# Your first AI workflow — free clinic brief

AI-native training by LockedIn Labs · Edition 1, 28 October 2026

A 90-minute facilitator plan for professionals and new builders. This is a runnable practice plan, not a dated live event. Reading, setup and breaks are separate. Use the free [visual systems lesson](https://training.lockedinlabs.ai/training/labs/agent-systems); no account, model call or private business information is needed for its authored browser replay.

## Before the session

Choose a fictional administrative request. Agree who is facilitating and who will receive another person's design note. A real workflow may be discussed only at a permitted, non-sensitive level; use the supplied synthetic case for execution. Have each person state an initial explanation of chat, a workflow and an agent before teaching. Keep that baseline separate from later revisions.

## Run the clinic

| Planned minutes | Activity | Evidence to retain |
| --- | --- | --- |
| 0–15 | State the user task, desired outcome and who may decide or act. | Initial workflow sketch and assumptions. |
| 15–35 | Follow the same request through chat, a workflow and an agent. Identify which actor chooses each next step. | One trace with model, application, tool and person marked. |
| 35–55 | Change the evidence from available to missing. Predict before replaying. | Original prediction, changed trace and actual stop or continuation. |
| 55–75 | Choose the simplest design for a new requirement. Exchange design notes. | Each person's design note, own reasoning and help received. |
| 75–90 | Receiver attempts to explain the design without author takeover. Agree a bounded next build. | Receiving action, missing evidence and accountable next role. |

The browser lesson is an authored explanation, not a live model or deployed workflow. If learners proceed to the [first LangChain build](https://training.lockedinlabs.ai/training/guides/agent-engineering#0b-your-first-langchain-agent), prepare its Python environment and pinned packages separately. Its scripted model does not establish the behavior of a real provider.

## Individual record

```text
Synthetic alias / role:
My starting explanation:
Original request / selected design / expected result:
Who chose each next step:
Changed evidence / prediction before replay:
What I observed and where:
My contribution / assistance:
Receiver's actual explanation / author takeover, if any:
What I changed after receiving feedback:
Next bounded build / accountable role:
```

Close by distinguishing an explanation, a local execution and evidence from real operation. Keep unresolved access, data and deployment questions visible. The clinic does not establish engineering independence, qualification or client placement.