inbox. Treat it that way.](https://www.getwilco.ai/guides/whatsapp-patient-trust): How clinics can use WhatsApp without spending patient trust: relevant contact, useful replies, clear exits and respect for channel choice.

- [Human handoff is part of the product.](https://www.getwilco.ai/guides/human-in-the-loop-healthcare-ai): How to design human-in-the-loop care that respects patient choice, transfers useful context and turns resolved exceptions into better agent behaviour.

- [A good call is not a passing test.](https://www.getwilco.ai/guides/evaluating-healthcare-ai-agents): How to turn agent evaluations into release decisions: verified task completion, failure severity, clinic-specific rules and regression after changes.

- [What should happen after a patient says, ‘I’ll think about it’?](https://www.getwilco.ai/guides/following-up-pending-treatment-plans): A practical framework for following up pending treatment plans: understand the barrier, agree the next action and record the patient’s decision.