AI Agents · Guided demo
Watch an agent work, step by step.
Five scenarios, each showing the conversation, the systems the agent touches and the checks it makes along the way, including where it stops and asks a person.
Guided interactive simulation A scripted walkthrough with fictional data. No live AI is running and nothing is sent. We would rather show the design honestly than fake a live system.
How to read it
Three panels, one request.
Connected tools and data
The systems the agent may use, with what it can do in each. They light up as they are used, and the data it retrieves appears underneath.
The conversation
What the person asked and what the agent answered. When an action is proposed, a person approves or declines it here.
Execution events
The sequence of checks and steps: who is asking, which tool was chosen, what came back, and what happened next.
Written example: clinic appointments
Front desk (WhatsApp): “Which appointments are pending confirmation tomorrow?”
Agent checks the sender’s role, reads tomorrow’s pending appointments through an approved read-only tool, and replies with five appointments by time and visit type, using initials only.
Agent proposes sending the approved reminder template to the five patients. Nothing is sent until a staff member approves.
How agents are engineered Design your own agent Laboratory copilot walkthrough